最新刊期
卷 27 ， 期 3 ， 2023
-
摘要：GF-1—GF-7 satellite series with 19 major payloads has been launched with the continuous implementation of the high-resolution Earth Observation System (referred to as GF) in the past decade. This progress is vital in forming the multispectral and multimode observation capability of China’s Earth Observation System. Remote sensing data with high spatial, temporal, and spectral resolution have been obtained and widely used in scientific research and remote sensing applications. However, obtaining high-quality remote sensing information products from the original satellite data is a complicated scientific issue and faces huge challenges. Hence, the conversion chain from GF data to information must be urgently set up to reduce the remote sensing application threshold and improve the effectiveness of application services.The errors of remote sensing quantitative products are determined by accumulating a series of errors, such as sensor imaging error, calibration error, remote sensing data processing error, and quantitative inversion error. Thus, improving the accuracy of quantitative remote sensing products is a complex system engineering. Completing the whole process, including data processing, retrieval algorithm development, product generation, and validation independently, is challenging. Remote sensing algorithm test and product validation are the two crucial ways for the quality improvement of remote sensing products. Hence, this study proposes the technique system of GF common product generation and validation to improve the quality of GF remote sensing products further, thereby guaranteeing the improvement of the application quality and the extensive application area of GF remote sensing products. Lastly, the current progress of the GF common product validation and algorithm determination system platform is introduced and discussed.GF common products are required by more than two thematic remote sensing products. They can be validated using in situ observations. According to the GF common product system, the number of 39 + 6 products in seven categories are sorted out for the common requirements of multiple users, including geometric products, basic radiation products, land cover and land type products, energy balance products, vegetation products, water products, and atmosphere products. This study presents the technique flowchart of GF common product algorithm determination and product generation. The key technologies of algorithm testing, algorithm optimization, product generation, and validation are developed. Eleven national standards for remote sensing product validation are issued and implemented. Other group standards, such as GF common product generation, ground in situ observation, and validation of GF common remote sensing products, are being designed and compiled. Based on these validation technologies and the in situ data from the national network of GF remote sensing product validation field sites, the GF common product validation platform and product algorithm determination system platform can ensure the high quality of GF common products.Building such a technical system for GF common product generation and validation has great relevance for ensuring high accuracy and high quality to improve the efficiency of application services further. It requires the cooperation of multiple researchers from different units to research and develop common product retrieval algorithms. Moreover, the algorithm should be continuously tested to improve the accuracy of common products.关键词：GF satellite;remote sensing retrieval;common product;algorithm test;ground truth;product validation- 1608
- |
- 939
- |
- 10
发布时间：2023-04-18 -
摘要：The major special deployment of the High Resolution Earth observation system is to build a national validation system for common products. The purpose is to form the ability to verify the authenticity of high resolution common products, build the mass production capacity of common products, and improve the overall level of quantitative application of high resolution remote sensing in China. Therefore, how to define common products, what are common products, what are the generating relationships between common products, and between common products and thematic products is a problem that must be solved first. Focusing on these problems, this paper proposes the overall architecture of the national authenticity inspection system for high resolution special projects, that is, the bottom layer is the authenticity inspection station network observation subsystem, building a authenticity inspection station network composed of 42 stations, obtaining the actual authenticity inspection data through sensors, unmanned aerial vehicles and manual measurement, and collecting them in the system database of the middle layer in real time. The middle layer is a subsystem of authenticity inspection and finalization of common products, with integrated data management and sharing, authenticity inspection station network management, authenticity inspection of common products, algorithm model integration testing and finalization of common products and other functions. The top layer is the national remote sensing data and application service platform, which is the portal for authenticity testing. This paper defines the category and grading method of high resolution common products, that is, the standard products are divided into 0—2 levels, which refer to the original image data generated by the special ground system for high resolution, or the image data after system geometric correction and/or radiometric correction; Common products are divided into 3 to 5 levels, which refers to the data obtained by the high resolution special application system using quantitative remote sensing feature parameter inversion method based on standard products as the production input of at least two industry specific products; Thematic products are classified into 6—7 levels, which refer to the data supporting the business application of the industry and department obtained by professional users using quantitative remote sensing characteristic parameter inversion method or multi-source data superposition analysis method based on standard products or generic products. The paper discusses the construction process of the generic product system. Taking agriculture and transportation as examples, it analyzes and deduces the demand for generic products in the production of industry specific products. Furthermore, it deduces the derivative relationship between generic products in the form of product tree, and clarifies the inspection object of the authenticity inspection system and the product catalog that should be focused on, which has important guiding significance for the implementation of the system. After the completion and operation of the national authenticity inspection system, in a sense, it will become a “challenge arena” for common product algorithms. Any algorithm that has been proved to be better through algorithm evaluation will be integrated into the system to replace the original algorithm. These algorithms will also further update the generation process of existing common products iteratively, which will promote each other.关键词：GF satellite;common product;thematic product;product system;product classification;authenticity inspection system- 1041
- |
- 543
- |
- 0
发布时间：2023-04-18 -
摘要：Remote sensing experiments and products validation have always played an important role in the development of remote sensing science and technology. As an important way of understanding the natural characteristics, remote sensing experiments have played an important role in remote sensing science and technology research. Remote sensing product validation is crucial to ensuring the high accuracy of remote sensing algorithms and improving the quality of remote sensing products. Remote sensing experiments are the basis of validation because the ground truth, which is derived from field experiments, is the core of the validation. However, obtaining the ground truth is a complex systematic process, including optimized sampling, field observation, scale transformation, and validation. Thus, reducing the error of each process is important in obtaining high-quality ground truth.On April 6, 2022, China held the “First Forum on Remote Sensing Field Experiments and Products Validation” where intense exchanges and discussions on the key topics occurred, including remote sensing experiments and validation theory and technology, ground field observation and uncertainty, validation practice and platform, and the future development of remote sensing experiments and validation. More than 1000 scholars and students from different domestic scientific research institutes and universities attended the event.On the basis of reviewing the main progress in the field of remote sensing experiments and validation in China, thus paper summarizes the opinions and suggestions of the Chinese scientists who attended the meeting and expounds on the problems and challenges faced by China