In Hyperspectral Imaging(HSI),the detrimental influence of noise and distortions on data quality is profound,which has severely affected the following-on analytics and decisionmaking such as land mapping.This study pr...In Hyperspectral Imaging(HSI),the detrimental influence of noise and distortions on data quality is profound,which has severely affected the following-on analytics and decisionmaking such as land mapping.This study presents an innovative framework for assessing HSI band quality and reconstructing the low-quality bands,based on the Prophet model.By introducing a comprehensive quality metric to start,the authors approach factors in both spatial and spectral characteristics across local and global scales.This metric effectively captures the intricate noise and distortions inherent in the HSI data.Subsequently,the authors employ the Prophet model to forecast information within the low-quality bands,leveraging insights from neighbouring high-quality bands.To validate the effectiveness of the authors’proposed model,extensive experiments on three publicly available uncorrected datasets are conducted.In a head-to-head comparison,the framework against six state-ofthe-art band reconstruction algorithms including three spectral methods,two spatialspectral methods and one deep learning method is benchmarked.The authors’experiments also delve into strategies for band selection based on quality metrics and the quality evaluation of the reconstructed bands.In addition,the authors assess the classification accuracy utilising these reconstructed bands.In various experiments,the results consistently affirm the efficacy of the authors’method in HSI quality assessment and band reconstruction.Notably,the authors’approach obviates the need for manually prefiltering of noisy bands.This comprehensive framework holds promise in addressing HSI data quality concerns whilst enhancing the overall utility of HSI.展开更多
Recent advances in convolution neural network (CNN) have fostered the progress in object recognition and semantic segmentation, which in turn has improved the performance of hyperspectral image (HSI) classification. N...Recent advances in convolution neural network (CNN) have fostered the progress in object recognition and semantic segmentation, which in turn has improved the performance of hyperspectral image (HSI) classification. Nevertheless, the difficulty of high dimensional feature extraction and the shortage of small training samples seriously hinder the future development of HSI classification. In this paper, we propose a novel algorithm for HSI classification based on three-dimensional (3D) CNN and a feature pyramid network (FPN), called 3D-FPN. The framework contains a principle component analysis, a feature extraction structure and a logistic regression. Specifically, the FPN built with 3D convolutions not only retains the advantages of 3D convolution to fully extract the spectral-spatial feature maps, but also concentrates on more detailed information and performs multi-scale feature fusion. This method avoids the excessive complexity of the model and is suitable for small sample hyperspectral classification with varying categories and spatial resolutions. In order to test the performance of our proposed 3D-FPN method, rigorous experimental analysis was performed on three public hyperspectral data sets and hyperspectral data of GF-5 satellite. Quantitative and qualitative results indicated that our proposed method attained the best performance among other current state-of-the-art end-to-end deep learning-based methods.展开更多
Accurate mapping of wetlands is crucial for wetlands conservation, as well as for monitoring and assessing coastal resources and the environment. Multispectral(MSI) satellite image time series have rich temporal evolu...Accurate mapping of wetlands is crucial for wetlands conservation, as well as for monitoring and assessing coastal resources and the environment. Multispectral(MSI) satellite image time series have rich temporal evolution characteristics, which can reveal dynamic changes in surface cover and environmental conditions. However, due to the limited number of bands, the ability to express the difference of ground features is limited, resulting in an inability to capture surface objects' changes in the finer spectral range. Therefore,this paper proposed a dual-branch spatial-temporal spectral feature fusion network(Fusion-Former), which combined MSI time series data with hyperspectral(HSI) data to achieve accurate mapping of wetlands in Liaohe River Delta, China in 2022. Fusion-Former achieved an overall accuracy(OA) of 96.36% in the Liaohe River Delta wetland, significantly outperforming all benchmark methods.Experimental results demonstrate that utilizing the temporal phenological information from multi-temporal MSI and the fine-grained spatial-spectral features from HSI can effectively resolve the misclassification between spectrally similar vegetation and water bodies.Furthermore, a continuous improvement in accuracy was observed as the length of the input time series increased, underscoring the critical role of temporal information. Therefore, by integrating these complementary information sources, the proposed method enables the generation of accurate wetland maps to support decision-makers in formulating more precise conservation and management strategies.展开更多
In recent years,nonvascular epiphytic communities have been increasingly subjected to extreme climatic conditions,with heavy rains and prolonged droughts.Therefore,understanding their management of water resources pro...In recent years,nonvascular epiphytic communities have been increasingly subjected to extreme climatic conditions,with heavy rains and prolonged droughts.Therefore,understanding their management of water resources provides insight into their ecosystem-level contributions.However,until now,little has been done to assess this feature at a micro-scale level considering species-species interactions.In this context,this study develops an analytical strategy based on hyperspectral imaging(HSI)and chemometrics to map the water content(WC)of nonvascular epiphytic communities during a dehydration process,while considering interactions among life forms.Exploratory analysis of data by means of principal component analysis(PCA)demonstrates that the highest source of variability along the process is due to water loss,though differences among communities can be observed as well.Indeed,the generation of false color RGB score maps enables the evaluation of different life forms'responses,giving an initial understanding of facilitation and competition mechanisms based on community composition.Moreover,the use of multivariate regression using partial least squares(PLS)regression to predict water content at a pixel level,with a final error in prediction around 3%,leads to the visualization of maps representing the WC of each pixel composing the sample,permitting the evaluation of communities'response at a detailed scale,providing a valuable method for recovering spatial information while monitoring dehydration.The analytical impact and novelty of the approach are supported by the consistency in results obtained from developing the model with two different strategies,image-based and pixel-based,and by the complementarity of the information obtained by the two strategies themselves.展开更多
基金National Natural Science Foundation Major Project of China,Grant/Award Number:42192580Guangdong Province Key Construction Discipline Scientific Research Ability Promotion Project,Grant/Award Number:2022ZDJS015。
文摘In Hyperspectral Imaging(HSI),the detrimental influence of noise and distortions on data quality is profound,which has severely affected the following-on analytics and decisionmaking such as land mapping.This study presents an innovative framework for assessing HSI band quality and reconstructing the low-quality bands,based on the Prophet model.By introducing a comprehensive quality metric to start,the authors approach factors in both spatial and spectral characteristics across local and global scales.This metric effectively captures the intricate noise and distortions inherent in the HSI data.Subsequently,the authors employ the Prophet model to forecast information within the low-quality bands,leveraging insights from neighbouring high-quality bands.To validate the effectiveness of the authors’proposed model,extensive experiments on three publicly available uncorrected datasets are conducted.In a head-to-head comparison,the framework against six state-ofthe-art band reconstruction algorithms including three spectral methods,two spatialspectral methods and one deep learning method is benchmarked.The authors’experiments also delve into strategies for band selection based on quality metrics and the quality evaluation of the reconstructed bands.In addition,the authors assess the classification accuracy utilising these reconstructed bands.In various experiments,the results consistently affirm the efficacy of the authors’method in HSI quality assessment and band reconstruction.Notably,the authors’approach obviates the need for manually prefiltering of noisy bands.This comprehensive framework holds promise in addressing HSI data quality concerns whilst enhancing the overall utility of HSI.
