Accurate fine-grained geospatial scene classification using remote sensing imagery is essential for a wide range of applications.However,existing approaches often rely on manually zooming remote sensing images at diff...Accurate fine-grained geospatial scene classification using remote sensing imagery is essential for a wide range of applications.However,existing approaches often rely on manually zooming remote sensing images at different scales to create typical scene samples.This approach fails to adequately support the fixed-resolution image interpretation requirements in real-world scenarios.To address this limitation,we introduce the million-scale fine-grained geospatial scene classification dataset(MEET),which contains over 1.03 million zoom-free remote sensing scene samples,manually annotated into 80 fine-grained categories.In MEET,each scene sample follows a scene-in-scene layout,where the central scene serves as the reference,and auxiliary scenes provide crucial spatial context for fine-grained classification.Moreover,to tackle the emerging challenge of scene-in-scene classification,we present the context-aware transformer(CAT),a model specifically designed for this task,which adaptively fuses spatial context to accurately classify the scene samples.CAT adaptively fuses spatial context to accurately classify the scene samples by learning attentional features that capture the relationships between the center and auxiliary scenes.Based on MEET,we establish a comprehensive benchmark for fine-grained geospatial scene classification,evaluating CAT against 11 competitive baselines.The results demonstrate that CAT significantly outperforms these baselines,achieving a 1.88%higher balanced accuracy(BA)with the Swin-Large backbone,and a notable 7.87%improvement with the Swin-Huge backbone.Further experiments validate the effectiveness of each module in CAT and show the practical applicability of CAT in the urban functional zone mapping.The source code and dataset will be publicly available at https://jerrywyn.github.io/project/MEET.html.展开更多
Remote sensing scene image classification is a prominent research area within remote sensing.Deep learningbased methods have been extensively utilized and have shown significant advancements in this field.Recent progr...Remote sensing scene image classification is a prominent research area within remote sensing.Deep learningbased methods have been extensively utilized and have shown significant advancements in this field.Recent progress in these methods primarily focuses on enhancing feature representation capabilities to improve performance.The challenge lies in the limited spatial resolution of small-sized remote sensing images,as well as image blurring and sparse data.These factors contribute to lower accuracy in current deep learning models.Additionally,deeper networks with attention-based modules require a substantial number of network parameters,leading to high computational costs and memory usage.In this article,we introduce ERSNet,a lightweight novel attention-guided network for remote sensing scene image classification.ERSNet is constructed using a deep separable convolutional network and incorporates an attention mechanism.It utilizes spatial attention,channel attention,and channel self-attention to enhance feature representation and accuracy,while also reducing computational complexity and memory usage.Experimental results indicate that,compared to existing state-of-the-art methods,ERSNet has a significantly lower parameter count of only 1.2 M and reduced Flops.It achieves the highest classification accuracy of 99.14%on the EuroSAT dataset,demonstrating its suitability for application on mobile terminal devices.Furthermore,experimental results from the UCMerced land use dataset and the Brazilian coffee scene also confirm the strong generalization ability of this method.展开更多
Mudflat vegetation plays a crucial role in the ecological function of wetland environment,and obtaining its fine spatial distri-bution is of great significance for wetland protection and management.Remote sensing tech...Mudflat vegetation plays a crucial role in the ecological function of wetland environment,and obtaining its fine spatial distri-bution is of great significance for wetland protection and management.Remote sensing techniques can realize the rapid extraction of wetland vegetation over a large area.However,the imaging of optical sensors is easily restricted by weather conditions,and the backs-cattered information reflected by Synthetic Aperture Radar(SAR)images is easily disturbed by many factors.Although both data sources have been applied in wetland vegetation classification,there is a lack of comparative study on how the selection of data sources affects the classification effect.This study takes the vegetation of the tidal flat wetland in Chongming Island,Shanghai,China,in 2019,as the research subject.A total of 22 optical feature parameters and 11 SAR feature parameters were extracted from the optical data source(Sentinel-2)and SAR data source(Sentinel-1),respectively.The performance of optical and SAR data and their feature paramet-ers in wetland vegetation classification was quantitatively compared and analyzed by different feature combinations.Furthermore,by simulating the scenario of missing optical images,the impact of optical image missing on vegetation classification accuracy and the compensatory effect of integrating SAR data were revealed.Results show that:1)under the same classification algorithm,the Overall Accuracy(OA)of the combined use of optical and SAR images was the highest,reaching 95.50%.The OA of using only optical images was slightly lower,while using only SAR images yields the lowest accuracy,but still achieved 86.48%.2)Compared to using the spec-tral reflectance of optical data and the backscattering coefficient of SAR data directly,the constructed optical and SAR feature paramet-ers contributed to improving classification accuracy.The inclusion of optical(vegetation index,spatial texture,and phenology features)and SAR feature parameters(SAR index and SAR texture features)in the classification algorithm resulted in an OA improvement of 4.56%and 9.47%,respectively.SAR backscatter,SAR index,optical phenological features,and vegetation index were identified as the top-ranking important features.3)When the optical data were missing continuously for six months,the OA dropped to a minimum of 41.56%.However,when combined with SAR data,the OA could be improved to 71.62%.This indicates that the incorporation of SAR features can effectively compensate for the loss of accuracy caused by optical image missing,especially in regions with long-term cloud cover.展开更多
Considering the important applications in the military and the civilian domain, ship detection and classification based on optical remote sensing images raise considerable attention in the sea surface remote sensing f...Considering the important applications in the military and the civilian domain, ship detection and classification based on optical remote sensing images raise considerable attention in the sea surface remote sensing filed. This article collects the methods of ship detection and classification for practically testing in optical remote sensing images, and provides their corresponding feature extraction strategies and statistical data. Basic feature extraction strategies and algorithms are analyzed associated with their performance and application in ship detection and classification.Furthermore, publicly available datasets that can be applied as the benchmarks to verify the effectiveness and the objectiveness of ship detection and classification methods are summarized in this paper. Based on the analysis, the remaining problems and future development trends are provided for ship detection and classification methods based on optical remote sensing images.展开更多