in terms of remote sensing field experiments and products validation. The development process of remote sensing experiments and products validation in China has three main development stages, namely, the remote sensing application experiment stage from the year of 1978, the quantitative remote sensing mechanism experiment stage from the year of 1994, and the comprehensive experiment stage of the quantitative remote sensing mechanism and validation. Significant progress has been made in experimental technology, validation technology, national specifications of product validation, and the validation system platform.Five related topics of remote sensing experiments and product validation have been discussed by 80 scientists, namely, the theoretical framework and basic theory of remote sensing product validation, ground observation and standardized data processing, land surface heterogeneity and scale transformation, product validation and evaluation, and system platform and data sharing.It should be considered systematically from the entire link key process of obtaining ground truth for remote sensing experiment and validation to truly become an important basis in ensuring the quality of satellite remote sensing products. Finally, the prospect of remote sensing experiment and validation in China is discussed. Remote sensing experiment and validation should be implemented from the joint efforts of national and even global scientific researchers to achieve global implementation through international cooperation and improve the ability of remote sensing experiments and product validation.关键词：remote sensing experiment;product validation;optimize sampling;scale transformation;ground truth- 864
- |
- 531
- |
- 0
发布时间：2023-04-18 -
摘要：In the context of remote sensing research and application, complete and reliable “ground a priori knowledge” datasets play an essential role in physical-based model construction, land surface parameter inversion, and remote sensing product production and validation. Ill-posed inversion problems, such as the case in which the observation information is less than the inversion target parameters that results in underdetermined inversion parameters, lead to uncertainty in the solution. A priori knowledge is an important support to solving the ill-posed problem of parameter inversion based on physical and empirical models. However, its completeness, accuracy, and timeliness are limited. The traditional methods of obtaining ground a priori knowledge include experimental measurements using various surface/near-surface sensors and numerical simulations using many physical models, such as one- or three-dimensional radiative transfer models. These current methods have their own advantages and disadvantages but cannot meet the need of comprehensive dataset production in spectral, temporal, angular, and spatial aspects for supporting the research and development of remote sensing science and technology when used alone. On the basis of the studies on experimental measurement, modeling of radiative transfer and ecological processes, and land surface parameter inversion and validation, we propose an innovative strategy to support remote sensing research by building a digital twin of the remote sensing experimental field. Several steps are designed for generating remote sensing a priori knowledge on the basis of the digital twin of the remote sensing experimental field. The three-dimensional structure of a scene is digitally reproduced from the surface by multiple experimental measurements of structural descriptors or the near-surface by remotely obtained data, such as the high-resolution visible and near-infrared images and light detection and ranging (lidar) point-clouds from the observation on Unmanned Aerial Vehicle (UAV) or other platforms based on a cooperative observation technology, recorded in a format accessible by simulation models. The systemic evolution of the surface simulations of physical processes can be realized by coupling radiative transfer, energy balance, evapotranspiration, and plant growth modeling theories as a synthesized model and applying the model to an experimental site in virtual space to illuminate and realize the dynamic progression of the remote sensing experiment field. Driven and constrained by the surface/near-surface collaborative observation data processed by data science and statistical methods, such as data fusion and data augmentation, the synthesized model is optimized by the feedback from the data assimilation of observation measurements and corresponding simulation data, increasing the consistency of the simulation results with the actual dynamic evolution of the remote sensing experimental field in the real world. Through the optimized model and the field measurements, a complete and coherent a priori knowledge of the remote sensing experimental field is achieved with high numerical precision and temporal continuity, supporting the development of remote sensing mechanism model construction and remote sensing inversion method and validation and improving the level of basic remote sensing research. The conditions for the development and application of digital twins in remote sensing experimental sites are gradually maturing. The construction of the remote sensing experimental field digital twin is expected to become the rudiment of digital twin construction theory of a small-scale ecosystem, which may in turn promote the comprehensive and collaborative development of various disciplines in geoscience.关键词：remote sensing experiment field;radiative transfer;computer simulation;data assimilation;digital twin;comprehensive dataset- 1011
- |
- 756
- |
- 0
发布时间：2023-04-18
Review
-
摘要：The on-orbit absolute radiometric calibration of satellite remote sensing payloads is crucial to ensuring the radiometric consistency of products from different sources and improving the inversion accuracy of high spatial resolution products. The traditional site calibration method based on manual synchronous measurement is limited by the imaging, surface and atmospheric measurement conditions at the time of satellite overpassing, which is difficult to meet the needs of correcting the radiometric performance change accurately during the life cycle. In this paper, according to the characteristics of large viewing angle observation of the wide width sensor onboard Gaofen6 satellite (GF-6/WFV), a time series absolute radiometric calibration method for the GF-6/WFV sensor with automatic radiometric calibration site is proposed. The proposed method not only can track the trend of satellite payload radiation performance by significantly increasing the frequency of on-orbit calibration, but also can give the calibration results with uncertainties. Based on proposed method, 41 of frequencies automatic radiometric calibration results of GF-6/WFV sensor were calculated from August 2019 to October 2021, and the traceability uncertainty of the calibration results was better than 5% (except for the 7th band). The trend analysis of the time series calibration results shows that the annual variation of the radiometric calibration coefficients of the GF-6/WFV sensor are from 1.10% to 4.59%. In addition, the high-precision in-orbit satellite payloads MODIS and Sentinel-2/MSI are used as reference to cross validate the GF-6/WFV sensor’s calibration results. And, the averaged absolute relative differences of the cross validation were between 2.12% to 6.09% respect to the bands 1 to 4 of GF-6/WFV sensor. The proposed method in this paper can realize the time-series on-orbit absolute radiometric calibration of the GF-6/WFV sensor and its radiometric performance trend change monitoring during the whole life cycle, which can effectively support the production of high spatial resolution remote sensing products and improve the quantitative application level of Chinese remote sensing data.关键词：time-series radiometric calibration;GF-6/WFV;uncertainty analysis;trend analysis;cross calibration- 558
- |
- 316
- |
- 1
发布时间：2023-04-18 -
