最新刊期
卷 22 ， 期 1 ， 2018
-
摘要：Solar-Induced chlorophyll Fluorescence (SIF) is directly related to photosynthesis and therefore considered a promising tool for grossprimary productivityestimation and vegetation environmental stress monitoring. Interest in SIF has increased since satellite remote sensing of SIF became feasible, especially after the first global SIF map was depicted in 2011. However, methods for retrieving SIF at Top-Of-Atmosphere (TOA) are still under investigation and argumentation, as decoupling SIF from total at-sensor radiance is challenging in the presence of atmospheric scattering and absorption. This paper aims to review the methods proposed for SIF retrieval at TOA over the past 10 years, to illustratethe advantages/disadvantages of those methods and to provide a technical instruction for remote sensingof SIF at airborne/space level. All the methods were categorized into three types: methods based on Radiative Transfer (RT) calculations, simplified physically-based methods and data-driven approaches. Methods based on RT calculations, including improved Fraunhofer Line Discrimination (FLD) methods and Spectral Fitting Methods (SFM), aim to retrieve SIF using atmospheric absorption lines. Atmosphere is characterized through RT calculations and then the TOA problem is converted to Bottom-Of-Atmosphere (BOA). These methods are applicable in situations with medium to low spectral resolution (0.3—5 nm) whereas imperfect characterization of atmosphere and RT process will lead to retrieval errors. Physically-based methods utilize single or several solar Fraunhofer Lines located in atmospheric windows, using solar irradiance spectra (measured or simulated through spectra convolution) as reference, decoupling SIF signal from earth radiances. Atmosphere scattering and absorption are neglected under these situations. Physically-based methods were developed for high spectral resolution measurements (e.g. 0.025 nm for GOSAT) and are sensitive to noise. Data driven approaches consider any fluorescent radiance spectrum consists non-fluorescent portion and SIF signal. Features extracted from large training dataset consisting non-fluorescent spectra (cloud, ice/snow, desert…) are used to express the non-fluorescent portion in the fluorescent spectra while SIF signal is expressed as mathematical function or spectra with fixed shape. Data driven approaches are popular because they do not require RT calculations while are applicable for medium to high spectral resolution situations. The performance of data driven approaches depends on the representativeness of training dataset and other empirical settings of the model, including the number of features used, function used to describe SIF spectrum and retrieval window selected. With several satellite sensors with medium spectral resolution (0.3—0.5 nm) being available for SIF retrieval (including currently available ones and scheduled to be launched in near future ones), including MetOp-GOME-2, Sentinel-5-TROPOMI and the FLuorescence EXplorer (FLEX) mission, RT-based methods and data-driven approaches are considered most promising SIF retrieval methods in the future. The RT-based methods are mainly developed by the FLEX team and are applicable for low spectral resolution and airborne data, but the performance of these methods on global SIF retrieval needs to be validated with real satellite data. On the other hand, several global SIF products have been generated using data driven approaches. However, representative training dataset needs to be built carefully and optimal parameters need to be determined according to different sensors.关键词：SIF retrieval;hyperspectral remote sensing;radiative transfer;data-driven approaches- 766
- |
- 292
- |
- 11
发布时间：2021-06-07 -
摘要：Hyperspectral imaging, which measures electromagnetic energy scattered in the instantaneous field view in hundreds of bands, contributes significantly to earth observation. As the electromagnetic spectrum covers visible, near infrared, and shortwave infrared spectral bands, it holds abundant spectral information. The subtle discriminative spectral characteristics also make material identification possible. However, the signals recorded by low-resolution hyperspectral sensors from certain pixels under a complex background are a mixture of substances. To improve the accuracy of classification and subpixel object detection, Hyperspectral Unmixing (HU) represents an important solution, especially in the remote sensing area. Mixed pixels comprise a few materials called endmembers. The fraction of each endmember in a pixel is called abundance. Given the hyperspectral data exhibiting high dimension, special spectral correlation, and huge quantity, the determination of the number of endmembers can be regarded as dimensionality reduction. The objective of HU is to estimate and extract the spectral signatures and abundance of endmembers. In this study, mixing models are systematically discussed. On the basis of mixing models, the core problem of HU is presented, which mainly comprises three aspects: estimation of number of endmembers, endmember extraction, and abundance fraction estimation. For each aspect, we summarize the basic theory and development of the HU processing methods. Through the basic theory model, this work performs an initial classification of each aspect. The physical or mathematical problems involved are also discussed along with classical and state-of-the-art methods used to address the problems. Furthermore, a general overview of each category is provided, and a comprehensive analysis of the advantages and disadvantages of classic algorithms is performed. Through the comparison of different methods, this study offers a perspective on the potential and emerging challenges in the process of HU. The research on HU has been continuous. Even though extensive processing results have been achieved and numerous analyses have been performed, obstacles remain and call for solutions. Five challenges are identified in this work: curse of dimensionality, establishment of a high-accuracy mixed inversion model, endmember variation, establishment of remote sensing operational product, and development of algorithms for real-time processing. According to each challenge, the study performs further analysis and presents several remarks. Then, suggestions with regard to future research directions are offered. This study provides a brief overview of HU. With the development of hyperspectral imaging, the performance of imaging data has undergone a qualitative leap. However, different application requires different data types. That is, the most suitable data, and not the data with the highest index, should be adopted to solve the problem. Considering the future development trend of HU, this work suggests that new methods be combined with the demands of users to promote the operational application of HU and support hyperspectral remote sensing engineering applications.关键词：hyperspectral image;linear unmixing;endmember extraction;abundance estimation- 745