基金the National Natural Science Foundation of China(No.51975374)。
文摘Recent advances in convolution neural network (CNN) have fostered the progress in object recognition and semantic segmentation, which in turn has improved the performance of hyperspectral image (HSI) classification. Nevertheless, the difficulty of high dimensional feature extraction and the shortage of small training samples seriously hinder the future development of HSI classification. In this paper, we propose a novel algorithm for HSI classification based on three-dimensional (3D) CNN and a feature pyramid network (FPN), called 3D-FPN. The framework contains a principle component analysis, a feature extraction structure and a logistic regression. Specifically, the FPN built with 3D convolutions not only retains the advantages of 3D convolution to fully extract the spectral-spatial feature maps, but also concentrates on more detailed information and performs multi-scale feature fusion. This method avoids the excessive complexity of the model and is suitable for small sample hyperspectral classification with varying categories and spatial resolutions. In order to test the performance of our proposed 3D-FPN method, rigorous experimental analysis was performed on three public hyperspectral data sets and hyperspectral data of GF-5 satellite. Quantitative and qualitative results indicated that our proposed method attained the best performance among other current state-of-the-art end-to-end deep learning-based methods.
基金Under the auspices of the National Natural Science Foundation of China(No.42471380,42325104,42101350)。
文摘Accurate mapping of wetlands is crucial for wetlands conservation, as well as for monitoring and assessing coastal resources and the environment. Multispectral(MSI) satellite image time series have rich temporal evolution characteristics, which can reveal dynamic changes in surface cover and environmental conditions. However, due to the limited number of bands, the ability to express the difference of ground features is limited, resulting in an inability to capture surface objects' changes in the finer spectral range. Therefore,this paper proposed a dual-branch spatial-temporal spectral feature fusion network(Fusion-Former), which combined MSI time series data with hyperspectral(HSI) data to achieve accurate mapping of wetlands in Liaohe River Delta, China in 2022. Fusion-Former achieved an overall accuracy(OA) of 96.36% in the Liaohe River Delta wetland, significantly outperforming all benchmark methods.Experimental results demonstrate that utilizing the temporal phenological information from multi-temporal MSI and the fine-grained spatial-spectral features from HSI can effectively resolve the misclassification between spectrally similar vegetation and water bodies.Furthermore, a continuous improvement in accuracy was observed as the length of the input time series increased, underscoring the critical role of temporal information. Therefore, by integrating these complementary information sources, the proposed method enables the generation of accurate wetland maps to support decision-makers in formulating more precise conservation and management strategies.
基金supported by the University of Genoa as part of the Science and Technologies for the Earth and Environment(STAT)doctoral programFinancial support provided by the Italian Ministry of Universities and Research–MUR(Research Project PRIN 2022 n.20223WBTH8,CUP:D53D23008950006)。
文摘In recent years,nonvascular epiphytic communities have been increasingly subjected to extreme climatic conditions,with heavy rains and prolonged droughts.Therefore,understanding their management of water resources provides insight into their ecosystem-level contributions.However,until now,little has been done to assess this feature at a micro-scale level considering species-species interactions.In this context,this study develops an analytical strategy based on hyperspectral imaging(HSI)and chemometrics to map the water content(WC)of nonvascular epiphytic communities during a dehydration process,while considering interactions among life forms.Exploratory analysis of data by means of principal component analysis(PCA)demonstrates that the highest source of variability along the process is due to water loss,though differences among communities can be observed as well.Indeed,the generation of false color RGB score maps enables the evaluation of different life forms'responses,giving an initial understanding of facilitation and competition mechanisms based on community composition.Moreover,the use of multivariate regression using partial least squares(PLS)regression to predict water content at a pixel level,with a final error in prediction around 3%,leads to the visualization of maps representing the WC of each pixel composing the sample,permitting the evaluation of communities'response at a detailed scale,providing a valuable method for recovering spatial information while monitoring dehydration.The analytical impact and novelty of the approach are supported by the consistency in results obtained from developing the model with two different strategies,image-based and pixel-based,and by the complementarity of the information obtained by the two strategies themselves.
