Loading...

Table of Content

    20 July 2023, Volume 52 Issue 7
    Previous Issue    Next Issue
    Special Issue of Hyperspectral Remote Sensing Technology
    Development and key technologies of spaceborne hyperspectral imaging payload
    LIU Yinnian, XUE Yongqi
    2023, 52(7):  1045-1058.  doi:10.11947/j.AGCS.2023.20220498
    Asbtract ( )   HTML ( )   PDF (11305KB) ( )  
    References | Related Articles | Metrics
    As spaceborne hyperspectral remote sensing technology is able to rapidly identify ground objects over a wide range based on spectrum characteristics, it has the potential to be widely used in natural resource exploration, ecological environment protection, precision agriculture, carbon emission monitoring, real-time detection of the Earth's surface anomalies. There has been an increasing emphasis on the research and development of hyperspectral imaging technology since NASA produced the first airborne hyperspectral imager in the early 1980s. Although the development of spaceborne hyperspectral payloads is more complex than that of airborne hyperspectral payloads, it has become the commanding height of science and technology in the international competition and also an important means for humans to detect the planet and perceive everything due to its enormous application value in rapid detection and identification on a global scale. The successful launch of our country's GF-5 satellite has elevated the international standard of spaceborne hyperspectral imaging technology to a new height, which contributes to many breakthroughs made in the application of carbon emission, soil organic matter, soil heavy metal pollution, trace pollution of water quality and large-scale deep earth prospecting. This article reviews the evolution of spaceborne hyperspectral payload technology, as well as summarizes the main points, key technologies, and applications of spaceborne hyperspectral payload with wide-spectrum and large-range. In addition, based on our team's actual research effort in this field over many years, several important development directions and key technologies have been proposed, such as geostationary orbit hyperspectral, fluorescence ultra-spectral, real-time hyperspectral detection, and so on, which will provide some essential and useful references for the key advancements of spaceborne hyperspectral payload research.
    Hyperspectral remote sensing image intrinsic information decomposition: advances and challenges
    LI Shutao, WU Qiong, KANG Xudong
    2023, 52(7):  1059-1073.  doi:10.11947/j.AGCS.2023.20220563
    Asbtract ( )   HTML ( )   PDF (3268KB) ( )  
    References | Related Articles | Metrics
    Hyperspectral imaging is a powerful image acquisition method which can record the rich spectral and spatial information of the scene in a high dimensional data cube. Due to this advantage, hyperspectral imaging has been very useful in many practical applications of earth observation and aerospace. However, as a branch of optical remote sensing, the performance of hyperspectral imaging may be affected by many factors such as atmosphere and illumination. The objective of hyperspectral intrinsic image decomposition is to decrease the influence of complex environmental factors, extract and represent the intrinsic spectral and spatial information of hyperspectral images accurately, so as to improve the performance of hyperspectral image recognition and interpretation. This paper reviews some representative work in hyperspectral intrinsic image decomposition. The principle, advantages, and disadvantages of some typical intrinsic image decomposition methods have been analyzed. Moreover, the challenging problems of intrinsic image decomposition faced in real remote sensing applications have been illustrated. At last, based on the requirements of practical remote sensing applications, we discuss the development trends of hyperspectral intrinsic image decomposition. This review could be a good guide for those researchers who are interested in the advances and applications of hyperspectral remote sensing. More importantly, it gives some important future research directions that could be investigated in the future.
