最新刊期
卷 24 ， 期 1 ， 2020
-
摘要：Compared with the single threshold segment method in SAR change detection, the dual thresholds segment method can simultaneously identify the change areas and confirm the change types. Although D-GKIT shows a superior performance, a strongly overlapping gray level is observed in the histogram of the difference image, thereby inaccurately identifying double thresholds on the same side of the peak. In this paper, we apply a dual-thresholds method combined with normalized maximal between-class variance and GKIT test on GF-3 images to verify its feasibility of our method and the ability of change detection ability.First, the normalized maximal between-class variance values of two sides surrounding the peak in the histogram are taken as the degrees of the overlapping gray level, and then the thresholds selection sequence and the candidate intervals are confirmed. Second, the side at which the gray level lightly overlaps is segmented by GKIT, and the threshold and the fitting function of the unchanged class are obtained. Third, the fitting function of the unchanged class is used to replace the corresponding part in the origin histogram to form a new histogram that is subsequently segmented to obtain the threshold in the second candidate interval. Finally, the two thresholds are applied on the difference image to obtain the final change result.The experiment on GF-3 SAR images reveals that the performance on our proposed method outperforms D-GKIT and can deal well with the overlapping gray level overlapped in the histogram of the difference image. The confusion matrix of the results for various local areas in the change image also shows that the proposed method has been slightly influenced by the overlapping gray level overlapped and obtains generally good results. Therefore, the feasibility of our method and its change detection ability by using GF-3 images are verified.We propose a method based on the normalized maximal between-class variance and GKIT to segment a difference image by applying dual thresholds in SAR change detection. The effectiveness of the proposed method and its change detection ability by using GF-3 images are validated by the experiment results.关键词：remote sensing;SAR change detection;dual thresholds segment;GF-3;GKIT;normalized maximal between-class variance- 579
- |
- 136
- |
- 4
发布时间：2021-10-18 -
摘要：The extraction of buildings from remote sensing imagery has an important application value. However, high-resolution images contain detailed information and complex features that hinder the difficulty of building extraction process.To address this problem, we propose a building extraction method of building extraction based on multi-scale SLIC-GMRF and FCNSVM that demonstrates an improved ability of extracting buildings from high-resolution remote sensing images to some extent. First, a multi-scale SLIC-GMRF segmentation algorithm is applied to determine the initial building area, and then the advantages of the FCN neural network in semantic segmentation are utilized to extract the building features. Second, the extracted building features are used to train an SVM classifier to refine the building extraction results of building.The results of three control experiments and two comparative tests reveal that the SLIC segmentation algorithm affects the initial segmentation results, the SVM classifier affects the extraction of building details, and the FCN features influence the performance of the SVM classifier. The precision rate, recall rate, and quality index of the proposed method are all better than the compared methods.The following conclusions can be drawn from the experimental results. For the study area with clear features and minimal obstructions, the proposed method can effectively extract buildings from an image. This method can also obtain ideal results for areas with a complex distribution of buildings can also get ideal results.关键词：remote sensing;building extraction;image segmentation;FCN neural network;SVM;high-resolution remote sensing image- 719
- |
- 363
- |
- 6
发布时间：2021-10-18 -
摘要：Remote sensing images are often affected by noise in the process of digitization and transmission processes. Denoising is an indispensable way of improving image quality. Despite showing an excellent noise removal performance, the existing denoising algorithms however typically suffer from a common drawback. Specifically, in the learning process, some edge information is lost, thereby over-smoothing the denoising result. Given the importance of details—including sharp edges and texture information—in remote sensing images, we propose a residual encoder–decoder denoising network with joint loss (REDJ) for GF-2 satellite data.Inspired by U-net, we use a deep convolutional framework is used to learn the end-to-end mapping from noisy images to the original ones. The encoder acts as a feature extractor that captures semantic information of image contents while eliminating noise, whereas the decoder recovers the image details. The high-resolution features from the encoder are combined with the up-sampled output by skip connection. We also introduce high-frequency decomposition and residual mapping to simplify the training process by reducing the solution