- |
- 279
- |
- 28
发布时间：2021-06-07
Review
-
摘要：Land Surface Temperature (LST), which controls the basic interactions between the Earth’s surface and the atmosphere, is significant in many aspects. By far, many algorithms have been proposed to retrieve LST from different satellite thermal infrared data. However, the influence of Air Temperature Inversion (ATI) on LST retrieval has not been considered in the development of existing algorithms. This study aims to analyze and reduce the influence of ATI on LST retrieval. Considering that the Generalized Split-Window (GSW) algorithm has been widely used, we choose this algorithm to retrieve LST in this study. Furthermore, considering that the LST retrieval error increases when ATI intensity increases, we add an error correction related to intensity to the GSW algorithm. To determine the relationship between the LST retrieval error and the ATI intensity, we manually change the normal profile in the Thermodynamic Initial Guess Retrieval (TIGR) database into the ATI profile with the intensity ranging from 1.0 K/100 m to 5.0 K/100 m and the step being 1.0 K/100 m because the intensities of the existing ATI profiles in the TIGR database are not large enough. The LST errors are calculated using the changed ATI profiles and the GSW coefficients derived from normal conditions. To improve the accuracy of the LST retrieval, we divide LST and Water Vapor Content (WVC) into different groups. After calculating the LST retrieval errors of all groups, we find that the LST retrieval error could be expressed as a quadratic function of ATI intensity. The coefficients that correspond to the correction of each group are derived by fitting the LST retrieval errors with various ATI intensities. Results show that the monomial coefficient and the constant of the quadratic function increase when the LST increases while the quadratic coefficient does not change significantly. In addition, the coefficients do not change regularly when the WVC increases. To test whether the proposed method could be used to reduce the influence of ATI on LST retrieval accuracy, we use both simulated data and in situ data. Simulation results show that the LST retrieval accuracy could be improved by 0.44 K when the ATI intensity is 1.7 K/100 m. In situ measurements at the Hailar site are also used to test this method. Results show that the proposed method could improve the LST retrieval accuracy by 0.47 K for the GSW algorithm in atmospheres with ATI. This study aims to add an error correction to the GSW algorithm to improve LST retrieval accuracy when the atmosphere shows ATI. Validation using both simulated data and in situ measurements indicates that the proposed method could effectively reduce the influence of ATI on LST retrieval. However, the application of the proposed method is restricted by the air temperature profile that it requires. A model by which the ATI could be determined from satellite data is expected to be developed in a future study.关键词：Air Temperature Inversion (ATI);ATI intensity;Generalized Split-Window algorithm;land surface temperature error correction;MODIS- 528
- |
- 261
- |
- 5
发布时间：2021-06-07 - 摘要：Himawari-8 (H8), as a new generation of geostationary meteorological satellites that observes full-disk images (images of the Earth as seen from the satellite) per 10 min, was launched by Japan Meteorological Agency to investigate aerosol characteristics. The Advanced Himawari Imager (AHI) onboard Himawari-8 has similar spectral bands with Moderate Resolution Imaging Spectroradiometer (MODIS). This study applies the Dark Target (DT) method for Aerosol Optical Depth (AOD) retrieval from AHI data. The atmospheric effect is established for the AHI data over AErosol RObotic NETwork sites. The ratio of surface reflectance between the shortwave infrared and visible bands is then obtained. This ratio serves as a priori knowledge for the surface reflectance estimation in the atmosphere-surface coupling model. Assuming that the aerosol type is continental, we build a look-up table through the radiative transfer model. With the retrieval algorithm, the AOD is determined in the case of a minimum difference between simulated apparent reflectance and satellite observations. The algorithm was used to retrieve AOD over the Beijing-Tianjing-Hebei area of China in May 2016. H8 AOD products were compared with MODIS products, and the results revealed a good spatial coincidence, with the correlation coefficientR being 0.852. The H8 AOD products were validated with AERONET observations, and they showed good linear relations, with the correlation coefficient R2 being better than 0.88. The high temporal resolution products were used to analyze aerosol spatial distribution and diurnal variation in the Beijing-Tianjin-Hebei region. Results show that H8 AOD retrieval based on the DT method has certain feasibility and potential. AOD products can express the high temporal variation of aerosol and are thus potentially useful in atmospheric environmental pollution monitoring.关键词：Himawari-8;dark target method;surface reflectance;aerosol optical depth;Beijing-Tianjin-Hebei area
- 846
- |
- 364
- |
- 17
