Article
Geochemistry & Geophysics
Bin Cui, Yao Peng, Hao Zhang, Wenmei Li, Peijun Du
Summary: This paper proposes a spectral-spatial convolutional network based on MRF and CoV for HSI classification. By combining MRF models and CoVs with spectral information, the fused features are used as input to produce reliable classification results.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Geochemistry & Geophysics
Chao Pan, Xiuping Jia, Jie Li, Xinbo Gao
Summary: This article introduces a novel approach for addressing the over-smoothing issue in MRF by class-by-class refinement and adaptive edge preservation. Experimental results demonstrate the superiority of aEPMs in evaluation metrics and detail preservation compared to traditional methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Geochemistry & Geophysics
Yujia Chen, Linlin Xu, Yuan Fang, Junhuan Peng, Wenfu Yang, Alexander Wong, David A. Clausi
Summary: The letter introduces a novel Bayesian approach for unsupervised subpixel mapping of hyperspectral imagery based on MRF and BDSMM, which allows adaptive adjustment of abundance and endmember information and fully explores endmember-abundance patterns in HSI to improve SPM performance. Experiments demonstrate better performance compared to traditional methods.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2021)
Article
Engineering, Electrical & Electronic
Huan Liu, Wei Li, Xiang-Gen Xia, Mengmeng Zhang, Chen-Zhong Gao, Ran Tao
Summary: This article addresses the issue of spectral shift in cross-scene hyperspectral imagery classification by proposing spectral shift mitigation (SSM) that includes amplitude shift mitigation (ASM) and adjacency effect mitigation (AEM). By reducing amplitude shift and spectral variation through amplitude normalization and weighted average spectral vector methods, and using a classifier trained with labeled samples from the source scene, superior classification performance is achieved on several cross-scene HSI data pairs.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)
Article
Engineering, Electrical & Electronic
Yujia Chen, Rongming Zhuo, Linlin Xu, Junhuan Peng, Xiaoman Qi, Zhaoxu Zhang, Zhongzheng Hu
Summary: The study proposes a new Bayesian subpixel mapping method that considers endmember variability, discrete subpixel class labels, and spatial information in HSI. Experimental results demonstrate that the method outperforms previously available SPM techniques.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Bing Tu, Chengle Zhou, Xiaolong Liao, Guoyun Zhang, Yishu Peng
Summary: The study introduces a novel spatial-spectral classification method for hyperspectral images based on structural-kernel collaborative representation (SKCR), which considers a weak assumption of spatial neighborhood. The method utilizes superpixel segmentation and dual kernels to achieve excellent classification performance even with relatively small training samples.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2021)
Article
Geochemistry & Geophysics
Fei Tong, Yun Zhang
Summary: In recent years, deep learning methods have been widely used in hyperspectral image (HSI) classification. A new classification method called spectral-spatial deep RF (SSDRF) is proposed to fully utilize the spatial information in HSIs for improved classification accuracy, combining fixed-size patches with shape-adaptive superpixels to exploit more accurate spatial information. This approach outperforms patched-based DCDRF and achieves satisfactory classification results.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Leyuan Fang, Yinglong Yan, Jun Yue, Yue Deng
Summary: Hyperspectral images (HSIs) provide rich spectral-spatial information and have been widely used in various fields. However, existing HSI interpretation methods often result in high memory costs and difficulties in editing. To address this, a novel end-to-end vectorization framework called HSI-VecNet is proposed, which integrates low-level geometry information and high-level semantic instance information to learn a vector representation from spectral-spatial data. Experimental results on four hyperspectral datasets show that the proposed method outperforms existing post-process vectorization methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Engineering, Electrical & Electronic
Yihang Lu, Xuan Zheng, Haonan Xin, Haoliang Tang, Rong Wang, Feiping Nie
Summary: In recent years, hyperspectral anomaly detection (HAD) has played a significant role in military and civilian fields. The collaboration representation-based detector (CRD) is a classic HAD method, but its computational cost and adaptability limitations have posed challenges. To address this, we propose a novel ensemble and random collaborative representation-based detector (ERCRD) that reduces computational complexity and improves accuracy through random sub-sampling and ensemble learning.
Article
Environmental Sciences
Xin He, Yushi Chen, Zhouhan Lin
Summary: This study explores the application of Transformer-based models in hyperspectral image classification and proposes two new classification frameworks to address the issues of sequential data processing and spatial feature extraction. Experimental results show that the proposed models are competitive compared to state-of-the-art methods, indicating the new potential of Transformer in HSI classification.