    Advances and prospects in hyperspectral and multispectral remote sensing image super-resolution fusion
    ZHANG Bing, GAO Lianru, LI Jiaxin, HONG Danfeng, ZHENG Ke
    2023, 52(7):  1074-1089.  doi:10.11947/j.AGCS.2023.20220499
    Asbtract ( )   HTML ( )   PDF (9353KB) ( )  
    References | Related Articles | Metrics
    As an important component of multimodal remote sensing data, hyperspectral images are capable of capturing the fine spectral profiles of objects. However, due to the limitations of the imaging mechanism, the loss of spatial details leads to degradation of the spatial representation capability of hyperspectral images, which to a certain extent limits the potential for further applications. Data fusion is an effective approach to resolve the contradiction in spatial and spectral domains, where related theories have intensively developed in recent years. This paper provides a comprehensive overview of the advances and prospects in hyperspectral and multispectral remote sensing image super-resolution fusion. Firstly, fusion algorithms are systematically introduced and classified into three categories, namely, detail injection-based, model optimization-based, and deep learning-based methods. The principles, models and representative algorithms of different methods are reviewed, with emphasis on matrix decomposition, tensor representation in model optimization-based methods, and supervised and unsupervised methods in deep learning-based methods. On this basis, successful applications of the techniques in the field of pixel-level classification,target extraction,and on-board fusion are given, finding that the potential of fusion products has not been fully exploited in subsequent applications. Then, prospective directions are discussed from four aspects, including degradation models, data-model-driven fusion, multi-task integration fusion and application-coupled fusion. Finally, the current process and prospects of the future development trend in this field are summarized, pointing out the strengths and weaknesses of various approaches while highlighting the importance of multi-approach association, external data assistance, application-driven fusion, et al.
    A capsule network for hyperspectral image classification employing spatial-spectral feature
    DU Peijun, ZHANG Wei, ZHANG Peng, LIN Cong, GUO Shanchuan, HU Zezhou
    2023, 52(7):  1090-1104.  doi:10.11947/j.AGCS.2023.20220565
    Asbtract ( )   HTML ( )   PDF (21778KB) ( )  
    References | Related Articles | Metrics
    An efficient and stable deep learning classifier can improve the classification accuracy of hyperspectral remote sensing images. In order to deal with the insufficiency of the scalar neuron's limited feature expression ability and the inability to effectively model the spatial hierarchical relationship among features in convolutional neural networks, an end-to-end hyperspectral capsule network (H-CapsNet) was designed considering the characteristics of hyperspectral image. The main body of H-CapsNet is composed of encoder (Conv, PrimaryCaps and DigitCats) and decoder (fully connection layer). It mainly embeds channel and spatial attention modules at the network input to enhance the model's capture and recognition of spatial and spectral features, thereby improving the network's ability to focus and express features. Taking the hyperspectral images of Zhang-jia-gang city and two public datasets:University of Pavia and University of Houston, as examples, the performance of the proposed H-CapsNet was compared with traditional machine learning algorithms and several deep neural networks. The experimental results show that the H-CapsNet has achieved the best classification accuracy on three hyperspectral images with different resolutions, and the overall accuracy is improved by 2.36%~7.67%, 0.16%~11.8% and 1.75%~15.58% compared with other methods. In particular, the H-CapsNet has good adaptability to small pixel neighborhoods. When the image patch size is limited, it can still achieve relatively ideal classification results.
    Adversarial autoencoder for hyperspectral anomaly detection
    DU Qian, XIE Weiying
    2023, 52(7):  1105-1114.  doi:10.11947/j.AGCS.2023.20220635
    Asbtract ( )   HTML ( )   PDF (9047KB) ( )  
    References | Related Articles | Metrics
    Autoencoder (AE) is a typical generative model. It has been widely used due to its simple learning process, good ability for convergence, and unsupervised nature. To improve the performance of AE whose objective function is merely input-output reconstruction error, adversarial autoencoder (AAE) has been proposed, which can provide variational inference to the network output. This paper reviews the use of unsupervised and semisupervised AAE in hyperspectral anomaly detection (HAD). The performance of AAE can be improved by adding adversarial learning between the input of the encoder and the output of the decoder, in addition to the adversarial learning in the latent space in the original AAE. In this way, the network can focus more on learning data distribution rather than point-to-point data reconstruction. The idea of using these deep learning models is beyond the concept of traditional HAD methods, and can significantly improve the detection performance, as demonstrated by real data experiments.