发布时间：2021-06-07 - 摘要：Satellite remote sensing has been widely used to retrieve Aerosol Optical Depth (AOD), which is an indicator of air quality and radiative forcing. The Dark Target (DT) algorithm is applied to low reflectance areas, such as vegetated or water areas, and the Deep Blue (DB) algorithm is adopted over bright-reflecting source regions. However, the spatial resolutions of the AOD products obtained with the DT and DB algorithms are relatively low, and the distribution details for urban areas are poor. In this study, a modified retrieval algorithm is proposed for the retrieval of AOD at a spatial resolution of 500 m over Beijing, China, on the basis of Landsat 8 OLI data. The key points are the accurate estimation of surface reflectance and the reasonable assumptions of the aerosol model. We developed a new algorithm to improve the accuracy of land surface reflectance for urban areas. A monthly Minimum Land Surface Reflectance (MLSR) database for China was established using Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance products. The construction of this database was based on the principle of minimum synthesis technique. The conversion model from the blue band reflectance of MODIS to Landsat 8 OLI was established using the ASTER spectrum database to compensate for the difference between the spectral settings of the two sensors. AErosol RObotic NETwork (AERONET) data were adopted to determine aerosol optical properties, such as Ångström exponent, complex refractive index, Single Scattering Albedo (SSA), and asymmetric factor (g). The AODs retrieved from the 36 OLI images were validated with the AERONET data and the NASA/MODIS Collection 6 aerosol products at a 10 km resolution. For this purpose, the DT and DB (DT/DB) algorithms were combined. Results indicated that the proposed algorithm accurately retrieved AODs over the Beijing area and that the retrieved aerosol distribution contained more spatial details and variability than the DT/DB AOD products. Ground-based AERONET observations from four sites (Beijing, Xianghe, Beijing_CAMS, and Beijing_RADI) were used to validate the retrieved AODs. The results from the proposed algorithm demonstrated the highest accuracy, with an average correlation coefficient (R) and Root-Mean-Square Error (RMSE) of approximately 0.963 and 0.156, respectively. Approximately 85.3% of the points fell within the expected error (EE) ± (0.05+0.15 AOD) envelope defined by NASA. The DT/DB AODs showed a small R (0.962) and slightly large RMSE (0.158), with 75.8% of the collocations falling within the EE. An aerosol retrieval algorithm for Landsat 8 OLI 500 m data was proposed in this study. The assumptions in this algorithm are as follows: the variation of surface reflectance needs to be small for a month, and the SSA and g are regionally constant for a particular day. The method can also be used to achieve AOD inversion of other terrestrial observation satellite data. However, according to the uncertainty analysis, the proposed algorithm has some limitations that must be addressed. (1) Errors may arise from the use of constant SSA and g values for the day of retrieval. (2) The MLSR database approach was sometimes unsuccessful over snow surfaces, particularly when seasonal changes such as the snow melting in March and the accumulation of snow in Novembers are significant. These factors will be explored in our future studies.关键词：aerosol optical depth;spectrum conversion;surface reflectance database;Landsat 8 OLI;AERONET
- 661
- |
- 130
- |
- 17
发布时间：2021-06-07 -
摘要：Remote sensing techniques offer a unique opportunity to monitor air quality and are thus crucial for the management and surveillance of the air quality of polluted megacities. MODIS Aerosol Optical Depth (AOD) products with a spatial resolution of 10 km have been widely used to monitor ground-level particulate matters. However, the demands of air quality estimation are difficult to meet in local areas due to the coarse resolution of 10 km AOD. Taking Fuzhou city as an example, this study used the newly released AOD with a spatial resolution of 3 km and meteorological data to map ground-level PM2.5 concentrations in the city and ultimately reveal the spatial details of the PM2.5 exposure. Two regression models, namely, the daily calibration model and site daily calibration model, were developed to estimate and map ground-level PM2.5 concentrations in Fuzhou, China. The MODIS 3 km AOD data for 2014—2015, in situ PM2.5 concentration data for the same period, and meteorological data of wind speed and relative humidity were used. A simple linear model was also derived and used for comparison with the two calibration models. Results showed that the PM2.5 concentrations and AOD had an extremely low agreement when a linear fit was applied, with the R2 value being 0.117 and the RMSE being 19.510 μg/m-3. Strong correlations were obtained with the use of the daily calibration model, which yielded an R2 of 0.762 and RMSE of 10.146 μg/m–3. A relatively high degree of agreement was achieved when the site daily calibration model was used; R2 was 0.814, and RMSE was 8.965 μg/m–3. Ten-fold Cross Validation (CV) was conducted to evaluate the performance of the regression models. The CV results showed that the site daily calibration model performed better than the daily calibration model. Correlation coefficients (R2) of the estimated PM2.5 concentrations with the in situ data were 0.781 (RMSE=9.687 μg·m–3) and 0.724 (RMSE=10.993 μg·m–3). In addition, the PM2.5 concentrations estimated by the site daily calibration model had a better agreement with the observed values for all seasons from 2014 to 2015. The R2 of the estimated and observed values of the seasonal average PM2.5 concentrations for the two models were 0.999 and 0.995, respectively, indicating that both models could reflect daily variations in the relationship among AOD, meteorological data, and PM2.5 concentrations. In this study, we proposed a daily calibration model and a site daily calibration model using the newly released MODIS 3 km AOD product and meteorological data to estimate ground-level PM2.5 concentrations in Fuzhou, China. The daily calibration model was used to retrieve the distribution of PM2.5 concentrations in Fuzhou, as the site effect parameters needed for the site daily calibration model is not available for every 3 km grid. Nevertheless, these two models perform similarly in PM2.5 estimation. The spatial distribution of PM2.5 concentrations in Fuzhou derived from the MODIS 3 km AOD exhibits high concentrations over central urban areas and low values over suburban districts. These results clearly reveal the spatial variation of PM2.5 in the area. This study indicated that the satellite-derived model based on the MODIS 3 km AOD product could work effectively in estimating PM2.5 concentrations on a local scale.关键词：MODIS 3 km AOD;PM2.5 concentration;remote sensing estimation;daily calibration model;site-daily calibration model- 570