Dae to complex terrain of the Loess Plateau, the classification accuracy is unsatisfactory when a single supervised classification is used in die remote sensing investigation of the sloping field. Taking the loess hil...Dae to complex terrain of the Loess Plateau, the classification accuracy is unsatisfactory when a single supervised classification is used in die remote sensing investigation of the sloping field. Taking the loess hill and gully area of northern Shaanxi Province as a test area, a research was conducted to extract sloping field and other land use categories by applying an integrated classification. Based on an integration of supervised classification aad unsupervised classification, sampling method is remarkably unproved. The results show that the classification accuracy is satisfactory by the method and is of critical significance in obtaining up-to-date information of the sloping field, which should be helpful in the state key project of converting farmland to forest and grassland on slope land in this area. This research sought to improve the application accuracy of image classification in complex terrain areas.展开更多
Although the Convolutional Neural Network(CNN)has shown great potential for land cover classification,the frequently used single-scale convolution kernel limits the scope of informa-tion extraction.Therefore,we propos...Although the Convolutional Neural Network(CNN)has shown great potential for land cover classification,the frequently used single-scale convolution kernel limits the scope of informa-tion extraction.Therefore,we propose a Multi-Scale Fully Convolutional Network(MSFCN)with a multi-scale convolutional kernel as well as a Channel Attention Block(CAB)and a Global Pooling Module(GPM)in this paper to exploit discriminative representations from two-dimensional(2D)satellite images.Meanwhile,to explore the ability of the proposed MSFCN for spatio-temporal images,we expand our MSFCN to three-dimension using three-dimensional(3D)CNN,capable of harnessing each land cover category’s time series interac-tion from the reshaped spatio-temporal remote sensing images.To verify the effectiveness of the proposed MSFCN,we conduct experiments on two spatial datasets and two spatio-temporal datasets.The proposed MSFCN achieves 60.366%on the WHDLD dataset and 75.127%on the GID dataset in terms of mIoU index while the figures for two spatio-temporal datasets are 87.753%and 77.156%.Extensive comparative experiments and abla-tion studies demonstrate the effectiveness of the proposed MSFCN.展开更多
The Sanjiang Plain, where nearly 20 kinds of wetlands exist now, is one of the largest wetlands distributed area of wetlands in China. To identify each of them and pick up them separately by means of automatic interpr...The Sanjiang Plain, where nearly 20 kinds of wetlands exist now, is one of the largest wetlands distributed area of wetlands in China. To identify each of them and pick up them separately by means of automatic interpretation of remote sensing from TM Landsat images is extremely important. However, most of the types of wetlands can not be divided each other due to the similarity and the illegibility of the wetland spectrum shown in TM images. Special disposals to remote sensing images include the spectrum enhancement of wetland information, the pseudo color composite of TM images of different bands and the algebra enhancement of TM images. By this way some kinds of wetlands such as Sparganium stoloniferum and Bolboschoenus maritimus can be identified. But in many cases, these methods are still insufficient because of the noise brought from the atmosphere transportation and so on. The physical features of wetlands reflecting the diversification of spectrum information of wetlands, which include the spatial temporal characteristics of the wetlands distribution, the landscape differences of wetlands from season to season, the growing environment and the vertical structure of wetlands vegetation and so on, must be taken into consideration. Besides these, the artificial alteration to spatial structure of wetlands such as the exploitation of some types of them can be also used as important symbols of wetlands identification from remote sensing images. On the basis of the above geographics analysis, a set of wetlands classification models of remote sensing could be established, and many types of wetlands such as paddy field, reed swamp, peat mire, meadow, CAREX marsh and paludification meadow and so on, will be distinguished consequently. All the ways of geographical analysis and model establishment will be given in detail in this article.展开更多
How to recognize targets with similar appearances from remote sensing images(RSIs) effectively and efficiently has become a big challenge. Recently, convolutional neural network(CNN) is preferred in the target classif...How to recognize targets with similar appearances from remote sensing images(RSIs) effectively and efficiently has become a big challenge. Recently, convolutional neural network(CNN) is preferred in the target classification due to the powerful feature representation ability and better performance. However,the training and testing of CNN mainly rely on single machine.Single machine has its natural limitation and bottleneck in processing RSIs due to limited hardware resources and huge time consuming. Besides, overfitting is a challenge for the CNN model due to the unbalance between RSIs data and the model structure.When a model is complex or the training data is relatively small,overfitting occurs and leads to a poor predictive performance. To address these problems, a distributed CNN architecture for RSIs target classification is proposed, which dramatically increases the training speed of CNN and system scalability. It improves the storage ability and processing efficiency of RSIs. Furthermore,Bayesian regularization approach is utilized in order to initialize the weights of the CNN extractor, which increases the robustness and flexibility of the CNN model. It helps prevent the overfitting and avoid the local optima caused by limited RSI training images or the inappropriate CNN structure. In addition, considering the efficiency of the Na¨?ve Bayes classifier, a distributed Na¨?ve Bayes classifier is designed to reduce the training cost. Compared with other algorithms, the proposed system and method perform the best and increase the recognition accuracy. The results show that the distributed system framework and the proposed algorithms are suitable for RSIs target classification tasks.展开更多
Wetland research has become a hot spot linking multiple disciplines presently. Wetland classification and mapping is the basis for wetland research. It is difficult to generate wetland data sets using traditional meth...Wetland research has become a hot spot linking multiple disciplines presently. Wetland classification and mapping is the basis for wetland research. It is difficult to generate wetland data sets using traditional methods because of the low accessibility of wetlands, hence remote sensing data have become one of the primary data sources in wetland research. This paper presents a case study conducted at the core area of Honghe National Nature Reserve in the Sanjiang Plain, Northeast China. In this study, three images generated by airship, from Thematic Mapper and from SPOT 5 were selected to produce wetland maps at three different wetland landscape levels. After assessing classification accuracies of the three maps, we compared the different wetland mapping results of 11 plant communities to the airship image, 6 plant ecotypes to the TM image and 9 landscape classifications to the SPOT 5 image. We discussed the different characteristics of the hierarchical ecosystem classifications based on the spatial scales of the different images. The results indicate that spatial scales of remote sensing data have an important link to the hierarchies of wetland plant ecosystems displayed on the wetland landscape maps. The richness of wetland landscape information derived from an image closely relates to its spatial resolution. This study can enrich the ecological classification methods and mapping techniques dealing with the spatial scales of different remote sensing images. With a better understanding of classification accuracies in mapping wetlands by using different scales of remote sensing data, we can make an appropriate approach for dealing with the scale issue of remote sensing images.展开更多