Article
Computer Science, Artificial Intelligence
Jun Liu, Zengfu Hou, Wei Li, Ran Tao, Danilo Orlando, Hongbin Li
Summary: This article discusses anomaly detection in hyperspectral imagery using two adaptive detectors, both of which are found to be equivalent. Analytical expressions for the false alarm probability of the detectors are derived, showing a constant false alarm rate against the noise covariance matrix. Experimental results demonstrate that the proposed detector performs better than its counterparts in detecting anomalies in real hyperspectral data sets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Geochemistry & Geophysics
Baokai Zu, Yafang Li, Jianqiang Li, Ziping He, Hongyuan Wang, Panpan Wu
Summary: In this paper, a cascaded convolution-based transformer with densely connected mechanism (CDCformer) is proposed for hyperspectral image classification. It introduces cascaded convolution feature tokenization and a densely connected transformer to enhance feature propagation and extract more discriminative spectral-spatial information from the HSI.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Remote Sensing
Lanxing Wang, Qunming Wang
Summary: In this paper, a fast spatial-spectral random forests (FSSRF) method is proposed for cloud removal in hyperspectral images. The FSSRF method improves computational efficiency while maintaining accuracy. Experimental results demonstrate that FSSRF outperforms other methods in terms of speed and prediction accuracy.
INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION
(2022)
Article
Geochemistry & Geophysics
Fei Tong, Yun Zhang
Summary: This article proposes a spectral-spatial and cascaded multilayer random forests (SSCMRF) method for classifying tree species in high-spatial-resolution hyperspectral images. The method achieves superior classification results by fully utilizing spatial information from shape-adaptive superpixels and shape-fixed patches, integrating two different types of spatial information.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Xianghai Cao, Zuji Liu, Xiangxiang Li, Qian Xiao, Jie Feng, Licheng Jiao
Summary: Most studies on hyperspectral imagery classification focus on improving accuracy, overlooking the influence of sampling strategy. Random sampling (RS) is commonly used, but is not feasible in practical applications where training and test samples are collected from different locations, leading to decreased performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Engineering, Industrial
Xinlai Liu, Yishuo Jiang, Zicheng Wang, Ray Y. Zhong, H. H. Cheung, George Q. Huang
Summary: This paper proposes a unified five-layer blockchain-enabled secure digital twin platform architecture for small and middle enterprises (SMEs) in the manufacturing industry to overcome the limitations of traditional manufacturing patterns. The experimental results show that the proposed platform, named imseStudio, effectively digitizes manufacturing resources and promotes the transformation towards service manufacturing.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
(2023)
Article
Computer Science, Artificial Intelligence
Qianqian Wang, Zhiqiang Tao, Wei Xia, Quanxue Gao, Xiaochun Cao, Licheng Jiao
Summary: In this article, we propose an adversarial MVC (AMvC) network to address the challenge of extracting consistent latent representations over multiple views for clustering. The AMvC network generates each view's samples based on fused latent representations and achieves a more consistent clustering structure. Experimental results show that our AMvC method outperforms several state-of-the-art deep MVC methods on video, image, and text datasets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Ronghua Shang, Jiarui Kong, Lujuan Wang, Weitong Zhang, Chao Wang, Yangyang Li, Licheng Jiao
Summary: This paper proposes an unsupervised feature selection method, FSDSC, which integrates discrete spectral clustering and feature weights. The method combines regression models and spectral clustering in a unified framework and introduces a feature weight matrix to improve feature selection performance.