摘要：Affected by imaging conditions, data transmission, and other factors, stripe noise is common in satellite remote sensing images. It seriously restricts the quality and further use of images. In early studies, various denoising methods, such as statistics-based methods, filtering-based methods, and optimization-based methods, have been proposed to overcome the above problems. These proposed methods have achieved inspiring results in some aspects. However, they still suffer from poor adaptability, low denoising efficiency, and the need for prior knowledge. Therefore, stripe noise removal remains a challenging task.In this study, we take advantage of the convolutional deep network while considering the characteristics of the stripe noise image itself. A deep-learning-based method is proposed, which includes three parts: a feature extraction module, a feature fusion module, and a stripe denoising module. The feature extraction module uses the convolutional layer of the same channel with different strides to extract features. As a result, different-scale feature maps of the noisy image are obtained for the following feature fusion module. The feature fusion module upsamples different-scale feature maps. It fuses these upsampled feature maps through the element-wise addition method. Finally, a denoising network is used to predict the components of stripe noise. The stripe component is subtracted from the noise image based on predictions. Given the difficulties in obtaining real noise samples, the network is trained by simulation samples. Then, it is extended to denoise real images.Experiments on simulation and real images show the excellent performance of our network. In the quantitative assessment, the PSNR and the SSIM of our network when simulated images are used are higher than those of the four methods. In the visual assessment, our network performs well on homogeneous and nonhomogeneous objects. Our network denoises more efficiently and retains more details of ground features than traditional methods and other denoising networks. In real noise images, our method achieves the best denoising performance with the highest ICV and the lowest MRD. Compared with traditional methods, our network has a fast denoising speed, approximately 100 times faster than the denoising speed of the optimization-based denoising method. The above experimental results demonstrate that our network has the best denoising performance in simulated and real images.In this study, a convolutional neural network denoising method based on multiscale feature fusion is proposed based on the fully convolutional neural network. The method uses residual learning to predict the strip-noise components on images. It achieves clean images by subtracting the strip components from noisy images. Experiments demonstrate that compared with traditional methods, deep learning denoising methods are adaptive for removing stripe noise of different intensities without losing image details. The strategy of feature fusion and residual learning can effectively improve the training speed and denoising accuracy of the network. In the future, skip-connected and batch normalization layers will be included to optimize training speed and improve denoising performance. Further studies will be conducted in terms of the transfer ability of the network and the extension of its application in other types of remote sensing images, such as aerial images and hyperspectral images.关键词：high-resolution image;deep learning;stripe noise;convolutional neural network;Feature fusion- 1599
- |
- 523
- |
- 1
发布时间：2023-04-18 -
摘要：The existence of a cloud reduces the application value of remote sensing images. Accurate and automatic cloud and cloud shadow detection and labeling for multispectral satellite images is conducive to the subsequent application of remote sensing images. China currently has a large number of high-resolution multispectral satellite images. However, standard data products rarely contain pixel-by-pixel cloud and cloud shadow tag data for quality analysis. Traditional cloud detection algorithms usually require parameters, such as satellite imaging geometry, imaging time, and calibration coefficients. However, many Chinese satellites’ images have lost parameter auxiliary files during multiple product iterations. Moreover, many military application satellite images are missing or do not provide parameter files. Multispectral satellite image cloud and cloud shadow detection with missing parameters requires special research.The present study investigates the cloud and cloud shadow detection method of domestic four-band multispectral satellite imagery with missing related parameters. This algorithm process is based on the classic spectral threshold cloud and cloud shadow detection algorithm. It also uses image processing and morphological algorithms to improve detection accuracy. A morphology-based method for estimating the azimuth and distance of the cloud shadow relative to the cloud area is proposed for the data with missing parameters.The experimental data in this study is from the GF-1 satellite Wide-Field View (WFV) sensor, and the 86 test images are from Dunhuang, Gansu, China. The experimental area contains a large area of a bright surface and snow-capped mountains that are easily misdetected in cloud and cloud shadow detection. The result of the cloud and cloud shadow detection experiment in this study for the case of missing parameters achieves accuracy similar to that achieved by normal algorithms. This study also analyzes the misdetection of the algorithm and clarifies the challenges of subsequent research.In this study, we propose a set of refined cloud and cloud shadow detection algorithms in case of missing parameters for domestic four-band multispectral satellite imagery. The algorithms are based on the classical spectral threshold cloud and cloud shadow detection algorithms. They also use image processing and morphological algorithms to improve accuracy further. Moreover, a morphology-based method for estimating the orientation and distance of cloud shadow relative to the cloud area is proposed for the data with missing parameters. The experimental results of GF-1 WFV data show that the detection results of this algorithm achieve an accuracy similar to that of the widely used MFC algorithm in the case of missing parameters.关键词：cloud detection;cloud shadow detection;GF-1;multispectral satellite image;missing parameters- 2927
- |
- 568
- |
- 1
发布时间：2023-04-18 -
摘要：Orthorectification is the foundation for the subsequent processing, analysis, and application of satellite remote sensing images. However, the accuracy of the autonomous geo-positioning of the optical satellite data cannot reach 1—2 pixels at present. Ground Control Points (GCPs) are still required for correcting the geometric model to generate Digital Orthophoto Map (DOM) products. This study introduces the DOM products of Chinese optical satellites and automated processing algorithms independently developed by China Remote Sensing Satellite Ground Station. Given the current situation, characteristics, and application requirements of the Chinese optical satellite data, a complete and automated DOM generation algorithm has been developed. It consists of several key steps and techniques, including automated GCP collection via accurate image registration, optimization of image geometry model, image orthorectification, pixel-wise geometric uncertainty estimation, pixel-wise solar irradiation and satellite observation angle calculation, and image division and encoding based on a global or regional geographic grid. (1) Given that the conventional affine transformation correction in the image space for the Rational Polynomical Coefficients (RPC) model cannot achieve satisfactory accuracy when applied to perform a geometric model if the swath width of the image is large or the geometric calibration accuracy is insufficient, a full-parameter optimization method of RPC model based on L1-norm constrained least squares (L1LS) is applied to improve the accuracy of the geometric model. (2) A novel block adjustment method that considers the accuracy of GCPs is proposed. When the GCPs obtained from the reference image with the low spatial resolution are used to correct the geometric model, the random error of the geometric model can be reduced through multiple observations under the premise of meeting the geometric constraints of the reference image, thereby improving the geo-positioning accuracy of the image. (3) Based on uncertainty propagation theory, the per-pixel geometric uncertainty is derived from the geometric model, the Digital Elevation Model (DEM) data, and the residual errors of GCPs to enable the geometric accuracy traceability of DOM products. When applied to GF-1 MSS, GF-1 WFV, GF-6 PAN, HJ-2A CCD4, CB-4A MSS, and ZY-1E PAN images, the L1LS-based full-parameter optimization method of the RPC model outperforms the conventional affine transformation correction in the image space for the RPC model, particularly for the images with a large field of view. The proposed block adjustment approach considering the accuracy of GCPs achieves higher accuracy than the ordinary block adjustment method when GCPs are obtained from a reference image with a resolution (10—15 m) lower than that of the target image (2—2.5 m).Conclusion The experimental results show that the algorithm in this study can be used for the large-scale production of orthophoto products. The field-measured checkpoints distributed all over China show that the absolute geometric accuracy of the produced 16 m-resolution DOM products can achieve high accuracy within two pixels.关键词：DOM;block adjustment;L1 norm;Chinese satellite;geometric uncertainty;pixel observation angle;geographic grid- 939
- |
- 514
- |
- 1
发布时间：2023-04-18 -