Landform classification is commonly done using topographic altitude only. However practice indicates that locations at a same altitude may have distinctly different landforms, depending on characteristics of soils und...Landform classification is commonly done using topographic altitude only. However practice indicates that locations at a same altitude may have distinctly different landforms, depending on characteristics of soils underneath those locations. The objectives of this study were to： 1） develop a landform classification approach that is based on both altitude and soil characteristic; and 2） use this approach to determine landforms within a watershed located in northern Ordos Plateau of China. Using data collected at 134 out of 200 sampling sites, this study determined that D10 （the diameter of soil particles 10% finer by weight） and long-term average soil moisture acquired in 2010, which can be estimated at reasonable accuracy from remote sensing imagery, can be used to represent soil characteristics of the study watershed. Also, the sampling data revealed that this watershed consists of nine classes of landforms, namely mobile dune （MD）, mobile semi-mobile dune （SMD）, rolling fixed semi-fixed dune （RFD）, flat sandy land （FD）, grassy sandy land （GS）, bedrock （BR）, flat sandy bedrock （FSB）, valley agricultural land （VA）, and swamp and salt lake （SW）. A set of logistic regression equations were derived using data collected at the 134 sampling sites and verified using data at the remaining 66 sites. The verification indicated that these equations have moderate classification accuracy （Kappa coefficients K 〉 43%）. The results revealed that the dominant classes in the study watershed are FD （36.3%）, BR （27.0%）, and MD （23.5%）, while the other six types of landforms （i.e., SMD, RFD, GS, FSB, VA, and SW） in combination account for 13.2%. Further, the landforms determined in this study were compared with the classes presented by a geologically-based classification map. The comparison indicated that the geologically-based classification could not identify multiple landforms within a class that are dependent upon soil characteristics.展开更多
Classification is always the key point in the field of remote sensing. Fuzzy c-Means is a traditional clustering algorithm that has been widely used in fuzzy clustering. However, this algorithm usually has some weakne...Classification is always the key point in the field of remote sensing. Fuzzy c-Means is a traditional clustering algorithm that has been widely used in fuzzy clustering. However, this algorithm usually has some weaknesses, such as the problems of falling into a local minimum, and it needs much time to accomplish the classification for a large number of data. In order to overcome these shortcomings and increase the classifi-cation accuracy, Gustafson-Kessel (GK) and Gath-Geva (GG) algorithms are proposed to improve the tradi-tional FCM algorithm which adopts Euclidean distance norm in this paper. The experimental result shows that these two methods are able to detect clusters of varying shapes, sizes and densities which FCM cannot do. Moreover, they can improve the classification accuracy of remote sensing images.展开更多
Spatial and temporal informationon urban infrastructure is essential and requires various land-cover/land-use planning and management applications.Besides,a change in infrastructure has a direct impact on other land-c...Spatial and temporal informationon urban infrastructure is essential and requires various land-cover/land-use planning and management applications.Besides,a change in infrastructure has a direct impact on other land-cover and climatic conditions.This study assessed changes in the rate and spatial distribution of Peshawar district’s infrastructure and its effects on Land Surface Temperature(LST)during the years 1996 and 2019.For this purpose,firstly,satellite images of bands7 and 8 ETM+(Enhanced Thematic Mapper)plus and OLI(Operational Land Imager)of 30 m resolution were taken.Secondly,for classification and image processing,remote sensing(RS)applications ENVI(Environment for Visualising Images)and GIS(Geographic Information System)were used.Thirdly,for better visualization and more in-depth analysis of land sat images,pre-processing techniques were employed.For Land use and Land cover(LU/LC)four types of land cover areas were identified-vegetation area,water cover,urbanized area,and infertile land for the years under research.The composition of red,green,and near infra-red bands was used for supervised classification.Classified images were extracted for analyzing the relative infrastructure change.A comparative analysis for the classification of images is performed for SVM(Support Vector Machine)and ANN(Artificial Neural Network).Based on analyzing these images,the result shows the rise in the average temperature from 30.04℃ to 45.25℃.This only possible reason is the increase in the built-up area from 78.73 to 332.78 Area km^(2) from 1996 to 2019.It has also been witnessed that the city’s sides are hotter than the city’s center due to the barren land on the borders.展开更多
This paper presents a new kind of back propagation neural network(BPNN)based on rough sets,called rough back propagation neural network(RBPNN).The architecture and training method of RBPNN are presented and the survey...This paper presents a new kind of back propagation neural network(BPNN)based on rough sets,called rough back propagation neural network(RBPNN).The architecture and training method of RBPNN are presented and the survey and analysis of RBPNN for the classification of remote sensing multi_spectral image is discussed.The successful application of RBPNN to a land cover classification illustrates the simple computation and high accuracy of the new neural network and the flexibility and practicality of this new approach.展开更多
Deep learning significantly improves the accuracy of remote sensing image scene classification,benefiting from the large-scale datasets.However,annotating the remote sensing images is time-consuming and even tough for...Deep learning significantly improves the accuracy of remote sensing image scene classification,benefiting from the large-scale datasets.However,annotating the remote sensing images is time-consuming and even tough for experts.Deep neural networks trained using a few labeled samples usually generalize less to new unseen images.In this paper,we propose a semi-supervised approach for remote sensing image scene classification based on the prototype-based consistency,by exploring massive unlabeled images.To this end,we,first,propose a feature enhancement module to extract discriminative features.This is achieved by focusing the model on the foreground areas.Then,the prototype-based classifier is introduced to the framework,which is used to acquire consistent feature representations.We conduct a series of experiments on NWPU-RESISC45 and Aerial Image Dataset(AID).Our method improves the State-Of-The-Art(SOTA)method on NWPU-RESISC45 from 92.03%to 93.08%and on AID from 94.25%to 95.24%in terms of accuracy.展开更多