Article
Automation & Control Systems
Vincent Havyarimana, Zhu Xiao, Thabo Semong, Jing Bai, Hongyang Chen, Licheng Jiao
Summary: This article proposes a reliable fusion technique, called non-Gaussian Redheffer weighted least squares (nGRWLSs), for intervehicle positioning estimation in various GNSS outage environments. The method combines the Gaussian dynamical matrix principle and the Redheffer distribution function to accurately estimate their positions.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Automation & Control Systems
Ronghua Shang, Weitong Zhang, Jingwen Zhang, Licheng Jiao, Yangyang Li, Rustam Stolkin
Summary: This article proposes a new local community detection algorithm that utilizes alternating strong fusion and weak fusion strategies to fuse nodes, improving the solution in each stage. A new membership function is proposed in the strong fusion phase, considering both node information and connection information, leading to higher quality fused nodes while preserving community structure. In the weak fusion phase, a parameter-based similarity measure is proposed to detect influential nodes in local communities. Additionally, a local community evaluation metric is proposed that does not require true division to determine the optimal local community under different parameters.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Automation & Control Systems
Shuo Li, Fang Liu, Licheng Jiao, Xu Liu, Puhua Chen
Summary: This paper introduces an unsupervised salient object detection method that achieves salient object detection by learning salient features from the data itself. The method enhances salient features, suppresses nonsalient features, and roughly locates the salient features to obtain the salient activation map. A saliency map update strategy is then used to remove noise and strengthen boundaries. The results show that the proposed method can effectively learn salient visual objects.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Automation & Control Systems
Changzhe Jiao, Chao Chen, Shuiping Gou, Xiuxiu Wang, Bo Yang, Xiaoying Chen, Licheng Jiao
Summary: This article proposes an L1 sparsity-regularized attention multiple-instance neural network (L1-attention MINN) for hyperspectral target detection with imprecise labels. The proposed algorithm achieves superior performance in both simulated and real-field scenarios, demonstrating its effectiveness in handling imprecisely labeled hyperspectral data.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Computer Science, Artificial Intelligence
Haoran Wang, Licheng Jiao, Fang Liu, Lingling Li, Xu Liu, Deyi Ji, Weihao Gan
Summary: This article explores the problem of social relation recognition in an open environment and introduces a new video dataset and a corresponding neural network architecture that can effectively recognize social relations at both individual and pair levels.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Automation & Control Systems
Qigong Sun, Xiufang Li, Licheng Jiao, Yan Ren, Fanhua Shang, Fang Liu
Summary: This article proposes a novel sequential single-path search (SSPS) method for mixed-precision model quantization, which introduces given constraints to guide the searching process and improves search efficiency and convergence speed. The experiments demonstrate that the method significantly outperforms uniform-precision models for different architectures and datasets.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Geochemistry & Geophysics
Zhonghua Li, Biao Hou, Zitong Wu, Bo Ren, Zhongle Ren, Licheng Jiao
Summary: This article proposes a Gaussian OBB algorithm for object detection in aerial image scenes, which eliminates border shift and improves the performance and accuracy of the detector through synthesis and decoding methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Xiaotong Li, Licheng Jiao, Hao Zhu, Zhongjian Huang, Fang Liu, Lingling Li, Puhua Chen, Shuyuan Yang
Summary: This study proposes a novel dynamic polar spatio-temporal encoding method to improve the tracking performance of visual Transformer models in video scenes. By utilizing spiral functions in polar space and a dynamic relative encoding mode for continuous frames, the method captures the spatio-temporal motion characteristics among video frames more effectively.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Jing Bai, Junjie Ren, Zhu Xiao, Zheng Chen, Chengxi Gao, Talal Ahmed Ali Ali, Licheng Jiao
Summary: In recent years, there has been increasing attention on object localization and detection methods in remote sensing images (RSIs) due to their broad applications. Weakly supervised object localization (WSOL) is a cost-effective alternative to fully supervised methods as it only requires image-level labels instead of time-consuming and labor-intensive instance-level annotations. In this article, a self-directed weakly supervised strategy (SD-WSS) is proposed to perform WSOL in RSIs by enhancing the spatial feature extraction capability of the RSIs' classification model and utilizing GradCAM++ to address the discriminative region problem. A novel self-directed loss is also designed to eliminate interference from complex backgrounds. Additionally, new WSOL benchmarks in RSIs, named C45V2 and PN2, are created to evaluate the proposed method alongside six mainstream WSOL methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Geochemistry & Geophysics
Zhonghua Li, Biao Hou, Zitong Wu, Zhengxi Guo, Bo Ren, Xianpeng Guo, Licheng Jiao
Summary: Traditional 2-D Gaussian distribution loses angular information when dealing with square-like objects, leading to inaccurate localization. To address this issue, this study modifies the 2-D Gaussian function using the Lame curve to create a super-Gaussian distribution. This distribution maintains angular information at arbitrary aspect ratios, and the distance between two super-Gaussian distributions is measured using KL divergence, converted into localization loss. Experimental results on multiple datasets confirm the effectiveness of the proposed algorithm, achieving state-of-the-art performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Dou Quan, Huiyuan Wei, Shuang Wang, Yu Gu, Biao Hou, Licheng Jiao
Summary: This article proposes a deep learning registration framework for multimodal remote sensing images, which utilizes different deep models for coarse and fine registration stages. Experimental results demonstrate the significant performance advantages of this framework in rotation correction and modality change.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Yunpeng Li, Xiangrong Zhang, Xina Cheng, Xu Tang, Licheng Jiao
Summary: Tremendous progresses have been made in remote sensing image captioning (RSIC) task in recent years. This work focuses on injecting high-level visual-semantic interaction into RSIC model. The experiments on three benchmark data sets show the superiority of our approach compared with the reference methods.
PATTERN RECOGNITION
(2024)