    Potential analysis and prospect of hyperspectral ground object recognition
    YU Xuchu, LIU Bing, XUE Zhixiang
    2023, 52(7):  1115-1125.  doi:10.11947/j.AGCS.2023.20220495
    Asbtract ( )   HTML ( )   PDF (1559KB) ( )  
    References | Related Articles | Metrics
    Hyperspectral remote sensing technology has fully released its potential in geospatial information acquisition due to the unique technical advantages in massive data analysis, information fusion, and recognition of ground objects. Due to the nonlinearity of hyperspectral data and complexity of ground objects, there are problems of data heterogeneity and sample scarcity when using hyperspectral images for land cover classification. Classical hyperspectral image analysis methods have the characteristics of a small amount of calculation, fewer labeled samples requirement, and strong theoretical interpretability, which play an important role in the extraction of ground objects attribute information. To cope with the rapid increase in the amount of data and more diverse applications, it is urgent to develop automatic and intelligent hyperspectral recognition technology for land cover classification. In recent years, artificial intelligence methods have been studied and widely applied in the field of hyperspectral remote sensing. In particular, image analysis and information extraction technology based on deep learning has become a persistent hot spot, which has effectively promoted the refinement and intelligence of hyperspectral recognition of ground objects. Based on a brief analysis of the potential and demand of recognition of ground objects, this article systematically summarizes several advancements in hyperspectral image analysis, and focuses on new ideas of deep learning in recent years for the intelligent recognition of hyperspectral ground objects. Firstly, combining the terrain elements and the capability of hyperspectral detection, the hyperspectral ground objects are divided into four categories (i.e., vegetation, soil, water, and artificial buildings) and several sub-categories, and spectral response characteristics of four ground objects are also analyzed; thereafter, in terms of image analysis research, the image analysis technologies especially band selection, feature extraction, pattern classification, and post-classification processing are reviewed, and new research directions and hotspots are also analyzed. In terms of intelligent processing, according to the schedule of supervised learning, semi-supervised learning, and self-supervised learning, the deep neural network models applied in hyperspectral ground object recognition are systematically summarized, and the applications of transfer learning as well as meta learning are also analyzed. Finally, based on the above analysis, the development trends of hyperspectral ground object recognition are also prospected, so as to expand the research ideas in the future.
    Progress of hyperspectral remote sensing applications on cultural relics protection
    ZHANG Lifu, WANG Sa, ZHANG Yan, YUAN Deshuai, SONG Ruoxi, QI Wenchao, QU Liang, LU Zhiyong, TONG Qingxi
    2023, 52(7):  1126-1138.  doi:10.11947/j.AGCS.2023.20220655
    Asbtract ( )   HTML ( )   PDF (13443KB) ( )  
    References | Related Articles | Metrics
    Cultural relics are the crystallization of historical, artistic, and scientific values in the development of human civilization, and it contains historical and cultural information. Cultural relics digitization is an important link for their protection, restoration, and reconstruction. As an important supplement of traditional cultural relics digitization ways, hyperspectral remote sensing can quickly and non-destructively obtain the material, pigment, and processing traces of cultural relics, and provide physical and chemical properties of cultural relics surface for the digital resources of cultural relics. In this paper, the equipment and methods of hyperspectral information collection in the field are summarized and analyzed, and then the key technologies of hyperspectral information processing in the field of cultural relics protection are reviewed, and summarized the four aspects of information increase, information extraction, information classification and visualization, and compares and analyzes the common methods. Based on the comparison and analysis of the common methods, the typical application of hyperspectral technology in the field of cultural preservation is summarized, including the analysis and recognition of pigment species, the extraction of hidden information and the recognition of repaired traces. Finally, the future research directions of hyperspectral cultural relics applications are prospected.