NOAA-AVHRR data have been more and more used by scientists because of its short temporal resolution,large scope, inexpensive cost and broad wave bands. On macro and middle scale of vegetation remote sensing, NOAAAVHRR...NOAA-AVHRR data have been more and more used by scientists because of its short temporal resolution,large scope, inexpensive cost and broad wave bands. On macro and middle scale of vegetation remote sensing, NOAAAVHRR possesses an advantage when compared with other satellites. However, because NOAA-AVHRR also problem of low resolution, data distortion and geometrical distortion, in the area of application of NOAA-AVHRR in largescale vegetation - mapping, the accuracy of vegetation classification should be improved. This paper discuss the feasibilityof integrating the geographic data in GIS(Geographical Information System) and remotely sensed data in GIS. Under theenvironment of GIS, temperature, precipitation and elevation, which serve as main factors affecting vegetation growth,were processed by a mathematical model and qualified into geographic image under a certain grid system. The geographicimage were overlaid to the NOAA-AVHRR data which had been compressed and processed. In order to evaluate the usefulness of geographic data for vegetation classification, the area under study was digitally classified by two groups of interpreter: the proposed methodology using maximum likelihood classification assisted by the geographic database and a conventional maximum likelihood classification only. Both result were compared using Kappa statistics. The indices to both theproposed and the conventional digital classification methodology were 0. 668(yew good) and 0. 563(good), respetively.The geographic database rendered an improvement over the conventional digital classification. Furthermore, in this study,some problems related to multi-sources data integration are also discussed.展开更多
The development of image classification is one of the most important research topics in remote sensing. The prediction accuracy depends not only on the appropriate choice of the machine learning method but also on the...The development of image classification is one of the most important research topics in remote sensing. The prediction accuracy depends not only on the appropriate choice of the machine learning method but also on the quality of the training datasets. However, real-world data is not perfect and often suffers from noise. This paper gives an overview of noise filtering methods. Firstly, the types of noise and the consequences of class noise on machine learning are presented. Secondly, class noise handling methods at both the data level and the algorithm level are introduced. Then ensemble-based class noise handling methods including class noise removal, correction, and noise robust ensemble learners are presented. Finally, a summary of existing data-cleaning techniques is given.展开更多
The classification of hyperspectral remote sensing data is an important problem theoretically and practically. With the increase of spectral bands, the separability of objects on remote sensing image should be improve...The classification of hyperspectral remote sensing data is an important problem theoretically and practically. With the increase of spectral bands, the separability of objects on remote sensing image should be improved. But the effects of traditional algorithm on feature extraction such as principal component analysis(PCA) is not so good for hyperspectral image. The key problem is that PCA can only represent the linear structure of data set; while the data clouds of different objects on hyperspectral image usually distribute on a nonlinear manifold. This paper established an algorithm of nonlinear feature extraction named as nonlinear principal poly lines, based on the algorithm, a classifier is constructed and the classification accuracy of hyperspectral image can be improved.展开更多
ANNs （Artificial neural networks） are used extensively in remote sensing image processing. It has been proven that BPNNs （back-propagation neural networks） have high attainable classification accuracy. However, th...ANNs （Artificial neural networks） are used extensively in remote sensing image processing. It has been proven that BPNNs （back-propagation neural networks） have high attainable classification accuracy. However, there is a noticeable variation in the achieved accuracies due to different network designs and implementations. Hence, researchers usually need to conduct several experimental trials before they can finalize the network design. This is a time consuming process which significantly reduces the effectiveness of using BPNNs and the final design may still not be optimal. Therefore, there is a need to see whether there are some common guidelines for effective design and implementation of BPNNs. With this aim in mind, this paper attempts to find and summarize the common guidelines suggested by different authors through literature review and discussion of the findings. To provide readers with background and contextual information, some ANN fundamentals are also introduced.展开更多
Recently,the convolutional neural network(CNN)has been dom-inant in studies on interpreting remote sensing images(RSI).However,it appears that training optimization strategies have received less attention in relevant ...Recently,the convolutional neural network(CNN)has been dom-inant in studies on interpreting remote sensing images(RSI).However,it appears that training optimization strategies have received less attention in relevant research.To evaluate this problem,the author proposes a novel algo-rithm named the Fast Training CNN(FST-CNN).To verify the algorithm’s effectiveness,twenty methods,including six classic models and thirty archi-tectures from previous studies,are included in a performance comparison.The overall accuracy(OA)trained by the FST-CNN algorithm on the same model architecture and dataset is treated as an evaluation baseline.Results show that there is a maximal OA gap of 8.35%between the FST-CNN and those methods in the literature,which means a 10%margin in performance.Meanwhile,all those complex roadmaps,e.g.,deep feature fusion,model combination,model ensembles,and human feature engineering,are not as effective as expected.It reveals that there was systemic suboptimal perfor-mance in the previous studies.Most of the CNN-based methods proposed in the previous studies show a consistent mistake,which has made the model’s accuracy lower than its potential value.The most important reasons seem to be the inappropriate training strategy and the shift in data distribution introduced by data augmentation(DA).As a result,most of the performance evaluation was conducted based on an inaccurate,suboptimal,and unfair result.It has made most of the previous research findings questionable to some extent.However,all these confusing results also exactly demonstrate the effectiveness of FST-CNN.This novel algorithm is model-agnostic and can be employed on any image classification model to potentially boost performance.In addition,the results also show that a standardized training strategy is indeed very meaningful for the research tasks of the RSI-SC.展开更多
Remote sensing technique, replacing conventional sonar bathymetry technique, has become an effective complementary method of mapping submarine terrain where special conditions make the sonar technique difficult to be ...Remote sensing technique, replacing conventional sonar bathymetry technique, has become an effective complementary method of mapping submarine terrain where special conditions make the sonar technique difficult to be carried out. At the same time, as one kind of data set, multispectral remote sensing data has the disadvantage of being influenced by the variable bottom types in shallow seawater, when it is applied in bathymetry. This paper puts forward a new method to extract water depth information from multispectral data, considering the bottom classification and the true water depth accuracy. That is the Principal Component Analysis (PCA) technique based on the bottom classification. By the least square regression with significance, the experiment near Qingdao City has obtained more satisfactory bathymetry accuracy than that of the traditional single-band method, with the mean absolute error about 2.57m.展开更多
基金supported by the National Natural Science Foundation of China(42030102,42371321).