space. As for the loss function, we modify the traditional denoising per-pixel loss. Given a well-trained convolutional neural network for defining perceptual loss, we instead to learn the perceptual differences of the extracted features instead of merely matching the low-level pixel information. Unlike the loss of detail resulting from normal per-pixel MSE loss, we recommend a new joint loss that combines the advantages of both per-pixel reconstruction and feature reconstruction, preserves additional edge and texture information, and generates clear denoised results. We employ the GF-2 satellite images in the experiments. To obtain enough training and testing data, we divide the entire high-resolution image is divided into 1200 pictures of size 512 and then allocate 70% of these images for training and the other 30% for testing. We generate the noisy images by adding Gaussian noise.To verify the effectiveness of our proposed network, we compare our quantitative and qualitative results with those of other state-of-the-art methods, including wavelet threshold, total variation, and K-SVD. Our proposed method REDJ can obtain the best index values both of PSNR and average gradient. In the qualitative visual sense, REDJ obtains clear denoising results because of the joint of perceptual loss. Compared with other methods that produce blurred regions generated by other methods, REDJ preserves more edge information and texture details. We also compare the run times of different methods for denoising images and find that REDJ has a relatively high CPU speed and achieves an excellent computational efficiency on GPU time.This paper successfully applies deep learning theory for denoising remote sensing images. We use the proposed network is used to remove noise from high-resolution GF-2 remote sensing images and to preserve the edge contours and fine details, which is conducive to facilitate later detection, classification, and other remote sensing applications. In our future work, we will explore to handle other types of noises, especially the complex real-world noises, and consider a single comprehensive network for more image restoration tasks.关键词：remote sensing image denoising;convolutional neural network;edge enhancement;perceptual loss;GF-2 remote sensing image- 607
- |
- 187
- |
- 6
发布时间：2021-10-18
Chinese Satellites
-
摘要：Multi-feature fusion has been widely employed for high-resolution remote sensing images change detection given its ability to reduce the influence of radiation difference, projection errors, and shadows. However, most multi-feature fusion methods depend on artificially designed fusion rules or man-made samples. Meanwhile, many methods for target change detection on bi-temporal images have successfully detected the changed areas yet fail to recognize the number and location of changed targets. To address these limitations, this paper proposes an aircraft target change detection method for high-resolution remote sensing images that combines adaptive multi-feature fusion for change detection with deep learning for target recognition.First, Iteratively Reweighted Multivariate Alteration Detection (IR-MAD) is applied to generate a change intensity map, and then adaptive histogram statistics are employed to calculate the threshold for the automatic acquisition of conspicuous changed and unchanged samples. Second, the edge and textural features are extracted respectively from multi-temporal images, and a multi-feature fusion change detection is carried out by using a support vector machine and the previous samples. The change polygons are obtained after morphological processing. Third, to prevent a false detection by a single threshold, a multi-threshold strategy is adopted, and raw images from the change polygons are recognized by using the aircraft recognition model that is trained by a few plane slices and the convolutional neural network with network in network.The experiments for airport images change detection and final aircraft target change detection experiments are performed using two datasets. First, we compare our proposed method is compared with other thresholds and check whether to use multi-features should be used to highlight the effectiveness of our multi-feature fusion change detection method with adaptive samples. To verify its performance, we compare our method with some popular change detection methods are compared, including Multivariate Alteration Detection (MAD), IR-MAD, principal component analysis, change vector analysis, robust change vector analysis, and iterative slow feature analysis. The experiments show that the overall accuracy of our proposed change detection method outperforms the other compared methods in terms of accuracy and false alarm rate. We obtain our target change detection results based on the change map, and validate the excellent performance of our proposed method based on its accuracy.To fully use of the spectral, spatial, and textural features of high-resolution remote sensing images, we design an adaptive multi-feature fusion method for change detection that requires less manual work and reduces the influence of radiation difference, projection error, and shadows. We also propose an aircraft target change detection method by combining the multi-feature fusion change detection with target recognition using deep learning. The experimental results validate the excellent performance and reliability of our method.关键词：remote sensing;change detection;multi-feature fusion;aircraft target;high-resolution remote sensing images;IR-MAD;convolutional neural network- 739