摘要：A fast development in remote sensing science and technologies has been witnessed in recent years with the launch of various remote sensing satellites and the establishment of numerous industrial application systems. Satellite series, like GaoFen, has pushed the richness of data to a new level. Moreover, quantitative product-driven application systems have become increasingly influential in many disciplines. By contrast, the bridge between data and application, namely, the production capability of quantitative products, seems weak, greatly limiting the usage and influence of GaoFen data. The improvement of production capability comes from multiple aspects, including product hierarchy, algorithm model, and production system. In this study, we propose, from the production system perspective, an integrated system design that responds to the four main challenges of the system: uniform data access, heterogeneous algorithm integration, layered workflow orchestration, and cloud infrastructure adaptation.The system is based on algorithm containerization, where executable algorithms and all their dependencies are encapsulated. Thus, it can run uniformly and consistently on different infrastructures without worrying about the complexity caused by deployment. This scenario helps the system to manage diverse remote sensing algorithms uniformly. We employ the Kubernetes container orchestration platform to automate the execution, scaling, and management of containerized algorithms. A containerized cluster consists of multiple master nodes, many computing nodes, and multiple data centers. Multiple algorithm repositories are constructed to support the system and cope with the high computing and data throughput density of remote sensing algorithms. Each algorithm repository is further divided into several subrepositories to improve load balancing. User-defined role-based access control for algorithms is set up to protect the intellectual properties of algorithm owners. A recommended algorithm image architecture is introduced to standardize algorithm encapsulation. A set of nine properties are abstracted to describe uniformly any data entity parameter of an algorithm. This approach ensures that suitable input data can be found for user-uploaded algorithms to run in the system. For data visualization and quantitative computing scenarios, a multiscenario data organization strategy is proposed to avoid excessive data operations, such as projection transform or subdivision. The business logic of the system, from user order creation to product calculation, is detailed for clear implementation. The production sometimes involves workflow batches. We propose a stratified workflow aggregation strategy to optimize workflow execution.The system has been used for large-scale production of various GF quantitative remote sensing products, including surface reflectance product, normalized difference vegetation index product, leaf area index product, and surface albedo product. These products fully cover China’s area for eight successive years from 2013 to 2020, with quantities of more than five million and storages of nearly 300 TB. The proposed system completes the production task smoothly and efficiently.During the routine support for many large-scale production tasks, each part of the system performed consistently with the system design proposed in this study, demonstrating that the study can help build a stable and efficient quantitative remote sensing production system on cloud-native infrastructures关键词：production system;quantitative product;algorithm integration;container;cloud computing;system architecture;workflow- 2776
- |
- 404
- |
- 1
发布时间：2023-04-18
Gaofen Satellite Data Fine Processing
-
摘要：The rapid development of remote sensing technology has promoted the generation of different vegetation index products. Normalized Difference Vegetation Index (NDVI) is the vegetation index with the highest utilization rate. However, the existing NDVI products have insufficient time resolution and spatial resolution, thereby limiting the fine dynamic monitoring application in a certain region. The Wide-Field View (WFV) of GF-1 satellite data has a 4-day revisit period and 16 m spatial resolution, indicating its great potential in long-time series dynamic monitoring. The objective of this study is to establish a method for generating a 16 m/10-day NDVI product based on GF-1 images from 2018 to 2020.In this study, the GF-1 NDVI products of 16 m and 10 days from 2018 to 2020 are produced based on GF-1 WFV. Moreover, Landsat NDVI and sentinel-2 NDVI products are produced based on landsat7, landsat8, and sentinel-2 data in the Google Earth engine database. The quantitative analysis and evaluation of time, spatial consistency, spatial continuity, and product comparison are performed from the space and time scale.In January and August, the spatial distribution of MuSyQ NDVI, Landsat NDVI, and sentinel-2 NDVI products in China is reasonable and consistent. MuSyQ NDVI’s lack of space in January is lower than that of two other products. The frequency distribution histogram of MuSyQ NDVI, Landsat NDVI, and sentinel-2 NDVI differs. The difference among the three products is concentrated in the range of ±0.2, the peak value is at 0, and the frequency is close to 70%. These findings indicate that MuSyQ NDVI has good spatial consistency with the two other NDVI products. In Northeast China, Northwest China, and Qinghai Tibet Plateau, MuSyQ NDV has a lower loss rate and better spatial continuity than the two other products. Moreover, the spatial continuity of products is high. On the whole, the effective value ratio of the MuSyQ NDVI product is better than that of the two other products; in particular, the effective value ratios of the MuSyQ NDVI product in farmland and grassland areas are 28.6 (70%) and 30.27 (70%), respectively, which are higher than those of the two other NDVI products. In the forest area, the effective value ratio of MuSyQ NDVI is also slightly better than that of the two other products. The three NDVI products have good consistency and phenological characteristics in the time series of farmland and grassland. In the deciduous broad-leaved forest area, the three products have similar seasonal variation laws. They fluctuate greatly in the time series curve in the evergreen broad-leaved forest and evergreen coniferous forest area. The consistency of MuSyQ NDVI, Landsat NDVI, and sentinel-2 NDVI in nonforest sites is higher than that in forest sites.In terms of spatial and temporal scales, the high spatial-temporal resolution NDVI products provided by GF-1/NDVI products are better than the existing products. They also provide useful information for selecting NDVI products in subsequent vegetation dynamic research. Moreover, they have advantages for long-time series fine monitoring in an extensive spatial range.关键词：GF-1;vegetation index;high resolution;spatiotemporal characteristics;cross validation- 846
- |
- 431
- |
- 1
发布时间：2023-04-18 -
摘要：Leaf Area Index (LAI), is a critical variable in models of climate, meteorology, hydrology, and biogeochemistry to characterize vegetation canopy structure. Remote sensing provides a practical approach to estimating dynamic LAI on a large scale and some global LAI products were generated in the past decades. However, these products are mainly focused on the low-medium resolution satellite data and there is no standardized high-resolution LAI product worldwide. The object of this work is to propose an LAI inversion algorithm for high-resolution satellite GF-1 Wide Field View (GF-1 WFV) images and generate the GF-1 LAI product across China.Three-dimensional stochastic radiative transfer (3D-SRT) model, which can take the 3D-canopy architecture into consideration, is a widely-used model in LAI inversion. Parameters of Single Scattering Albedo (SSA) and uncertainty in the 3D-SRT model are highly correlated with the band setting and band stability. To acquire the optimal values of these parameters, 94824 homogenous samples of six vegetation types across China are selected and the characteristics of the difference in their surface reflectance are analyzed. SSA and uncertainty are adjusted to the values when the GF-1 retrieved LAI and MODIS LAI share the most similarity across the homogeneous samples. Based on the 3D-SRT model and the adjusted key parameters, an look up table (LUT) was constructed for the LAI retrieval in this work.There are 359 ground-measured LAI data in Shihezi, Xinjiang, and Sidaoqiao, Inner Mongolia in the validation. The overall result shows compared with the inversion result before adjusting the parameters, the root mean square error (RMSE) of the optimized algorithm can be reduced from 1.209 to 0.804, the determination coefficient (R2) can be increased from 0.659 to 0.883, and the retrieval index (RI) can be increased from 25.3% to 73.8 %, suggesting the higher accuracy and stability of the algorithm and more suitable for GF-1 LAI retrieval. The accuracy and stability of the algorithm also improved for each vegetation type individually. Based on the algorithm, the GF-1 leaf area index product of 16 m/10 days resolution across China from 2018 to 2020 was generated. The temporal profiles extracted from the product can indicate reasonable phenological characteristics of different vegetation types.Based on the algorithm proposed in this study, the high-resolution (16 m /10 days) LAI products for 2018-2020 across China were generated based on domestic satellite GF-1 Wide Field View. It can provide accurate and effective data which supports vegetation change research, agricultural and forestry application, ecological environment monitoring, and government decision-making. However, due to the short revisit time of medium and high-resolution satellites and the cloud contamination, the miss rate of current products is still high. In the future, more works can focus on how to generate spatial and temporal continuous products.关键词：remote sensing;leaf area index;GF-1;three dimensional radiative transfer model- 672