文摘Accurate fine-grained geospatial scene classification using remote sensing imagery is essential for a wide range of applications.However,existing approaches often rely on manually zooming remote sensing images at different scales to create typical scene samples.This approach fails to adequately support the fixed-resolution image interpretation requirements in real-world scenarios.To address this limitation,we introduce the million-scale fine-grained geospatial scene classification dataset(MEET),which contains over 1.03 million zoom-free remote sensing scene samples,manually annotated into 80 fine-grained categories.In MEET,each scene sample follows a scene-in-scene layout,where the central scene serves as the reference,and auxiliary scenes provide crucial spatial context for fine-grained classification.Moreover,to tackle the emerging challenge of scene-in-scene classification,we present the context-aware transformer(CAT),a model specifically designed for this task,which adaptively fuses spatial context to accurately classify the scene samples.CAT adaptively fuses spatial context to accurately classify the scene samples by learning attentional features that capture the relationships between the center and auxiliary scenes.Based on MEET,we establish a comprehensive benchmark for fine-grained geospatial scene classification,evaluating CAT against 11 competitive baselines.The results demonstrate that CAT significantly outperforms these baselines,achieving a 1.88%higher balanced accuracy(BA)with the Swin-Large backbone,and a notable 7.87%improvement with the Swin-Huge backbone.Further experiments validate the effectiveness of each module in CAT and show the practical applicability of CAT in the urban functional zone mapping.The source code and dataset will be publicly available at https://jerrywyn.github.io/project/MEET.html.
文摘Remote sensing scene image classification is a prominent research area within remote sensing.Deep learningbased methods have been extensively utilized and have shown significant advancements in this field.Recent progress in these methods primarily focuses on enhancing feature representation capabilities to improve performance.The challenge lies in the limited spatial resolution of small-sized remote sensing images,as well as image blurring and sparse data.These factors contribute to lower accuracy in current deep learning models.Additionally,deeper networks with attention-based modules require a substantial number of network parameters,leading to high computational costs and memory usage.In this article,we introduce ERSNet,a lightweight novel attention-guided network for remote sensing scene image classification.ERSNet is constructed using a deep separable convolutional network and incorporates an attention mechanism.It utilizes spatial attention,channel attention,and channel self-attention to enhance feature representation and accuracy,while also reducing computational complexity and memory usage.Experimental results indicate that,compared to existing state-of-the-art methods,ERSNet has a significantly lower parameter count of only 1.2 M and reduced Flops.It achieves the highest classification accuracy of 99.14%on the EuroSAT dataset,demonstrating its suitability for application on mobile terminal devices.Furthermore,experimental results from the UCMerced land use dataset and the Brazilian coffee scene also confirm the strong generalization ability of this method.
基金Under the auspices of the National Key Research and Development Program of China(No.2023YFC3208500)Shanghai Municipal Natural Science Foundation(No.22ZR1421500)+3 种基金National Natural Science Foundation of China(No.U2243207)National Science and Technology Basic Resources Survey Project(No.2023FY01001)Open Research Fund of State Key Laboratory of Estuarine and Coastal Research(No.SKLEC-KF202406)Project from Science and Technology Commission of Shanghai Municipality(No.22DZ1202700)。
文摘Mudflat vegetation plays a crucial role in the ecological function of wetland environment,and obtaining its fine spatial distri-bution is of great significance for wetland protection and management.Remote sensing techniques can realize the rapid extraction of wetland vegetation over a large area.However,the imaging of optical sensors is easily restricted by weather conditions,and the backs-cattered information reflected by Synthetic Aperture Radar(SAR)images is easily disturbed by many factors.Although both data sources have been applied in wetland vegetation classification,there is a lack of comparative study on how the selection of data sources affects the classification effect.This study takes the vegetation of the tidal flat wetland in Chongming Island,Shanghai,China,in 2019,as the research subject.A total of 22 optical feature parameters and 11 SAR feature parameters were extracted from the optical data source(Sentinel-2)and SAR data source(Sentinel-1),respectively.The performance of optical and SAR data and their feature paramet-ers in wetland vegetation classification was quantitatively compared and analyzed by different feature combinations.Furthermore,by simulating the scenario of missing optical images,the impact of optical image missing on vegetation classification accuracy and the compensatory effect of integrating SAR data were revealed.Results show that:1)under the same classification algorithm,the Overall Accuracy(OA)of the combined use of optical and SAR images was the highest,reaching 95.50%.The OA of using only optical images was slightly lower,while using only SAR images yields the lowest accuracy,but still achieved 86.48%.2)Compared to using the spec-tral reflectance of optical data and the backscattering coefficient of SAR data directly,the constructed optical and SAR feature paramet-ers contributed to improving classification accuracy.The inclusion of optical(vegetation index,spatial texture,and phenology features)and SAR feature parameters(SAR index and SAR texture features)in the classification algorithm resulted in an OA improvement of 4.56%and 9.47%,respectively.SAR backscatter,SAR index,optical phenological features,and vegetation index were identified as the top-ranking important features.3)When the optical data were missing continuously for six months,the OA dropped to a minimum of 41.56%.However,when combined with SAR data,the OA could be improved to 71.62%.This indicates that the incorporation of SAR features can effectively compensate for the loss of accuracy caused by optical image missing,especially in regions with long-term cloud cover.