    Hyperspectral image classification method based on hierarchical transformer network
    ZHANG Yichao, ZHENG Xiangtao, LU Xiaoqiang
    2023, 52(7):  1139-1147.  doi:10.11947/j.AGCS.2023.20220540
    Asbtract ( )   HTML ( )   PDF (2589KB) ( )  
    References | Related Articles | Metrics
    Hyperspectral image classification, which assigns each pixel to predefined land cover categories, is of crucial importance in various Earth science tasks such as environmental mapping and other related fields. In recent years, scholars have attempted to utilize deep learning frameworks for hyperspectral image classification and achieved satisfactory results. However, these methods still have certain deficiencies in extracting spectral features. This paper proposes a hierarchical self-attention network (HSAN) for hyperspectral image classification based on the self-attention mechanism. Firstly, a skip-layer self-attention module is constructed for feature learning, leveraging the self-attention mechanism of Transformer to capture contextual information and enhance the contribution of relevant information. Secondly, a hierarchical fusion method is designed to further alleviate the loss of relevant information during the feature learning process and enhance the interplay of features at different hierarchical levels. Experimental results on the Pavia University and Houston2013 datasets demonstrate that the proposed framework outperforms other state-of-the-art hyperspectral image classification frameworks.
    Singular spectrum analysis method for hyperspectral imagery feature extraction: a review and evaluation
    SUN Genyun, FU Hang, ZHANG Aizhu, REN Jinchang
    2023, 52(7):  1148-1163.  doi:10.11947/j.AGCS.2023.20220542
    Asbtract ( )   HTML ( )   PDF (18679KB) ( )  
    References | Related Articles | Metrics
    Hyperspectral remote sensing imagery (HSI) usually contains dozens to hundreds of continuous spectral bands, with the syncretism of spectrum and image, spectral continuity, which can realize fine classification of ground objects and has been widely used in agriculture, forestry, urban and marine areas. The feature extraction of HSI is the premise of hyperspectral applications and has become one of the research hotspots and frontier topics in remote sensing. In recent years, singular spectrum analysis (SSA) has been applied in HSI, achieving superior results in the extraction of spectral and spatial features, and gradually becoming an effective feature extraction method. In this paper, firstly, the research progress and existing problems of HSI feature extraction are analyzed. Secondly, the existing SSA methods are systematically summarized and reviewed. The functions, effects, advantages, and disadvantages of three types of methods, namely, spectral domain 1D-SSA, spatial domain 2D-SSA, and combined spectral-spatial domain SSA, are introduced respectively, and the classification results are verified on two publicly available HSI datasets and one China Gaofen-5 satellite HSI dataset. Finally, the SSA feature extraction is summarized and future research directions are discussed.
    Inter-spectral contrast learning based unsupervised feature extraction for hyperspectral images
    HANG Renlong, LI Chengxiang, LIU Qingshan
    2023, 52(7):  1164-1174.  doi:10.11947/j.AGCS.2023.20220493
    Asbtract ( )   HTML ( )   PDF (5784KB) ( )  
    References | Related Articles | Metrics
    Deep learning is able to extract high-level features from input data via layer by layer abstraction. In recent years, it has been widely used in hyperspectral image classification. Most of the existing deep learning-based feature extraction methods for hyperspectral images belong to supervised learning models, which require a large number of labeled samples in the training process, but it is difficult and time-consuming to label hyperspectral images pixel by pixel. Therefore, we propose an unsupervised deep learning model based on inter-spectral contrast learning in this paper. It can extract features by modeling the relationship between different spectral bands without annotation of samples. Specifically, because different spectral channels of hyperspectral image depict the response degree of the same object in different electromagnetic spectrum, there must be a feature space, which makes the spectral information of different channels have similar characterization. Inspired by this, we first divide the high-dimensional spectral information into two groups, and then extract the features of each group using multi-layer convolution operations. Finally, the features extracted from different samples are compared and a contrastive loss function is constructed to optimize the model parameters. To test the performance of the proposed model, it was applied to a hyperspectral image classification task and validated on three commonly used data sets, including Houston 2013, Pavia University and WHU-Hi-Longkou. Experimental results show that using only 10 training samples in each class, the proposed unsupervised learning model can obtain better classification performance than the commonly used unsupervised models such as principal component analysis and auto-encoder.