- |
- 230
- |
- 7
发布时间：2021-10-18 -
摘要：This paper presents a multispectral-image-based model for eliminating the shadow effect on NDVI. This NDVI Shadow-Effect-Eliminating model (NSEE) is derived from simulated data and then applied on two Landsat 8 OLI images (one for the experiment and one for verification).NDVI plays a key role in the multispectral remote sensing retrieval of vegetation and has been widely used in many areas. However, the shadow normally existing in remote sensing images always influences the accuracy of NDVI. These effects will be transmitted in the process of further remote sensing retrieval, thereby resulting in errors. In this case, eliminating the shadow effect on vegetation is crucial and has a positive application value and high necessity.The difference between the shaded region and non-shaded regions in an image depends on how much solar irradiance these regions have received (i.e., supposing that the shaded region receives a solar diffuse radiation whereas the bright region receives total solar radiation, including direct and diffuse ones). The total solar irradiance (E0), solar direct irradiance (Ed), and solar diffuse irradiance (Ef) are simulated by using MODTRAN 4.0, the typical vegetation reflectance spectra (R) are selected from the spectra library in ENVI 5.3, and the radiances of vegetations (LR, LR′) in the shaded and non-shaded regions are calculated (using E0, Ef, and R. The mechanism behind the shadow effects on the NDVI of vegetation is analyzed by using the aforementioned simulated data. A normalized dark pixel index (NDPI) that shows high sensitivity in shadow detection and low relativity to NDVI is introduced. By analyzing the relationships between two sets of simulated NDVI (under solar diffuse radiation and under total solar radiation) of the same vegetation spectrum (to simulate shaded and non-shaded situations in remote sensing image), the NSEE model of NDVI Shadow-Effect-Eliminating(NSEE) is constructed to correct the NDVI in shaded regions based on the NDVI in the bright regions of an image.The NSEE model is applied on two Landsat 8 OLI images. The results show that, the NDVI values in the shaded regions are basically corrected to be normal, whereas the NDVIs in the bright regions remain stable in both the experimental and verification images. The NSEE model can also normalize the skewness of the NDVI statistical histogram caused by the shadow effect. The NDVI values of the experimental and verification images are compared pixel by pixel along the two transect lines, and the result shows that the reduction in NDVI due to shadow is eliminated and that the NDVI in the bright region belonging to either vegetation or non-vegetation pixels remains stable. The total RMSE is 0.067, thereby validating the effectiveness of the model is effective.The NSEE model effectively eliminates the shadow effects of shadow on the NDVI of vegetations. This model can also distinguish the NDVI-decreasing pixels of NDVI-decreasing (due to shadow effects) from those pixels with relatively low original NDVI values, thereby suggesting that the model fits well with land type. This model is entirely based on the image information itself, it can effectively maintain the relative spatial relations of NDVI, and effectively eliminate the influence of shadow. The proposed NESS model is based on a physical mechanism, it is concise and can be easily applied. This model only depends on the information of the multispectral image, does not require different data sources, and shows a convenient and efficient calculation.关键词：remote sensing;shaded vegetation canopy;NDVI;NDPI;shadow effects;Landsat 8 OLI;multispectral remote sensing- 817
- |
- 221
- |
- 6
发布时间：2021-10-18 -
摘要：Hyperspectral remote sensing images, which collect rich spectral and spatial information of observed targets, usually contain dozens to hundreds of narrow bands with wavelengths ranging from the visible light region to the near-infrared spectra. With such an abundant number of spectral features, hyperspectral images (HSI) allow us to distinguish different of objects or targets by rule and line. Unfortunately, such high-dimensionality data pose a challenge in data transmission, storage, and processing. Specifically, those HSIs with high redundancy information and strong correlation are prone to a Hughes phenomenon during the image classification process. Therefore, dimensionality reduction is necessary for targets classification. Moreover, without using prior label samples, unsupervised dimensionality reduction can effectively simplify the HSI feature space, and prevent the Hughes phenomenon in the targets classification.In this paper, the Jeffries–Matusita (JM) modified adaptive band selection (JM2ABS) method is proposed to extract proper features from HSI datasets. Generally speaking, a band that contains many information and demonstrates strong independence is a very