文摘Considering the important applications in the military and the civilian domain, ship detection and classification based on optical remote sensing images raise considerable attention in the sea surface remote sensing filed. This article collects the methods of ship detection and classification for practically testing in optical remote sensing images, and provides their corresponding feature extraction strategies and statistical data. Basic feature extraction strategies and algorithms are analyzed associated with their performance and application in ship detection and classification.Furthermore, publicly available datasets that can be applied as the benchmarks to verify the effectiveness and the objectiveness of ship detection and classification methods are summarized in this paper. Based on the analysis, the remaining problems and future development trends are provided for ship detection and classification methods based on optical remote sensing images.
基金National Nature Science Foundation of China,No.40271089High-visiting scholar fund of The Key Laboratory of LIESMARS
文摘Dae to complex terrain of the Loess Plateau, the classification accuracy is unsatisfactory when a single supervised classification is used in die remote sensing investigation of the sloping field. Taking the loess hill and gully area of northern Shaanxi Province as a test area, a research was conducted to extract sloping field and other land use categories by applying an integrated classification. Based on an integration of supervised classification aad unsupervised classification, sampling method is remarkably unproved. The results show that the classification accuracy is satisfactory by the method and is of critical significance in obtaining up-to-date information of the sloping field, which should be helpful in the state key project of converting farmland to forest and grassland on slope land in this area. This research sought to improve the application accuracy of image classification in complex terrain areas.
基金supported by the National Natural Science Foundation of China[grant number 41671452].
文摘Although the Convolutional Neural Network(CNN)has shown great potential for land cover classification,the frequently used single-scale convolution kernel limits the scope of informa-tion extraction.Therefore,we propose a Multi-Scale Fully Convolutional Network(MSFCN)with a multi-scale convolutional kernel as well as a Channel Attention Block(CAB)and a Global Pooling Module(GPM)in this paper to exploit discriminative representations from two-dimensional(2D)satellite images.Meanwhile,to explore the ability of the proposed MSFCN for spatio-temporal images,we expand our MSFCN to three-dimension using three-dimensional(3D)CNN,capable of harnessing each land cover category’s time series interac-tion from the reshaped spatio-temporal remote sensing images.To verify the effectiveness of the proposed MSFCN,we conduct experiments on two spatial datasets and two spatio-temporal datasets.The proposed MSFCN achieves 60.366%on the WHDLD dataset and 75.127%on the GID dataset in terms of mIoU index while the figures for two spatio-temporal datasets are 87.753%and 77.156%.Extensive comparative experiments and abla-tion studies demonstrate the effectiveness of the proposed MSFCN.
文摘The Sanjiang Plain, where nearly 20 kinds of wetlands exist now, is one of the largest wetlands distributed area of wetlands in China. To identify each of them and pick up them separately by means of automatic interpretation of remote sensing from TM Landsat images is extremely important. However, most of the types of wetlands can not be divided each other due to the similarity and the illegibility of the wetland spectrum shown in TM images. Special disposals to remote sensing images include the spectrum enhancement of wetland information, the pseudo color composite of TM images of different bands and the algebra enhancement of TM images. By this way some kinds of wetlands such as Sparganium stoloniferum and Bolboschoenus maritimus can be identified. But in many cases, these methods are still insufficient because of the noise brought from the atmosphere transportation and so on. The physical features of wetlands reflecting the diversification of spectrum information of wetlands, which include the spatial temporal characteristics of the wetlands distribution, the landscape differences of wetlands from season to season, the growing environment and the vertical structure of wetlands vegetation and so on, must be taken into consideration. Besides these, the artificial alteration to spatial structure of wetlands such as the exploitation of some types of them can be also used as important symbols of wetlands identification from remote sensing images. On the basis of the above geographics analysis, a set of wetlands classification models of remote sensing could be established, and many types of wetlands such as paddy field, reed swamp, peat mire, meadow, CAREX marsh and paludification meadow and so on, will be distinguished consequently. All the ways of geographical analysis and model establishment will be given in detail in this article.
基金supported by the National Natural Science Foundation of China(U1435220)
文摘How to recognize targets with similar appearances from remote sensing images(RSIs) effectively and efficiently has become a big challenge. Recently, convolutional neural network(CNN) is preferred in the target classification due to the powerful feature representation ability and better performance. However,the training and testing of CNN mainly rely on single machine.Single machine has its natural limitation and bottleneck in processing RSIs due to limited hardware resources and huge time consuming. Besides, overfitting is a challenge for the CNN model due to the unbalance between RSIs data and the model structure.When a model is complex or the training data is relatively small,overfitting occurs and leads to a poor predictive performance. To address these problems, a distributed CNN architecture for RSIs target classification is proposed, which dramatically increases the training speed of CNN and system scalability. It improves the storage ability and processing efficiency of RSIs. Furthermore,Bayesian regularization approach is utilized in order to initialize the weights of the CNN extractor, which increases the robustness and flexibility of the CNN model. It helps prevent the overfitting and avoid the local optima caused by limited RSI training images or the inappropriate CNN structure. In addition, considering the efficiency of the Na¨?ve Bayes classifier, a distributed Na¨?ve Bayes classifier is designed to reduce the training cost. Compared with other algorithms, the proposed system and method perform the best and increase the recognition accuracy. The results show that the distributed system framework and the proposed algorithms are suitable for RSIs target classification tasks.
基金Under the auspices of National Natural Science Foundation of China (No. 40871241, 40771170)National High Technology Research and Development Program of China (No. 2007AA12Z176)
文摘Wetland research has become a hot spot linking multiple disciplines presently. Wetland classification and mapping is the basis for wetland research. It is difficult to generate wetland data sets using traditional methods because of the low accessibility of wetlands, hence remote sensing data have become one of the primary data sources in wetland research. This paper presents a case study conducted at the core area of Honghe National Nature Reserve in the Sanjiang Plain, Northeast China. In this study, three images generated by airship, from Thematic Mapper and from SPOT 5 were selected to produce wetland maps at three different wetland landscape levels. After assessing classification accuracies of the three maps, we compared the different wetland mapping results of 11 plant communities to the airship image, 6 plant ecotypes to the TM image and 9 landscape classifications to the SPOT 5 image. We discussed the different characteristics of the hierarchical ecosystem classifications based on the spatial scales of the different images. The results indicate that spatial scales of remote sensing data have an important link to the hierarchies of wetland plant ecosystems displayed on the wetland landscape maps. The richness of wetland landscape information derived from an image closely relates to its spatial resolution. This study can enrich the ecological classification methods and mapping techniques dealing with the spatial scales of different remote sensing images. With a better understanding of classification accuracies in mapping wetlands by using different scales of remote sensing data, we can make an appropriate approach for dealing with the scale issue of remote sensing images.