    Adaptive context aggregation network for H2 remote sensing imagery classification
    HU Xin, WANG Xinyu, ZHONG Yanfei
    2023, 52(7):  1175-1186.  doi:10.11947/j.AGCS.2023.20220237
    Asbtract ( )   HTML ( )   PDF (8374KB) ( )  
    References | Related Articles | Metrics
    High spectral and spatial resolution (H2) remote sensing imagery can achieve more comprehensive and precise attribute recognition of ground objects. However, the details of ground objects are gradually revealed with the significant improvement of the spatial resolution, which makes the H2 images show extremely high spectral variability and spatial heterogeneity, and then the phenomenon of the same class with different spectrums occurs in large numbers; the intraclass variance increases significantly. As a result, an adaptive aggregation context network was proposed for H2 image classification, which uses a full convolution network with encoder-decoder architecture to achieve global spectrum-spatial fusion. A local-to-global long-distance context module was designed to alleviate the intraclass variance in the encoder module. Then an adaptive context aggregation module was constructed in the decoder module for the adaptive aggregation of local and global context information. ACANet has achieved excellent performance in the WHU-Hi benchmark dataset, and the experiments show that it can sufficiently alleviate the impact of the spatial-spectrum heterogeneity of the H2 image in the precise classification.
    Review of hyperspectral remote sensing image subpixel information extraction
    FENG Ruyi, WANG Lizhe, ZENG Tieyong
    2023, 52(7):  1187-1201.  doi:10.11947/j.AGCS.2023.20220491
    Asbtract ( )   HTML ( )   PDF (1521KB) ( )  
    References | Related Articles | Metrics
    Hyperspectral remote sensing image provides abundant data for precise landcover classification, target detection and object recognition, attributed to its unique advantages in very high spectral resolution, continuous spectum as well as the synchronous acquisition of both image and spectra of objects. Due to the limitation of the spatial resolution and the complicated scenes, mixed pixels are common in hyperspectral remote sensing images. The mixed pixel problem hinders the information extraction and analysis, and consequently, greatly weaks the potential application in various fields. It has become an important frontier scientific issue and hot spot to tackle the mixed pixel problem and realize the information extraction and analysis deep into subpixel scale for hyperspectral remote sensing imagery. This review generates a systematic summary for hyperspectral remote sensing image subpixel information extraction, and conducts a comprehensive review of the classical approaches from three aspects, namely, hyperspectral unmixing, subpixel mapping and subpixel target detection. Additionaly, this paper also analyzes and evaluates the current progress, development frontier and main challenges in related fields at home and abroad. Finally, the research trends and directions are discussed, especially in the aspects of model construction, optimization algorithm, and theoretical research and practical application combination.
    Classification of hyperspectral forest tree species based on morphological transform and spatial logical integration
    ZHANG Mengmeng, LI Wei, LIU Huan, ZHAO Xudong, TAO Ran
    2023, 52(7):  1202-1211.  doi:10.11947/j.AGCS.2023.20220492
    Asbtract ( )   HTML ( )   PDF (8542KB) ( )  
    References | Related Articles | Metrics
    By recording reflectance spectral information of the ground on an aircraft or satellite platform, hyperspectral imagery (HSI), occupying dozens of or even hundreds of contiguous narrow bands, possesses abundant discriminative information for land use. Compared with visible light images and multispectral images, HSI can reveal subtle spectral characteristics, which contribute to a more accurate identification of the materials and classes of land covers. However, most existing methods overly focus on spectral knowledge while neglecting the potential morphological and spatial information within the hyperspectral input. In the classification of complex objects, the capture of morphological differences is much more necessary for searching out the class boundaries of fine-grained classes, e.g., forestry tree species. In this paper, the importance of morphological structure utilization is analyzed, and different feature extractors are designed. Specifically, focusing on fine-grained traits extraction, we propose a coarse-to-fine spatial information integration network, called MS-NET (morphological and spatial information based network), for tree species classification. The morphological operators are effectively embedded with the trainable structuring elements, which contributes to acquiring distinctive morphology representations, enhancing the classification accuracy. We evaluate the classification performance of the proposed method on two tree species datasets, and the results demonstrate that the proposed method provides superior performance when compared with other state-of-the-art classifiers.