- |
- 200
- |
- 11
发布时间：2021-06-07
Atmospheric Remote Sensing
- 摘要：The Space Environment Monitor (SEM) aboard the FY-3C satellite monitors energetic particles, including protons, electrons, and heavy ions, and the radiation dose and surface potential of the satellite caused by the particles. The independent data on the energetic particles of the FY-3C satellite are essential in space weather operation. Quantitative evaluations of such data are conducted to examine their validity. Cross comparisons are performed with similar satellite data to evaluate the energetic particle data of the FY-3C satellite. Normalization is performed based on the basis of certain assumptions to eliminate the differences caused by observation time, location (longitude and latitude of the sub-satellite point, and satellite height), orientation, and energy range. Statistical parameters such as correlation coefficient, slope, and standard deviation are calculated to evaluate the consistency between the compared data. The cross comparison of the FY-3C data with the NOAA-18 and FY-3B data during a quiet space weather period shows that the energetic particle data of the FY-3B satellite are consistent with the compared data. The consistencies of the FY-3C data with the FY-3B data are satisfactory as they have the same specifications. Results show the consistency of the instruments. Through observations of a solar proton event and energetic electron storm, this work finds that the energetic particle data of the FY-3C satellite can accurately reflect the characteristics and intensities of space weather events. Results show the reasonable quality of the energetic particle data of the FY-3C satellite and their suitability for space weather monitoring, warning, and research.关键词：FY-3C satellite;energetic particles data;cross comparison;space weather
- 503
- |
- 143
- |
- 1
发布时间：2021-06-07 -
摘要：Hyperspectral remote Sensing Images (HSIs), which contain rich spectral and spatial information, are important in the precise classification of earth objects. However, HSIs are usually highly dimensional and non-linear, and they contain large amounts of data. These characteristics increase the difficulty of data processing and bring about the Hughes phenomenon. Conventional methods such as Neural Networks (NN) and support vector machine have solved these problems by reducing the dimensions of HSIs with PCA, ICA, or MNF. Although these methods are effective in the classification of HSIs, they may cause information loss in the original data. Therefore, improving the accuracy of HSI classification with inadequate data is difficult. Recently, deep learning method, especially Convolutional Neural Network (CNN), has achieved remarkable performance in many fields. Therefore, the application of CNNs to HSI classification shows immense potential. To avoid the Hughes phenomenon and improve the accuracy of his classification, this study proposes a Coefficient of Variation–Convolution Neural Network (CV-CNN) method for the classification of HSIs. After the calculation of the Coefficient of Variation of the IntrA-class (CVIA) and the Coefficient of Variation of the IntEr-class (CVIE) of each band, the bands with low (CVIE)2/CVIA values are excluded. Then, a target pixel with the spectral information of its eight neighbors is organized as multi-layer spectral–spatial information. The spectral–spatial information of the target pixel should then be converted into matrix form. The two-dimensional image suitable for the input of CNN was subsequently obtained. Furthermore, a seven-layer CNN model was constructed with two convolution layers, two max-pooling layers, two full-connection layers, and one softmax layer. Using the seven-layer CNN can effectively improve the accuracy of HSI classification. Experiments were conducted on the Indian Pines dataset and Pavia University dataset to evaluate the performance of the presented CV-CNN method. The results are as follows: (1) Compared with the method that only considers spectral information for HSI classification, the use of spectral–spatial information can actually improve the accuracy of HSI classification. (2) Compared with other CNN methods, the seven-layer CNN model with no band removed can increase the overall accuracy by 2.61% for the Indian Pines dataset and 0.04% for the Pavia University dataset. (3) Based on the same seven-layer model, the experiments show that the classification of HSI with poor bands excluded shows increased accuracy from 97.9% to 98.69% for the Indian Pines dataset and from 99.6% to 99.66% for the Pavia University dataset. This outcome verifies the validity of excluding poor bands on the basis of the CV method. (4) Based on the same seven-layer model, the experiments show that the CV method, compared with the PCA and MNF methods, can increase the overall accuracy by 1.38% and 0.44%, respectively, for the Indian Pines dataset and by 3.26% and 1.83%, respectively, for the Pavia University dataset. The CNN model developed in this study and the method of removing poor bands with the CV technique can improve the accuracy of HSI classification.关键词：Convolutional Neural Networks(CNN);deep learning;coefficient of variation;hyperspectral remote sensing images;classification- 599
- |
- 115
- |
- 19
发布时间：2021-06-07 - 摘要：Land surface albedos represent an important parameter in the Earth’s energy balance, and they play a crucial role in the global change and climate systems. Numerous global or regional land surface albedo products with various spatial resolutions and temporal frequencies have been released. Thus, the quality analysis of albedo products has become increasingly important. On the basis of the algorithm of a version 5 albedo product (MODIS V005), a new daily product version (MODIS V006) with a high temporal resolution is generated with MODLAND. The new albedo product requires urgent validation, and its accuracy should be known before extensive use. On the basis of sites with adequate spatial representativeness in the FLUXNET network, direct comparisons between different versions of albedo products and in situ albedos were conducted. To investigate the accuracy under various land cover type conditions of the two versions of albedo products, we divided the