- |
- 335
- |
- 2
发布时间：2023-04-18 -
摘要：Fractional Vegetation Cover (FVC) is a critical parameter for monitoring vegetation growth status. Remote sensing effectively generates FVC at a large scale. However, the spatial resolution of the existing FVC products at the global scale is more than 300 m. The major limitation of the FVC produced from high-spatial-resolution satellite images is the lack of effective observations, mainly due to a small range of view and long periods of revisit time for satellite images with 30 m or higher spatial resolutions. The Chinese GaoFen No. 1 satellite (GF-1) wide-field view data with 16 m spatial resolution and 4-day revisit time provide an available data resource for FVC extraction. The objective of this study is to assess the quality of the 16 m/10-day FVC product based on GF-1 images from 2018 to 2020.The assessment of the GF-1 FVC product was accomplished through direct validation with ground measurements and indirect validation with the GEOV3 FVC product. Two FVC products were postprocessed with the same temporal (month) and spatial (300 m) resolution to compare GF-1 FVC with 16 m/10-day and GEOV3 FVC with 300 m/10-day. A total of 32 ground measurements (including 18 ground measurements for crops and 14 ground measurements for forest) throughout the growing season at the middle reach of Heihe River Basin, the Jingyuetan station, and the Saihanba plantation forestry farm were used to validate the FVC product in China. The indirect validation was evaluated by the spatial and temporal continuity with missing values and the product consistency with GEOV3 FVC.The percentage of the annual missing value lower than 70% accounted for 88% of the main land in China. During the growing season, the percentage of the annual missing value lower than 73.68% approached 82.73%. According to different inversion algorithms and input products, the percentage of the average missing value of forest types (>20%) was higher than that of nonforest types, such as crops and grassland (<10.6%). The GF-1 FVC agreed well with GEOV3 FVC for the nonforest type based on the homogeneous samples in China from January to December 2019. The direct validation results indicated that the accuracy for the FVC product achieved by the GF-1 FVC product is reasonable compared with the ground measurements (R2 = 0.57, root mean square error = 0.12, BIAS = -0.03) in China. Moreover, it is better than the accuracy achieved by the GEOV3 FVC product, particularly for forest type.In conclusion, the GF-1 FVC product of China with 16 m/10-day resolution reflects the seasonal characteristics of vegetation well. Moreover, the GF-1 FVC product with high spatial and temporal resolutions meets the requirements of vegetation monitor at the regional scale.关键词：Fractional Vegetation Cover (FVC);direct validation;indirect validation;China;GF-1- 863
- |
- 356
- |
- 1
发布时间：2023-04-18 -
摘要：The Fraction of absorbed Photosynthetically Active Radiation (FPAR) is one of the key parameters in the light use efficiency model of the carbon cycle. High-spatiotemporal-resolution data have been provided for the inversion of quantitative remote sensing products since the launch of GF satellites. The FPAR products derived from GF satellite data provide precise and accurate input parameters for the analysis and evaluation of the ecosystem’s carbon cycle. In this study, a deep learning algorithm was developed to retrieve FPAR over China based on the simulated data of the radiative transfer model. The inputs are surface reflectance, cloud detection, and land cover products of GF-1 satellite data, whereas the output is FPAR. The FPAR product has a spatial resolution of 16 m and a temporal resolution of 10 days. This method uses the SAIL model to simulate output canopy FPAR and reflectance under various input variables, such as solar and observing angles and atmospheric conditions. The FPAR inversion model of GF-1 satellite data was obtained using a deep belief network. The long-term crop and grassland FPAR observation data in Huailai and Heihe were used to compare and validate the FPAR products, with a root mean square error of 0.15 and 0.17, respectively. The inversed FPAR is in good agreement with the measured FPAR in the low values, but lower than the measured FPAR in the high values. The radiative transfer model, the representativeness of the simulated data, and the preprocessing (calibration and geometric and atmospheric correction) of the GF-1 satellite data inevitably introduce some biases in the inversion process. This method uses the multidimensional atmospheric and surface variables as the input and the simulated vegetation canopy by the radiative transfer model parameters as the output. The simulated dataset, used as the training samples for deep learning, makes up for the errors in the deep learning training process caused by the insufficient number of training samples and incomplete observation data. The input of the inversion is only the surface reflectance product with the information of the sun angle and the observation angle. It lessens the difficulty of obtaining input parameters, reduces the influence of the error transmission of the input parameters, and is conducive to the realization of the commercial production of the product. The high FPAR is mainly distributed in the northeast, north, central, east, southwest, and south parts of China. The interannual variation of the FPAR time series, combined with the vegetation growing cycle and phenology, is high in spring and summer and low in autumn and winter.关键词：quantitative remote sensing inversion;GF-1;FPAR;SAIL model;deep learning- 701
- |
- 307
- |
- 1
发布时间：2023-04-18 -
摘要：The Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) is a key parameter in characterizing the photosynthesis process of vegetation and widely used in many study areas, such as vegetation monitoring, NPP estimation, and global change. Remote sensing is the only way to obtain FAPAR at large scales. Compared with the multispectral instrument, the hyperspectral instrument has an advantage in analyzing the canopy reflectance and absorption on the basis of the high accuracy of the spectrum measurement, which is important in FAPAR retrieval. This study developed a new FAPAR retrieval algorithm for the Chinese GF-5 Visible-shortwave Infrared Advanced Hyperspectral Imager (AHSI) data on the basis of BRDF unified model and the neural network (NNT). The validation was performed in Hulun Buir Xeltala, which is a grassland and farming-pastoral area in Inner Mongolia. First, the simulated GF-5 AHSI reflectance-FAPAR datasets were generated by the BRDF unified model, and the characteristics of the data set were analyzed. Five groups of NNT input bands were selected based on the Optimal Index Factor (OIF) and a new factor OIFR, which was modified by the relevance of the band reflectance and the FAPAR. Different groups of bands were used to build the NNT, and the results were assessed by a test set in the simulation dataset. Finally, the best feature bands and the NNT were selected to generate the FAPAR map of the study area from the GF-5 AHSI image. Validation with in-situ observations was made. Overall, the new factor OIFR is more efficient than the origin factor OIF in band selection. As the amount of input bands increases, the NNT accuracy gradually increases, but the trend stops when the amount reaches a certain level. Considering both band information and instrument noise, 8 bands were selected as the feature bands of FAPAR retrieval with the FAPAR RMSE of NNT is 0.014. The FAPAR map of the study area was generated, and the comparison with in-situ FAPAR shows the applicability of the method with RMSE=0.048. The reflectance and absorption by the hyperspectral data when the NNT reduced the middle term and parameters of the traditional methods simultaneously can be analyzed, presenting a new approach to the surface parameters retrieval of domestic satellite hyperspectral instruments.关键词：FAPAR;hyperspectral remote sensing;Feature band;neural network;GF-5- 904
- |
- 325
- |
- 0
发布时间：2023-04-18
Vegetaion Variable Retrieval And Products
-