    Hyperspectral with high-spatial resolution remote sensing from observation, processing to applications
    ZHONG Yanfei, WANG Xinyu, HU Xin, WANG Shaoyu, WAN Yuting, TANG Ge, ZHANG Liangpei
    2023, 52(7):  1212-1226.  doi:10.11947/j.AGCS.2023.20220715
    Asbtract ( )   HTML ( )   PDF (9249KB) ( )  
    References | Related Articles | Metrics
    Hyperspectral remote sensing has always been a research hotspot in the field of remote sensing. However, limited by imaging aperture and energy, it is difficult to obtain the imagery with hyperspectral and high spatial resolution at the same time, which greatly limits the application of hyperspectral remote sensing in fine-scale tasks. In recent years, with the development of hyperspectral imaging technology and new observation platforms represented by unmanned aerial vehicles, hyperspectral and high-spatial resolution (H2, with both nanometer spectral resolution and submeter spatial resolution) has developed rapidly, promoting the application of hyperspectral remote sensing technology, but at the same time, it has also brought more problems.The extremely high spatial and spectral resolution makes the data more massive and high-dimensional, increases the spatial heterogeneity and spectral variability of hyperspectral data, and brings greater challenges to intelligent image information processing. Therefore, this article reviews the application and development status of H2 remote sensing image from three aspects:H2 remote sensing image benchmark dataset, H2 remote sensing image intelligent information processing and typical application of H2 remote sensing image.
    Summary of PhD Thesis
    Precise orbit determination of small body spacecraft and gravitational mass estimation
    JIN Weitong
    2023, 52(7):  1227-1227.  doi:10.11947/j.AGCS.2023.20210703
    Asbtract ( )   HTML ( )   PDF (716KB) ( )  
    Related Articles | Metrics
    Pattern mining and knowledge discovery of complex trajectory: a case study of mesoscale eddies in the South China Sea
    WANG Huimeng
    2023, 52(7):  1228-1228.  doi:10.11947/j.AGCS.2023.20210713
    Asbtract ( )   HTML ( )   PDF (722KB) ( )  
    Related Articles | Metrics
    Several improved methods for prediction of earth rotation parameters
    WU Fei
    2023, 52(7):  1229-1229.  doi:10.11947/j.AGCS.2023.20210716
    Asbtract ( )   HTML ( )   PDF (714KB) ( )  
    Related Articles | Metrics
    Spatio-temporal characteristics analysis of public perception and comprehensive evaluation of air quality
    SUN Yong
    2023, 52(7):  1230-1230.  doi:10.11947/j.AGCS.2023.20210719
    Asbtract ( )   HTML ( )   PDF (738KB) ( )  
    Related Articles | Metrics
    Research on the technologies of SBAS clock-ephemeris augmentation
    ZHENG Shuaiyong
    2023, 52(7):  1231-1231.  doi:10.11947/j.AGCS.2023.20210724
    Asbtract ( )   HTML ( )   PDF (735KB) ( )  
    Related Articles | Metrics
    Research on theory and methods of pulsar timing data
    ZHOU Qingyong
    2023, 52(7):  1232-1232.  doi:10.11947/j.AGCS.2023.20210725
    Asbtract ( )   HTML ( )   PDF (723KB) ( )  
    Related Articles | Metrics
    Research on building change detection from high-resolution remote sensing images in complex urban scenes
    GONG Jinqi
    2023, 52(7):  1233-1233.  doi:10.11947/j.AGCS.2023.20210728
    Asbtract ( )   HTML ( )   PDF (714KB) ( )  
    Related Articles | Metrics
    Research on the point cloud filtering and classification for airborne LiDAR bathymetry
    YANG Anxiu
    2023, 52(7):  1234-1234.  doi:10.11947/j.AGCS.2023.20220015
    Asbtract ( )   HTML ( )   PDF (747KB) ( )  
    Related Articles | Metrics
News
 
More...
links
More...