ground measurements into six classes and calculated several evaluation indicators. Two different versions of products were also compared using a cross-validation method under the prior knowledge that V005 has been extensively estimated and that the method achieves good performance in various validation works. At the same time, the Quality Control (QC) flag was investigated to analyze the stability of the new albedo products. The direct comparison results indicate that two versions of products show good consistency against the in situ albedos. However, V006 performs better than V005 during the same experimental time. For the accuracy under different land cover types, all the evaluation indicators briefly show that V006 products achieve a better match with in situalbedos than V005 products. As the V006 has a high temporal resolution, accurately capturing the rapid changes of land surface albedos is relatively easy, especially during snow-covered days. The comparison result obtained through the cross-validation method indicates that except for a few slightly high values, the V006 product effectively matches the V005 product. Although the percentage of QC flag indicating the good quality of the V006 product is lower than that of the V005 product, data show the stable performance of both products in the span of three years. The MODIS V006 product has a shorter temporal resolution and higher accuracy than the V005 product. The trends of the two albedo products over the time series highlight their good consistency. However, V006 performs better than MODIS V005 during snow-covered times. This finding is expected to play a significant role in many other research fields.关键词：remote sensing;validation of albedo;MODIS;FLUXNET;quality control flag
- 518
- |
- 185
- |
- 4
发布时间：2021-06-07 - 摘要：Hyperspectral imaging collects both spectral signals and images of ground objects on the Earth’s surface using hundreds of narrow bands. It has become a powerful technique for air-to-land observation. Unfortunately, numerous bands and strong intra-band correlations bring about information redundancy and heavy computational burdens for hyperspectral imagery (HSI) classification. Moreover, the “Hughes” problem traps the HSI data into a conflict between a high classification accuracy and an improbably large number of training samples. Therefore, proper bands should be selected from original hyperspectral data to reduce the high computations while achieving superior classification accuracies. In this study, a Weighted Probabilistic Archetypal Analysis (WPAA) method is proposed to extract proper bands from HSI data. WPAA considers the differences between pairwise bands and adopts a composite dissimilarity measure to construct a weighted matrix. This method improves the regular archetypal analysis by using the constructed weighted matrix in hyperspectral bands. Thereafter, the method considers the Dirichlet distribution of sparse coefficients and the quantum nature of hyperspectral imaging. It also introduces the Bayesian framework to construct its mathematical model for band selection. WPAA implements an iterative optimization scheme, transforms unconvex problems into two convex subproblems, and utilizes the Alternating Direction Method of Multipliers (ADMM) to estimate two sparse coefficient matrices. ADMM introduces proper auxiliary variables to augment the constraints in the objection function of WPAA, iteratively minimizes the Lagrangian function with respect to the primal variables, and maximizes it with respect to the Lagrange multipliers. The iteration procedure of two sparse coefficient matrices is repeated until the convergence conditions are satisfied or the number of iterations exceeds the predefined maximal number of iterations. The proper band subset is finally estimated using the sparse reconstruction equation. Two groups of classification experiments on the popular HSI datasets of PaviaU and Urban were designed to carefully test the performance of WPAA in band selection. Four state-of-the-art methods were compared against the proposed WPAA method, i.e., the sparse-based band selection, sparse nonnegative matrix factorization, improved sparse subspace clustering, and sparse self-representation. Experimental results show that WPAA achieves better overall classification accuracy than the other four state-of-the-art methods. WPAA could also obtain the best classification map regardless of the different sizes of the band subset and different training samples. The proposed composite dissimilarity weighted matrix makes great contributions to the improvement of the classification accuracy of the PAA model. WPAA could help select the proper bands for hyperspectral images, and it can serve as a good alternative for dimensionality reduction in hyperspectral image classification.关键词：hyperspectral imagery;archetypal analysis;weighted probabilistic archetypal analysis;band selection;sparse representation
- 535
- |
- 88
- |
- 8
发布时间：2021-06-07 -
摘要：With increasing image resolution, change detection for high-resolution images has become one of the most important aspects of remote sensing research. Change Vector Analysis (CVA) is an effective method that has been widely used in change detection for low- or moderate-resolution remote sensing images. However, processing high-resolution data involves limitations caused by spectral heterogeneity and objects with different scales. Thus, CVA must be combined with object-oriented methods. However, the performance of most object-oriented methods depends on the results of image segmentation, which are unstable due to the difficulty in determining the optimum scale. Taking advantage of the rich spatial information in high-resolution images is obviously important. Considering the aforementioned problems, this work proposes an object-oriented and multi-feature hierarchical change detection method based on CVA. Bi-temporal high-resolution remote sensing images are hierarchically segmented. Hierarchical image segmentation is realized with an image segmentation method based on a region adjacency graph. A logical OR is then applied to corresponding segmentation levels of the bi-temporal images. On the basis of spatial characteristics, spectral, texture, and shape features are extracted. A gray level co-occurrence matrix is used as a texture feature, and a geometric moment