摘要：Land surface albedo is an important variable for controlling Earth radiation budget. It is also recognized as an Essential Climate Variable by the Global Climate Observing System. The progress of the high-spatial-resolution satellite development allows high-spatial-resolution albedo to provide important data for the research of local radiation and energy balance, regional climate, and ecological environment. However, albedo is a variable related to solar angle, wavelength, and atmospheric status. Thus, the estimation of high-spatial-resolution land surface albedo becomes challenging. Different methods of land surface albedo remote sensing estimation have been developed in the last two decades. These methods greatly improve the high-resolution albedo mapping ability. Summarizing and analyzing the proposed methods of high-spatial-resolution albedo are important to improve the accuracy of its product estimation further.Two fundamental problems of insufficient multiangle observation and multisensor band information for high-resolution albedo estimation are proposed by analyzing the basic principle of the existing albedo estimation. Four main methods of high-spatial-resolution albedo are summarized in the characteristics of the algorithm and application cases according to how the problem of insufficient understanding of land surface reflectance anisotropy is overcome. Lastly, the conclusion and prospect of the high-spatial-resolution albedo method development are also summarized.According to whether the land surface anisotropic reflection characteristics are considered, the current high-resolution albedo retrieval methods are divided into two basic types: the method of narrowband to broadband conversions and the method of Bidirectional Reflectance Distribution Functionconsideration. The latter type considers the surface anisotropic reflection characteristics in different ways. It also includes estimation based on high-resolution multiangle reflectance data, estimation based on combining high-resolution reflectance data with low-resolution reflectance data, and estimation based on empirical knowledge of BRDF/albedo.These proposed methods of obtaining high-spatial-resolution land surface albedo with the surface bidirectional reflection characteristic information directly or indirectly are still the mainstream idea. They alleviate the problem of missing high-resolution valid data to a certain extent. However, they are still limited because of the lack of effective data, such as the angle and band of the high-resolution remote sensing data and the lack of high-resolution BRDF a priori knowledge information. The development of a high-resolution albedo algorithm is a prospect for future research. It can potentially provide theoretical support for high-resolution land surface albedo product development.关键词：land surface albedo;Bidirectional Reflectance Distribution Function (BRDF);remote sensing retrieval;narrowband to broadband;high spatial resolution satellite- 719
- |
- 362
- |
- 0
发布时间：2023-04-18 -
摘要：Land surface albedo is a critical parameter in radiation and energy budget. Using GF satellite data to produce the land surface albedo is beneficial for local-scale environmental monitoring.The challenge beneath the albedo estimation from the GF satellite data is the inadequate angular information, which complicates the BRDF inversion and the albedo derivation based on BRDF. We use the high spatial-and-temporal-resolution priori-knowledge BRDF database obtained from coarse spatial resolution multiangular information to help describe the GF BRDF features. Then, the GF albedo is estimated from the derived GF BRDF. The algorithm is applied in GF-1 data to generate the land surface albedo product in China. First, this algorithm and the production are introduced briefly. Then, the spatial-temporal features are evaluated by qualitative analysis and quantitative validation. In the validation, a long time series of field-measured albedo from the sites of different land covers is used. It includes the cropland (maize) in the Daman site from the Heihe remote sensing test site located in the northwest of China, the forest (Chaenomeles, metasequoia, Chinese pine) in the Huailai test remote sensing site located in the north of China, and the grass in the Dongbei remote sensing test site located in the northeast of China. These sites would be covered by bare soil or snow in winter. In this study, we preliminarily evaluate the algorithm feasibility in albedo production and product precision.The time series comparisons of the field measurements and the GF product present good agreement for different sites over 1 year to 2 years. For over 1-or 2-year time-continuous comparisons, the land surface status is changed driven by the phenology (vegetation growing cycle). Thus, this finding reveals the feasibility of the algorithm based on the spatial-temporal distributed BRDF a priori knowledge. The total root mean square error is 0.05, with a relative accuracy of 80.24%, which meets the application requirement.Therefore, this algorithm is feasible for the albedo estimation from the GF data. The validation results show a good agreement between the field measurements and the product. However, the remote sensing common products from GF satellite data have not been initiated long enough. Thus, evaluating the GF satellite data is still needed. Besides the algorithm itself, the albedo accuracy is directly affected by the land surface reflectance product’s precision, which may introduce uncertainties from the geometric and radiometric calibration, atmospheric correction, and even cloud production. Therefore, much work should be done to evaluate this product and clarify the effects of those factors. In this way, the algorithm and albedo production can be improved.关键词：GF-1;albedo;BRDF;priori-knowledge;GF remote sensing common products production- 541
- |
- 373
- |
- 2
发布时间：2023-04-18 -
摘要：Photosynthetically Active Radiation (PAR) is an important input of vegetation productivity models, and also a key parameter of terrestrial ecosystem models and biogeochemical models. The accuracy and availability of current global or regional products are still insufficient to better understanding Earth system. China has launched series of Gaofen (GF) Earth observation satellites and provide the possibility of high spatial resolution PAR products retrieval. In this paper, a new method based on a parametric model for remote sensing inversion of PAR was proposed. The surface albedo products and the aerosol optical depth products were retrieved from GF-1 satellites, the cloud optical thickness products were retrieved from Himawari-8 and FY-4 satellites. Under clear sky conditions, the attenuation of PAR by aerosol and Rayleigh scattering is mainly considered. The influence of cloud on incident radiation is mainly considered for cloudy skies, and the calculation is based on the Mie scattering theory of spherical particles. Surface received direct PAR for rugged surfaces were calculated with the input of incident angle, the slope and the aspect. The enhancement or attenuation effect of the scattered radiation were retrieved using the sky view factor the horizontal surface received diffuse radiation. The PAR products were compared and verified using the continuous observation data of the ground stations collected from the Hebei Huailai, the Heihe River Basin Surface Process Comprehensive Observation Network and Ganyansuo in Chengdu. The correlation coefficient, mean bias error and the root mean square error between the two datasets were 0.87, 1.56 W/m2 and 16.14 W/m2, respectively. The spatial resolution of the input atmospheric parameters (sub-satellite point 1 km) and the surface parameter resolution (16 m) of the PAR products have a large spatial scale difference. The atmospheric parameters with the resolution of 50 m provided by the GF-4 satellite will be used to further improve the spatiotemporal accuracy of GF PAR products, and more extensive and in-depth verification analysis will be carried out in our future study.关键词：GF-1 satellite;FY-4;Himawari-8;PAR;parameterized model;quantitative remote sensing products- 571
- |
- 323
- |
- 1
发布时间：2023-04-18 -