基金Foundation： National Natural Science Foundation of China, No.51139002 No.51069005+1 种基金 Inner Mongolia Agricultural University Innovation Team Building Program Cold-Arid Region Water Resources Utilization, No.NDTD2010-6 Inner Mongolia Scientific and Technology Bureau, No.20090516, Project of the Ministry of Science and Technology of China, No.2010DFA71460
文摘Landform classification is commonly done using topographic altitude only. However practice indicates that locations at a same altitude may have distinctly different landforms, depending on characteristics of soils underneath those locations. The objectives of this study were to： 1） develop a landform classification approach that is based on both altitude and soil characteristic; and 2） use this approach to determine landforms within a watershed located in northern Ordos Plateau of China. Using data collected at 134 out of 200 sampling sites, this study determined that D10 （the diameter of soil particles 10% finer by weight） and long-term average soil moisture acquired in 2010, which can be estimated at reasonable accuracy from remote sensing imagery, can be used to represent soil characteristics of the study watershed. Also, the sampling data revealed that this watershed consists of nine classes of landforms, namely mobile dune （MD）, mobile semi-mobile dune （SMD）, rolling fixed semi-fixed dune （RFD）, flat sandy land （FD）, grassy sandy land （GS）, bedrock （BR）, flat sandy bedrock （FSB）, valley agricultural land （VA）, and swamp and salt lake （SW）. A set of logistic regression equations were derived using data collected at the 134 sampling sites and verified using data at the remaining 66 sites. The verification indicated that these equations have moderate classification accuracy （Kappa coefficients K 〉 43%）. The results revealed that the dominant classes in the study watershed are FD （36.3%）, BR （27.0%）, and MD （23.5%）, while the other six types of landforms （i.e., SMD, RFD, GS, FSB, VA, and SW） in combination account for 13.2%. Further, the landforms determined in this study were compared with the classes presented by a geologically-based classification map. The comparison indicated that the geologically-based classification could not identify multiple landforms within a class that are dependent upon soil characteristics.
基金the Scientific Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry
文摘Classification is always the key point in the field of remote sensing. Fuzzy c-Means is a traditional clustering algorithm that has been widely used in fuzzy clustering. However, this algorithm usually has some weaknesses, such as the problems of falling into a local minimum, and it needs much time to accomplish the classification for a large number of data. In order to overcome these shortcomings and increase the classifi-cation accuracy, Gustafson-Kessel (GK) and Gath-Geva (GG) algorithms are proposed to improve the tradi-tional FCM algorithm which adopts Euclidean distance norm in this paper. The experimental result shows that these two methods are able to detect clusters of varying shapes, sizes and densities which FCM cannot do. Moreover, they can improve the classification accuracy of remote sensing images.
文摘Spatial and temporal informationon urban infrastructure is essential and requires various land-cover/land-use planning and management applications.Besides,a change in infrastructure has a direct impact on other land-cover and climatic conditions.This study assessed changes in the rate and spatial distribution of Peshawar district’s infrastructure and its effects on Land Surface Temperature(LST)during the years 1996 and 2019.For this purpose,firstly,satellite images of bands7 and 8 ETM+(Enhanced Thematic Mapper)plus and OLI(Operational Land Imager)of 30 m resolution were taken.Secondly,for classification and image processing,remote sensing(RS)applications ENVI(Environment for Visualising Images)and GIS(Geographic Information System)were used.Thirdly,for better visualization and more in-depth analysis of land sat images,pre-processing techniques were employed.For Land use and Land cover(LU/LC)four types of land cover areas were identified-vegetation area,water cover,urbanized area,and infertile land for the years under research.The composition of red,green,and near infra-red bands was used for supervised classification.Classified images were extracted for analyzing the relative infrastructure change.A comparative analysis for the classification of images is performed for SVM(Support Vector Machine)and ANN(Artificial Neural Network).Based on analyzing these images,the result shows the rise in the average temperature from 30.04℃ to 45.25℃.This only possible reason is the increase in the built-up area from 78.73 to 332.78 Area km^(2) from 1996 to 2019.It has also been witnessed that the city’s sides are hotter than the city’s center due to the barren land on the borders.
基金Projectsupported by the fund of the Scientific and Technological Development of Surveying and Mapping from the State Bureau of Surveying and Mapping(No.2001_02_04).
文摘This paper presents a new kind of back propagation neural network(BPNN)based on rough sets,called rough back propagation neural network(RBPNN).The architecture and training method of RBPNN are presented and the survey and analysis of RBPNN for the classification of remote sensing multi_spectral image is discussed.The successful application of RBPNN to a land cover classification illustrates the simple computation and high accuracy of the new neural network and the flexibility and practicality of this new approach.
基金supported in part by the National Natural Science Foundation of China(No.12302252)。
文摘Deep learning significantly improves the accuracy of remote sensing image scene classification,benefiting from the large-scale datasets.However,annotating the remote sensing images is time-consuming and even tough for experts.Deep neural networks trained using a few labeled samples usually generalize less to new unseen images.In this paper,we propose a semi-supervised approach for remote sensing image scene classification based on the prototype-based consistency,by exploring massive unlabeled images.To this end,we,first,propose a feature enhancement module to extract discriminative features.This is achieved by focusing the model on the foreground areas.Then,the prototype-based classifier is introduced to the framework,which is used to acquire consistent feature representations.We conduct a series of experiments on NWPU-RESISC45 and Aerial Image Dataset(AID).Our method improves the State-Of-The-Art(SOTA)method on NWPU-RESISC45 from 92.03%to 93.08%and on AID from 94.25%to 95.24%in terms of accuracy.