is used as a shape feature. Feature selection is realized with a random forest algorithm. Then, CVA is conducted to calculate the hierarchical magnitude images according to the optimal feature vectors. The final change magnitude image is obtained by fusing the hierarchical magnitude images using an adaptive fusion method. Finally, the Otsu algorithm is used to determine the change threshold values and thereby realize change detection. The change detection result of the object-oriented and multi-feature hierarchical CVA method is compared with those of the hierarchical CVA that only uses spectral features, the multi-feature pixel-based CVA, and the pixel-based CVA that only uses spectral features. Experiment outcomes show that the proposed method offers higher change detection accuracy and greater stability than the other three methods. It also achieves reduced false alarm rate and missing alarm rate in the change detection result. The proposed method can also reduce salt-and-pepper noise and produce entire changed objects. The unit for analysis in the proposed object-oriented and multi-feature hierarchical change detection based on CVA is an image object at different scales. The changes are detected at different hierarchy levels to adapt to the different characteristics of objects with different scales and to avoid the difficulty in determining the optimal segmentation scale. Thus, the impact of segmentation on change detection precision is minimized. In addition, multiple feature extraction and optimal feature selection ensure that the spectral and spatial information of high-resolution images is fully utilized while the characteristic redundancy is reduced. Therefore, the influences of complex spectral characters are avoided. The proposed method demonstrates high performance and high reliability.关键词：remote sensing change detection;change vector analysis;multiscale segmentation;feature selection;adaptive fusion- 639
- |
- 221
- |
- 21
发布时间：2021-06-07 -
摘要：The research on cloud detection is an important branch of remote sensing image research. In recent years, with the increasing number and type of remote sensing satellites, cloud detection based on reference map/sequence image has become a subject receiving close review in cloud detection. GF-4 is a geo-synchronous orbit satellite launched by China in December 2015. This satellite is equipped with a visible-light/near-infrared camera with a resolution of 50 m and has typical high-resolution and multi-spectral satellite data characteristics. GF-4 satellites have many common characteristics as meteorological satellites. These common features include the geostationary orbit, area array starring imaging, and the mid-infrared band. GF-4 has the capability to acquire the sequence data of the same area in a short time. This paper attached importance to the algorithm of automatic cloud detection for early GF-4 satellite data acquisition. The research is based on the same area of multiple GF-4 images. First, according to the characteristics of the image data and cloud in different images on the movement characteristics, this work performed automatic geometric registration and relative radiation normalization on multiple images. Then, the image was set to automatic threshold by cloud detection and was processed by the Savitzky–Golay filtering. Finally, this work implemented an automatic cloud detection algorithm for GF-4 sequence images. This paper selected 36 data in the eastern region of Inner Mongolia and 39 data in the middle and lower reaches of the Yangtze River for cloud testing to detect the practical feasibility of the new algorithm. The following preliminary conclusions were obtained. (1) The results of the cloud detection algorithm based on sequence image are superior to those of the single-image cloud detection in terms of overall accuracy. The main difference was observed in the image of the cloud boundary, highlighted surface, and thin cloud area. Through the experiment, this work showed that the algorithm has a high degree of automation and can satisfy the needs of engineering data. (2) Based on the single-image cloud detection, using the automatic threshold method can provide an overall stability for the test results. However, owing to the diversity of the cloud in the image, improving the accuracy is obviously difficult for the proposed algorithm. (3) The mid-infrared band data of GF-4 cannot be simply used for cloud detection due to differences in coverage area, spatial resolution, and acquisition time. The shortcoming of the current algorithm is that the acquisition time of the sequence data is extremely long. Eliminating the radiation difference between the data obtained in the morning and those obtained in the noon with the simple linear relation is difficult. Simultaneously, the follow-up research will focus on the systematic cloud detection accuracy evaluation method.关键词：GF-4;cloud detection;automatically matching;Savitzky-Golay filtering- 539
- |
- 195
- |
- 6
发布时间：2021-06-07
Technology and Methodology
-
摘要：High-precision and rapidly changing Net Primary Production (NPP) monitoring in regions depends on remote sensing data with high spatial and temporal resolution. However, high-quality remote sensing data with high spatial and temporal resolution are difficult to acquire with a single remote sensor. To solve the problem of missing data, we used the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) algorithm in blending MODIS and Landsat data. We obtained high-frequency temporal information from the MODIS data and high-resolution spatial information from the Landsat8 OLI data to predict NPP with high spatial and temporal resolution. Using the STARFM model, we obtained the time series data with 30 m spatial resolution and 16 day temporal resolution. Then, the fused time series NDVI data combined with meteorological data were used as inputs to the Carnegie–Ames–Stanford approach model to estimate the NPP in the Chang–Zhu–Tan urban agglomerations. An area with 500×500 pixels was randomly selected from true Landsat images and fused Landsat images. The correlation coefficient and RMSE between the true Landsat8 NDVI and fused Landsat NDVI image were above 0.7 and 0.08, respectively. This result indicates that the true data and fused data maintained good consistency. The NPP estimated with the fused Landsat NDVI