摘要：Evapotranspiration (ET) is one of the core variables for studying the surface water cycle and management of the field-scale water resources. Satellite remote sensing are widely adopted to obtain the variation of evapotranspiration at large spatial scale, but the resolution of existing ET products is mainly limited at low or medium resolution (1—25 km), which cannot satisfy the field-scale water management application. The Chinese GF-1 Wide Field of View (WFV) camera has the characteristics of high spatial and temporal resolution, with a spatial resolution of 16m, and it can support to generate ET products with high spatial and temporal resolution, which has not yet been well presented. The objective of this study is to present the capacity of using the remote sensing data from the GF-1 satellite as the driving force to produce high resolution (16 m) ET. The ETMonitor model was adopted to estimate ET at 16 m resolution in this study. ETMonitor is a combined model with multi-process parameterizations, and it has been proven to be able to generate accurate regional and global ET estimation at relative coarse resolution (e.g., 1 km) mainly using the biophysical and hydrological parameters/variables retrieved from satellite observations. During the ET estimation procedure, the adopted GF-1 remote sensing datasets include the Leaf Area Index (LAI), Fraction of Vegetation Cover (FVC), Albedo, and NDVI datasets, which are retrieved from previous studies. Ground observation data from 16 sites in China was collected to validate the estimated ET, including 6 grassland sites, 4 cropland sites, 1 mixed forest site, 2 shrubland sites, and 3 desert or Gobi sites. The validation results show that the overall Root Mean Square Error (RMSE) of estimated daily evapotranspiration based on GF-1 satellite remote sensing datasets is 0.85 mm d-1, the correlation coefficient (R) is 0.79, and the Bias is 0.16 mm d-1, which can demonstrate the high accuracy of estimated ET. The GF-1 based ET at 16 m resolution also presented better performance in terms of spatial variation of ET comparing with the low-resolution (e.g., 1 km) ET, especially in the regions with high surface heterogeneity. These highlight the ability of Chinese GF-1 satellite remote sensing dataset could produce accurate ET at high spatial variation, and it has potential to meet the application of field-scale agricultural water resources management, irrigation management, ecological environment monitoring and government decision-making in China. However, due to the impact of the revisit cycle of the GF-1 satellite and the impact of clouds, there are some gaps or missing values in GF-1 based LAI, FVC or Albedo, and these further cause gaps in the GF-1 ET data. In order to improve the availability of high-resolution ET products, it is necessary to produce spatially and temporally continuous high-resolution ET products, which will be the focus of follow-up research.关键词：evapotranspiration;ETMonitor;remote sensing;GF-1;validation;16 m resolution;China- 704
- |
- 409
- |
- 1
发布时间：2023-04-18 -
摘要：The GF-1 wide-field-of-view cameras (GF-1 WFV) has high spatial and temporal resolution, and has great potential in the application of water environment remote sensing. Existing studies mainly focus on region-specific algorithms, and lack of water-quality parameter estimation algorithms that can be used in a large range. Based on the above problems, this study carried out 28 water surface measurement and sampling experiments with 68 voyages in China, and obtained 647 typical and representative sampling point data for water quality parameter estimation modeling and validation. The study area included Taihu Lake, Chaohu Lake, Dianchi Lake, Three Gorges Reservoir, Guanting Reservoir, Yuqiao Reservoir, Shandong Pingyin Small Water Body, Shaanxi Yulin Small Water Body, Ningxia Ningdong Base Small Water Body. The GF-1 WFV images were used the relative atmospheric correction algorithm based on Sentinel2-MSI data of uniform invariant ground object spectral database to obtain the water remote sensing reflectance data. In order to meet the needs of large-scale estimation of water quality parameters in turbid water with complex optical characteristics, a GF-1 WFV water quality parameter estimation algorithm based on soft classification was developed. Firstly, the algorithm divided the water into four types (OWTs) by a stepwise iterative K-mean clustering method and calculated the centroid spectra of each type of water by the average of all spectra of this category, among them, OWT1 was jointly dominated by phytoplankton and non-algae particles, OWT2 was dominated by non-algae particles, OWT3 was dominated by phytoplankton, OWT4 was bloom (no water quality inversion in this water type); Then, the Spectral Angular distance (SAD) was used to calculate the distance from each pixel to each type of centroid spectra and the SAD was converted into distance weight, and the suitable estimation models of chlorophyll a concentration, total suspended solids concentration and transparency were selected and optimized for each type of water body, and the final estimation results of water quality parameters were obtained by weighted fusion with distance weight. In this paper, several band ratio and difference models were investigated. Chlorophyll a used the blue green ratio model in OWT1, the red green ratio model in OWT2, and the red near-infrared ratio model in OWT3. The total suspended concentration was applicable to the red green ratio model in OWT1, the green near-infrared ratio model in OWT2, and the red near-infrared ratio model in OWT3. The transparency models of the three types of water bodies all used the green band and match the blue and red band to constructed ratio model. The mean relative errors of chlorophyll a concentration, total suspended solids concentration and transparency estimation were 33.1%, 28.6% and 17.6% verified by satellite earth synchronization experiment data, and the transition of category boundary was smooth, which avoiding the numerical jump caused by different models. The results showed that this algorithm had the ability to generate water quality parameter production of wide range area. Due to the limitation of the GF-1 WFV sensor band setting (only four broad bands from visible light to near-infrared), the quantification processing and models have great limitations, and the applicability and scalability of the model need to be further improved.关键词：GF-1 WFV;water types (OWTs);chlorophyll a;total suspended solids;transparency- 683
- |
- 390
- |
- 1
发布时间：2023-04-18
Radiation And Water Variable Retrieval And Product
-
摘要：Given the development of China’s high-resolution earth observation system (referred to as GF), common products, standard products, and specific products play an important role in remote sensing applications of GF data. In particular, the common products promote remote sensing applications as the bridge between the GF standard products and specific products. However, GF common products are derived by retrieval algorithms. Inevitable errors bring great uncertainty to the followed-up remote sensing applications. Thus, the retrieval algorithms of GF common products should be tested. Moreover, the products should be validated to ensure the quality of GF common products.This study expounds on the relationship between the algorithm test and product validation and that between the test and validation technology system. Algorithm test and product validation are at the two ends of a common product-generating process. A complete algorithm test process of common products mainly includes the establishment of the algorithm test index system, the construction of the index calculation method, the preparation of the test dataset, the algorithm running and index calculation, and finally, the writing of the algorithm test report. All the generated common products should be validated using the referenced truth data through suitable methods. Thus, a service platform of GF common product validation and algorithm test is implemented to achieve these two functions.The test index system of the GF common product algorithm is constructed with four primary indexes and 10 secondary indexes through literature research, expert consultation, and questionnaire investigation. For example, the land surface albedo retrieval algorithm is tested to show the feasibility of the algorithm test in the service platform. The results show that it can support the validation and algorithm test of Gaofen common products in China.Algorithm test and the product validation technology system and its service platform are important for quality control in the generating of GF common products. The validation and algorithm test of GF common products is a systematic project that requires the cooperation of multiple communities to collect the sample data and reference data through continuous field observation. This field observation is an important foundation of validation and algorithm tests for common products.关键词：remote sensing;GF satellite;common product;algorithm test;product validation;land surface albedo- 1037
- |
- 415
- |
- 4
发布时间：2023-04-18 -