important feature that helps unsupervised band selection methods to classify targets. The JM2ABS method considers both the information capacity and independence of HSI bands. Given the significant differences in the measurements of a band’s information capacity and its independence, we introduce the JM transform function to normalize the distributions of the information capacity and the independence of HSI data. Thus JM2ABS shows that both the information capacity and the independence are equally important in unsupervised dimensionality reduction.We also compare our proposed JM2ABS method against three typical methods, namely, the modified adaptive band selection method, the Laplacian score feature selection method, and the infinite feature selection method. By using random training samples, we perform supervised classification experiments on two kinds of HSI public datasets (linear and planar arrays). The results demonstrate that JM2ABS outperforms the other three typical methods in terms of Kappa value, overall classification accuracy, and average classification accuracy. Moreover, under a small number of bands, JM2ABS can reach a high and stable level regardless of the different datasets and different classifiers used.The proposed JM2ABS can select the proper features of HSI datasets for their classification application. The JM transform function is a kind of nonlinear distribution function that can standardize variables from different scales to the same. To demonstrate the feasibility of JM transform optimization, we set the same weight for the information content and the independence. In our future work, we will explore the similarities and differences between the information capacity and the independence in dimensionality reduction.关键词：remote sensing;Jeffries-Matusita transform;normalization;adaptive dimensionality reduction;unsupervised band selection;hyperspectral image- 602
- |
- 166
- |
- 1
发布时间：2021-10-18 -
摘要：Localization is one of the core technologies of indoor surveying and mapping services. To achieve an accurate indoor navigation and positioning, many indoor navigation and positioning technologies have been introduced. Given that visual sensors can generate an abundant amount of information at low cost and can be easily implemented, the space navigation and positioning method based on vision sensors has become a research hotspot. This paper proposes a robust method for measuring the mobile platform position measurement based on the coding graphic images that are captured by a monocular vision sensor.First, a series of coding graphics for the positioning of the carrier are designed. Coding graphics can be accurately identified through contour matching, and the coordinates of the center of graphics can be obtained by means of moment calculation and coding matching. However, due to the shooting angle, the coding image is deformed, thereby resulting in the residual errors of the image coordinates. Second, the Tukey weight factor model is used to calculate the weight according to the residual of the observed value. The value with a residual error of less than 1- 606
- |
- 86
- |
- 1
发布时间：2021-10-18
Technology and Methodology
-
摘要：Mesoscale eddies are active in the North Pacific Ocean (NPO) sensitive to the global variation of the atmosphere and ocean and directly affect the climate and coastal areas of the country. Therefore, the influence of the mesoscale eddies in the NPO on the marine ecological environment needs to be examined. The relationship between Sea Surface Temperature (SST) and Chlorophyll-a (Chl-a) concentration inside these eddies as well as the response mechanism of these eddies to local ecological processes also warrant further research.In this paper, the Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO)-merged satellite altimeter data are used to identify and track 992 mesoscale eddies in the NPO during the years of 2007-2012, including 442 Cyclonic Eddies (CEs) and 550 Anticyclonic Eddies (AEs), in the NPO between 2007 and 2012. The spatial and temporal distributions of the SST and the Chl-a concentration inside these eddies are analyzed via Operational Sea surface Temperature and sea Ice Analysis (OSTIA) SST and MODIS data, and the variability of these parameters inside typical eddies is examined.The results show that the temperature intensity of CEs (ICE) has a higher tendency to demonstrate monthly variations compared with that of AEs (IAE). The seasonal variation of ICE tends to contrast that of IAE. Specifically, ICE shows an obvious annual variation, whereas IAE does not. Stronger ICE and IAE are observed in the Kuroshio Extension intensively. Both CEs and AEs have similar monthly variation tendencies in the temporal and spatial distributions of eddy Chl-a. The annual trends of Chl-a in both AEs and CEs are ambiguous. For AEs and CEs, the highest Chl-a concentration is observed in the high-latitude region. A study of the relationship between the eddy SST and eddy dynamic parameters (e.g., amplitude, vorticity, and Eddy Kinetic Energy (EKE)) reveals that SST inner AEs are either positively or negatively correlated to amplitude with a uniform distribution in space. A negative correlation of SST inner CEs can be observed in the Kuroshio Extension, whereas