data showed better detailed spatial information than that obtained with the MODIS data. The fused NPP data showed distinct boundaries between water, road, and building, whereas the NPP simulated from the MODIS did not. The mean values of the NPP data obtained with the fused Landsat NDVI and MODIS NDVI data were 323.01 and 260.88 gC·m–2·a–1, respectively. The mean value of the NPP from the fused Landsat NDVI data was higher than that of the NPP from the MODIS data due to the spectral difference between the MODIS and Landsat images. To further validate the fused NPP data, we employed the measured value of NPP. The correlation coefficients, RMSEs, and relative errors of the fused NPP and measured NPP were 0.89, 4.83 gC·m–2·a–1, and 9.82%, respectively, which indicated that the fused NPP had a good consistency with the measured data. The fused NPP was consistent with the NPP obtained with the MODIS data in terms of pixels; the vegetation coverage accounted for more than 80% of the mixed pixels. Meanwhile, the fused NPP was significantly higher than the NPP estimated from the MODIS data in terms of pixels; the vegetation coverage accounted for less than 20% of the mixed pixels. At the same time, the fused NPP data retained the time features of the time series MODIS data. With the rapid development of regions, remote sensing data from a single date cannot meet the requirements of the monitoring of dynamic vegetation growth in urban areas. A spatial–temporal fusion technique is an effective way to blend images from different sensors for applications that require high resolutions in both time and space. This technique should be able to support time series remote data with high spatial and temporal resolution for NPP monitoring in regional areas. However, due to the rapid expansion of cities and unreasonable planning of regions, vegetation patches become increasingly fragmented with high spatial heterogeneity. Therefore, on the one hand, we should improve the spatial resolution of remote sensing data. On the other hand, we have to solve spatial heterogeneity, which is the focus of our follow-up research.关键词：Landsat 8 OLI;MODIS;data fusion;high spatial and temporal;net primary production;remote sensing estimation- 782
- |
- 213
- |
- 12
发布时间：2021-06-07 -
摘要：The flow velocity of a glacier is not uniform in time and space. Thus, the flow characteristics of glaciers should be detected to fully monitor their situation through satellite images. In this study, we use the Normalized Cross Correlation (NCC) algorithm and the intensity tracking method to study the spatial distribution characteristics of Laohugou glacier No. 12 in Gansu province, to explore the relationship between glacier flow velocity and temperature, and to evaluate error sources and precision using Landsat 8 and SAR images. The flow velocities of different seasons between 2014 and 2016 are obtained from Landsat 8 images in the NCC registration algorithm. The land surface temperature is inversed from the TIRS band. Flow velocities from April to October 2008 are retrieved from TerraSAR-X images under an intensity tracking method. Results of the two types of data show that the velocity at the glacier terminal is lower than that in the central area and that it decreases from the axis to both sides. Flow velocity is faster in summer than in winter, and its trend is consistent with temperature changes. Moreover, the velocity at the west branch is relatively large, and the maximum velocities extracted from Landsat 8 and TerraSAR-X are 2.56 m·d-1 and 2.74 m·d-1, respectively. Finally, the reliability of monitoring glacier flow based on the two types of data is demonstrated by using the mean and standard deviation in the stable zones. Considering these values, we find that the velocity accuracy for Landsat 8 is between 1 and 9 cm·d-1 and that it is better than the other type of data by approximately 2 cm·d-1. Our methods can effectively monitor glacier flow conditions, and reliability evaluation shows that the velocity accuracy is up to several centimeters per day. The comparisons indicate that radar images are especially sensitive to surface deformation due to their high spatial resolution. The SNR of optical images is also relatively high to improve integrity. Therefore, the two types of data can be used in combination to extract glacier flow information comprehensively.关键词：glacier motion;NCC ( Normalized Cross Correlation);image registration;Landsat 8;TerraSAR-X- 581
- |
- 139
- |
- 4
发布时间：2021-06-07 -
摘要：DMSP/OLS nighttime light data have been widely applied to reveal the intensity of human activities and the process of urbanization.However, the limitation of sensors leads to a serious saturation problem in the NTL data,especially in the city center area of high light intensity. Saturated nighttime light data cover the differences and details of light intensity in the central area of a city and reduce the application scenarios of DMSP/OLS data.The method for alleviating the saturation of light data is mainly divided into radiometric calibration and non-radiometric calibration. Although the radiometric calibration method achieves high precision, the complex calibration algorithm and interference factors lead to a low data output rate. The non-radiometric calibration method uses auxiliary data (including natural factors, social indicators, and economic indicators) to construct the function model and thereby establish correct saturation data. We use GeoDetector to identify the factors of Normalized Difference Vegetation Index (NDVI), Enhancel Vegetation Index (EVI), and Unit Network Length (UNL) and propose the unit network length-adjusted (UNLI) NTL Index after quantifying the explanatory power of stable light data. Furthermore, we evaluated the capabilities of EANTLI and UNLI to eliminate saturation through the “detailed description of the degree of light intensity in the saturation area,” “correlation analysis of UNLI, EANTLI, and (RCNTL) radiometric calibration data,” and “goodness of fit of power consumption and GDP”. Both UNLI and EANTLI yield good results in the modified saturation data,but UNLI achieves greater accuracy in describing the difference in the light intensities of the saturation area. UNLI can break through the limitations of spatial resolution and increase the intensity of light detail differences. In terms of the light of the highest intensity saturation region, UNLI is closely correlated with RCNTL. However, with the decrease of light intensity, the linear regression coefficientR2 of the UNLI, EANTLI and RCNTL gradually approached, and the fitting advantages of UNLI and RCNTL is lost. In the fitting analysis of power consumption and GDP, the piecewise calibration model of “EANTLI&UNLI” can maximize the advantages of EVI and UNL, which has the best fitting effect (adjustedR2=0.873). In this study, we proposed a new method to correct the saturation of night light data. The new method can effectively eliminate the saturation of native NTL in urban central areas. Compared with previous correction methods, the new method is highly advantageous in terms of light intensity detail and spatial resolution. Thus, the proposed method can accurately reflect the characteristics of urban structure from the perspective of night light luminance.