摘要：Remote sensing retrieval is one of the core issues of quantitative remote sensing. High-precision retrieval can improve the utilization efficiency of satellite data and promote the development of quantitative remote sensing. Machine learning algorithms have been increasingly used in remote sensing domains due to the outstanding advantages in dealing with complex and nonlinear problems. They can avoid the complicated processing and calculation in physical models, and minimize the uncertainty resulting from data preprocessing such as geometric correction, radiometric correction, and atmospheric correction. However, the utility of machine learning algorithms in retrieval needs to be viewed objectively. For instance, which machine learning algorithm is the most appropriate or which parameter configuration is optimal in order to obtain more accurate retrieval results. What we are concerned about most is which factors might influence the accuracy of retrievals. Hence, this paper systematically combs the current situation and principles of the application of machine learning algorithms in remote sensing retrieval with the focus on the main uncertainty factors in machine learning-based retrieval.It was found that the number of relevant articles surged especially after 2018. Several mainstream algorithms including random forests, support vector machine, and artificial neural network have been widely used in remote sensing retrieval. And the retrievals of vegetation index, leaf area index, soil moisture, chlorophyll content, and biomass are the main research hotspots at present. MODIS and Sentinel datasets are the most widely used data. The process of machine learning-based retrieval can be summarized as the acquisition of training samples as well as the construction and application of the training model. The main factors causing the uncertainty of retrievals including the selection of machine learning models, the selection of auxiliary variables, the source and accuracy of the variable datasets, the selection of training samples, and the cross-regional and cross-area application of models were discussed in this paper. The findings are helpful in deepening the consciousness and understanding of the uncertainties of the retrieval results based on machine learning algorithms, which is necessary to select the most suitable algorithm and samples for improving the accuracy and reliability of the results.Moreover, it is important to note that there is a trade-off between the model accuracy and complexity for machine learning algorithms. Therefore, they may be not feasible in solving different remote sensing problems. Further retrieval based on machine learning models may be developed from the following four aspects: 1) Selecting more detailed information to fully capture the spatial heterogeneity of land surface and the spatiotemporal variation characteristics of parameters; 2) The training samples should be more representative to make the models more universal; 3) The scale mismatch between the in situ measurements and auxiliary variables extracted from satellite products should be fully taken into consideration to reduce the uncertainty caused by the scale effect; 4) The combination of physical models and machine learning algorithms can improve the representativeness of training samples, providing a new way for the construction of training samples.关键词：machine learning;Retrieval;uncertainty;quantitative remote sensing;deep learning;literature analysis- 1925
- |
- 866
- |
- 1
发布时间：2023-04-18 -
摘要：Scale effect is one of the classical and important problems in the field of quantitative remote sensing especially in surface validation field, in which the judgment of surface heterogeneity is a precursor to the problems of surface validation and station optimized layout, and is also one of the important error sources of surface parameters validation. The first way is to calculate the accuracy evaluation of the scale transformation results between the medium resolution remote sensing products and the ground measurement results or the very high resolution products acquired at the same time to express spatial heterogeneity indirectly, and a series of errors, such as different sensors optical parameters, different measurement angles, spatial and temporal scale inconsistency, geometric mismatching etc., they all affect the results directly or jointly, and the error contributions are difficult to quantitatively, it means that is difficult to describe the spatial heterogeneity clearly. The second way is to use geostatistical methods to describe the images for evaluation the spatial heterogeneity directly. Then how to express the surface heterogeneity with only very high resolution remote sensing measurement image based on the lack of moderation satellite retrieval products is a workable way to describe to spatial heterogeneity for further exploration and analysis of spatial heterogeneity in the next step. Therefore, this paper uses a typical algorithm to portray spatial heterogeneity and discusses the relationship between spatial resolution and spatial heterogeneity in the absence of a reference base of medium-resolution data, with a view to reflecting the relationship between resolution and spatial heterogeneity and conducting a preliminary analysis. Specifically, this paper calculates Normalized Difference Vegetation Index (NDVI) data using Unmanned Aerial Vehicle（UAV） spectral reflectance data with spatial resolution better than 0.2 m that has been Radiation calibration by reflector plates, and obtains results for 39 different spatial resolutions from 0.2 m to 30 m by cubic convolution upscaling algorithm, and obtains land use and land cover change (LULC) by visual interpretation. The spatial heterogeneity of the 1km×1km map area was evaluated with GeoDetector algorithm, and then the regional spatial heterogeneity was described to explore the relationship between resolution and spatial heterogeneity. The results showed that the thresholds of spatial heterogeneity evaluation q value were different in three regions with fragmented land-surface, but the overall q value tended were oscillate to stable with the increase of spatial resolution (30 m to 0.2 m), and the minimum threshold from oscillation to stability was 2 m resolution; then the change curve of q value with spatial resolution and done M-K mutation detection found that the thresholds and q values of spatial heterogeneity mutation points in Ganyansuo and Hutou Village oscillation curve existed for the oscillation to stable points basically matched, but there were multiple mutation points and mismatched in the Caoshang. There were pass the 5% significance test of M-K test for all three areas, which tested the relationship between q value and spatial resolution in the aforementioned in statistical significance. In conclusion, all this classification system was now regionally stable when the resolution was lower than 2 m, i.e., when the resolution was higher than 2 m, its spatial heterogeneity tends to stable, and its could provide some reference for the sampling of ground and space-based platforms.关键词：surface validation;NDVI;UAV;spatial heterogeneity;geographical detector;M-K test- 540
- |
- 293
- |
- 1
发布时间：2023-04-18 -
摘要：The optimal scale of land cover products varies by different applications. Consequently, it needs the scale conversion of land cover products from the base resolution to the objective one. However, the scale effects accompanying the scale conversion often lead to information distortion of land cover products. The state-of-art patch-level research on scale effect, which is vulnerable to patch classification accuracy, often classifies the patches of land cover products into different types. Consequently, a cross-type scale effect model (CSEM) with Patch Area Ratio (PAR) is proposed in this study.The CSEM is aimed to explore further the effect mechanism of Patch Morphological Indexes (PMIs) on scale conversion, predict the areas of land classes in land cover products with specific scales, and provide a rationality evaluation for the scale conversion. First, a candidate index library is defined with seven indexes to describe the patch morphology based on patch size, shape, edge complexity, and internal connectivity. Then, a set of PMIs is defined by information metric techniques. The scale is incorporated into the PMIs by mathematical transformation. Moreover, the CSEM with PAR is constructed by associating the PMIs with PAR, calculated according to the patch areas before and after scale conversion.In this study, a 30 m land cover product from domestic satellites is used with the support of a National Key Research and Development Program to construct and validate the proposed CSEM. According to the experimental results, the CSEM fits the relationship among Filling, logRsr, and PAR: the logRsr is the main factor affecting the value of PAR when logRsr > -0.05, whereas Filling is the main factor on PAR while logRsr < -0.05. The prediction results show that the maximum prediction errors for the four land-cover types of forest, cropland, grassland, and impervious surface on 12 converted scales, varying from 90 m to 750 m with an interval of 60 m, are 0.056, 0.051, 0.051, and 0.053; their average prediction errors are 0.036, 0.033, 0.034, and 0.035, respectively.The results confirm that the CESM explores the effects of Filling and logRsr on PAR and reveals the effect mechanism during the scale conversion, providing instruction for subsequent research of other indexes. The proposed CSEM has satisfactory prediction accuracy for various area types in land cover products with specific scales. It can also provide theoretical guidance for scale conversion and production of land cover products.关键词：land cover product;upscaling;scale effect;Patch area Ratio (PAR);Patch Morphology Index (PMI)- 2742
- |
- 319
- |
- 0
发布时间：2023-04-18
Algorithm Test And Products Validation
- Postal code：100190
- Tel：010-58887052 Email：nrsb@aircas.ac.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备20021838号-9
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0