a positive correlation is observed in offshore areas of California. A positive correlation is more frequently observed in AEs than in CEs. The correlation of SST with vorticity in AEs is either positive or negative, while in CEs, SST shows a negative correlation with vorticity. The correlation of SST with EKE in AEs is either positive or negative, but such correlation is only positive uniquely in CEs. The Chl-a concentration in AEs is positively correlated with amplitude and has a uniform distribution in space. In CEs, Chl-a concentration shows a positive correlation with amplitude in the Kuroshio Extension and in Alaska Bay. Chl-a concentration also shows a positive correlation with vorticity in both AEs and CEs, a positive correlation with EKE in AEs, and either a positive or negative correlation with EKE in CEs.We conclude that SST demonstrates obvious monthly and annual variation tendencies in CEs, and high ICE and IAE values are distributed in the Kuroshio Extension of the NPO. The Chl-a concentration in CEs and AEs demonstrate a similar monthly variation tendency and an ambiguous annual variation. Eddies with a high Chl-a concentration are mainly located in the high-latitude region of the NPO. The relationship between the eddy SST and the eddy dynamic parameters is treated as a local feature of the NPO, and a positive relationship between eddy Chl-a concentration and eddy dynamic parameters is observed in the Kuroshio Extension and Alaska Bay. The influence of mesoscale eddies on the ecological processes is related to the type and eddy lifetime of eddies in the NPO.关键词：remote sensing;Chlorophyll concentration;sea surface temperature;temperature intensity;mesoscale eddy;North Pacific Ocean- 696
- |
- 229
- |
- 2
发布时间：2021-10-18 -
摘要：Disastrous weather caused by Severe Convective Clouds (SCC), such as short-term heavy rainfall, hail, thunderstorms, squall line and tornado, has a serious impact on the life and production over sea surface. Therefore it is very essential to monitor SCC duly and accurately. However, the traditional methods are difficult to meet the requirements due to the short life cycle of severe convective weather and the lack of meteorological stations over sea surface. The geostationary meteorological satellite images have wide coverage area and high temporal resolution, so it has become an important means to monitor severe convective weather. With the rapid development of deep learning, it has been applied in the field of remote sensing image recognition.This paper proposes a method for automatic identification of SCC from geostationary meteorological satellite images based on Deep Belief Networks (DBN). First, in order to identify the characteristic parameters that can be used for recognizing severe convective clouds, the severe convective clouds are analyzed from spectral features and texture features. Second, building automatic recognition algorithm based on deep belief networks which consists of multiple Restricted Boltzmann Machines (RBM) and a softmax classifier. And it is divided into two stages: unsupervised pre training and supervised fine tuning. This method is summarized in the following four steps: (1)data preprocessing: images splicing and region cutting; (2)extracting features and constructing sample sets: extracting the spectral features: TBB13, TBB08-TBB13 and TBB13-TBB15. And extracting the texture features: Energy and Contrast, which extracted based on spectral feature TBB08-TBB13. And then constructing the sample sets automatically referring to CloudSat satellite cloud classification products;(3)training DBN model: determining the structure and parameters of the model, including parameters of RBM and depth of DBN;(4)recognizing SCC using DBN model, and doing the postprocessing: identifying the severe convective clouds using DBN model, and processing the recognition results by category merging, closing operation and edge detection.The Himawari-8 satellite image data and CloudSat cloud classification products from March to May in 2017 were used in the experiment. The research area is 70°E—150°E, 0°N—55°N. After training, the structure of DBN model was setting to 245-140-140-140-135-135-135-9. The accuracy of the DBN model is evaluated with the modified test sample, the Critical Success Index (CSI) is 71.28%, the Probability of Detection (POD) is 84.83%, and the False Alarm Ratio (FAR) is 18.31%. Compared with single band threshold method, multi band threshold method and SVM, the method proposed in this paper can effectively improve the recognition accuracy.The experiment results show that all kinds of severe convective clouds in different phases from initiation to dissipation can be effectively identified. Consequently, it has the advantages of finding severe convective weather in advance. The majority of the cirrus can be removed, but the results of recognition still contain some cirrus spissatus, and the recognition of the cloud edge is not accurate enough, which will be two directions for future research.关键词：remote sensing;severe convective clouds;deep belief networks;geostationary meteorological satellite;spectral feature;texture feature- 1078
- |
- 276
- |
- 6
发布时间：2021-10-18
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