关键词：DMSP/OLS;nighttime light;Geo detector;saturation;EANTLI;UNLI- 630
- |
- 333
- |
- 21
发布时间：2021-06-07 -
摘要：The information on vertical forest structure is important to accurately estimate aboveground biomass. Stereoscopic radar has the capability to detect the vertical structure of forest stands. However, the radar image resolution is low in the early stages. The emergence of high-resolution radar data such as TerraSAR-X provides an opportunity to detect vertical forest structures using stereoscopic radar data. The goal of the current study is to examine the influence of critical matching parameters on DSM extraction using stereoscopic radar data. We evaluated the DSMs extracted using the TerraSAR-X stereoscopic radar data of stripmap mode over the Daxing’anling forest area and Changbai Mountain nature reserve. The critical parameters examined included the number of pyramid layers, matching window size, and search range. Result: (1) In the Daxing’anling study area where the parallax was small, the DSM accuracy improved as the number of pyramid layers decreased. We used five, six, and seven pyramid layers in the matching process. DSM errors within ± 10 m accounted for 81.8%, 77.7%, and 77.1% of the total number of pixels. Over the area where the parallax was large, such as in the Changbai Mountain study area, increasing the number of pyramids proved useful in correcting matching errors. The DSM obtained by the seven pyramid layers could reflect the correct terrain. (2) A large matching window improved the correlation of the image pair and reduced the noise caused by the lack of texture information in the forest area. The average correlation of the two study areas increased from 0.43 and 0.40 to 0.49 and 0.45, respectively, when a large window was applied over the two test areas. (3) Compared with that of the reference data, the RMSEs of the extracted DSM were 6.682 and 10.384 m over the Daxing’anling area and Changbai Mountain, respectively. Large errors occurred where the slope changed significantly. Significant slope changes resulted in large geometric distortions such as foreshortening and overlap. In these cases, corresponding matching points could not be found with our matching algorithm, resulting in the failure of matching results. The influences of critical matching parameters on DSM extraction were obvious. The proper settings of the number of pyramid layers, matching window size, and search ranges should be carefully evaluated.关键词：TerraSAR-X;stereoscopic SAR;forested area;DSM;matching- 540
- |
- 151
- |
- 0
发布时间：2021-06-07 -
摘要：Low-altitude remote sensing recently became a hot technology with rapid development and com-prehensive application. Small consumer UAV attracted wide attention with rapid, flexible, and cost-effective ad-vantages. Large professional UAV, which is vulnerable to weather conditions, requires professional manipulation and airspace application. These factors restrict its ability to access terrain data agilely and rapidly. A small consumer UAV can compensate for large professional UAV limitations. This study comprehensively verifies the accuracy of the data obtained through this type of UAV to improve application reliability. This study focuses on the precision verification of topographic data obtained by small consumer UAV (Phantom 3 Professional). Six kinds of flight heights (50 m, 60 m, 70 m, 80 m, 90 m, and 100 m) are set to acquire a stereoscopic image and generate Point Cloud, Digital Surface Model, and Digital Orthophoto Map. Ground Control Points (GCPs) are laid out uniformly in the standard experimental field to verify measurement accuracy, and their three-dimensional coordinates are derived using differential GPS with high-precision. The absolute position of the stereoscopic images is calibrated by GCPs. Finally, the measurement accuracy of the result is analyzed using the mean error and relative root mean square error (RMSE). Results show that the resolution of the UAV image is 2.22—4.23 cm for flight height 50–100 m and will decrease with increasing flight altitude. The mean error is ± 0.51 cm in the horizontal direction and ± 4.39 cm in the vertical direction. RMSE is ± 2.79 cm in the horizontal direction and ± 9.98 cm in the vertical direction. The errors in horizontal and vertical directions are within normal distribution, but the error range is larger in the vertical direction. Five or more images in the same area are recommended when shooting to avoid errors caused by insufficient image overlap and to generate high-quality data. Simultaneously, GCPs should be evenly laid in the survey area to ensure absolute positioning accuracy and should be found in more than five images. Experimental precision can be influenced by a number of factors, such as light and weather con-ditions and flight stability. The GCP selection, measurement method, and image spike processes include some errors. The research shows that the measurement accuracy of small consumer UAV can reach centimeter level with reliable flight control system condition; this condition provides a new measurement method for low-cost, fast, flexible, and accurate terrain information acquisition to geography and ecology researchers.关键词：small consumer UAV;accuracy verification;topographic survey;differential GPS;GCP (Ground Control Points)- 654
- |
- 293
- |
- 19
发布时间：2021-06-07
Remote Sensing Applications
- 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