文摘NOAA-AVHRR data have been more and more used by scientists because of its short temporal resolution,large scope, inexpensive cost and broad wave bands. On macro and middle scale of vegetation remote sensing, NOAAAVHRR possesses an advantage when compared with other satellites. However, because NOAA-AVHRR also problem of low resolution, data distortion and geometrical distortion, in the area of application of NOAA-AVHRR in largescale vegetation - mapping, the accuracy of vegetation classification should be improved. This paper discuss the feasibilityof integrating the geographic data in GIS(Geographical Information System) and remotely sensed data in GIS. Under theenvironment of GIS, temperature, precipitation and elevation, which serve as main factors affecting vegetation growth,were processed by a mathematical model and qualified into geographic image under a certain grid system. The geographicimage were overlaid to the NOAA-AVHRR data which had been compressed and processed. In order to evaluate the usefulness of geographic data for vegetation classification, the area under study was digitally classified by two groups of interpreter: the proposed methodology using maximum likelihood classification assisted by the geographic database and a conventional maximum likelihood classification only. Both result were compared using Kappa statistics. The indices to both theproposed and the conventional digital classification methodology were 0. 668(yew good) and 0. 563(good), respetively.The geographic database rendered an improvement over the conventional digital classification. Furthermore, in this study,some problems related to multi-sources data integration are also discussed.
基金supported by the National Natural Science Foundation of China (62201438,61772397,12005169)the Basic Research Program of Natural Sciences of Shaanxi Province (2021JC-23)+2 种基金Yulin Science and Technology Bureau Science and Technology Development Special Project (CXY-2020-094)Shaanxi Forestry Science and Technology Innovation Key Project (SXLK2022-02-8)the Project of Shaanxi F ederation of Social Sciences (2022HZ1759)。
文摘The development of image classification is one of the most important research topics in remote sensing. The prediction accuracy depends not only on the appropriate choice of the machine learning method but also on the quality of the training datasets. However, real-world data is not perfect and often suffers from noise. This paper gives an overview of noise filtering methods. Firstly, the types of noise and the consequences of class noise on machine learning are presented. Secondly, class noise handling methods at both the data level and the algorithm level are introduced. Then ensemble-based class noise handling methods including class noise removal, correction, and noise robust ensemble learners are presented. Finally, a summary of existing data-cleaning techniques is given.
基金Project(40174003) supported by the National Natural Science Foundation of China
文摘The classification of hyperspectral remote sensing data is an important problem theoretically and practically. With the increase of spectral bands, the separability of objects on remote sensing image should be improved. But the effects of traditional algorithm on feature extraction such as principal component analysis(PCA) is not so good for hyperspectral image. The key problem is that PCA can only represent the linear structure of data set; while the data clouds of different objects on hyperspectral image usually distribute on a nonlinear manifold. This paper established an algorithm of nonlinear feature extraction named as nonlinear principal poly lines, based on the algorithm, a classifier is constructed and the classification accuracy of hyperspectral image can be improved.
文摘ANNs （Artificial neural networks） are used extensively in remote sensing image processing. It has been proven that BPNNs （back-propagation neural networks） have high attainable classification accuracy. However, there is a noticeable variation in the achieved accuracies due to different network designs and implementations. Hence, researchers usually need to conduct several experimental trials before they can finalize the network design. This is a time consuming process which significantly reduces the effectiveness of using BPNNs and the final design may still not be optimal. Therefore, there is a need to see whether there are some common guidelines for effective design and implementation of BPNNs. With this aim in mind, this paper attempts to find and summarize the common guidelines suggested by different authors through literature review and discussion of the findings. To provide readers with background and contextual information, some ANN fundamentals are also introduced.
基金Hunan University of Arts and Science provided doctoral research funding for this study (grant number 16BSQD23)Fund of Geography Subject ([2022]351)also provided funding.
文摘Recently,the convolutional neural network(CNN)has been dom-inant in studies on interpreting remote sensing images(RSI).However,it appears that training optimization strategies have received less attention in relevant research.To evaluate this problem,the author proposes a novel algo-rithm named the Fast Training CNN(FST-CNN).To verify the algorithm’s effectiveness,twenty methods,including six classic models and thirty archi-tectures from previous studies,are included in a performance comparison.The overall accuracy(OA)trained by the FST-CNN algorithm on the same model architecture and dataset is treated as an evaluation baseline.Results show that there is a maximal OA gap of 8.35%between the FST-CNN and those methods in the literature,which means a 10%margin in performance.Meanwhile,all those complex roadmaps,e.g.,deep feature fusion,model combination,model ensembles,and human feature engineering,are not as effective as expected.It reveals that there was systemic suboptimal perfor-mance in the previous studies.Most of the CNN-based methods proposed in the previous studies show a consistent mistake,which has made the model’s accuracy lower than its potential value.The most important reasons seem to be the inappropriate training strategy and the shift in data distribution introduced by data augmentation(DA).As a result,most of the performance evaluation was conducted based on an inaccurate,suboptimal,and unfair result.It has made most of the previous research findings questionable to some extent.However,all these confusing results also exactly demonstrate the effectiveness of FST-CNN.This novel algorithm is model-agnostic and can be employed on any image classification model to potentially boost performance.In addition,the results also show that a standardized training strategy is indeed very meaningful for the research tasks of the RSI-SC.
基金Foundation item: Under the auspices of Scientific Foundation Research Project of the Ministry of Science and Technology and Chinese Academy of Surveying and Mapping (No. F0610)
文摘Remote sensing technique, replacing conventional sonar bathymetry technique, has become an effective complementary method of mapping submarine terrain where special conditions make the sonar technique difficult to be carried out. At the same time, as one kind of data set, multispectral remote sensing data has the disadvantage of being influenced by the variable bottom types in shallow seawater, when it is applied in bathymetry. This paper puts forward a new method to extract water depth information from multispectral data, considering the bottom classification and the true water depth accuracy. That is the Principal Component Analysis (PCA) technique based on the bottom classification. By the least square regression with significance, the experiment near Qingdao City has obtained more satisfactory bathymetry accuracy than that of the traditional single-band method, with the mean absolute error about 2.57m.
