Thermal infrared (TIR) remote sensing is an important technology for detecting geothermal anomalies. However, detection results have been found to have a certain dependency on the distribution of ground objects and imaging conditions at different times, and pseudo-anomalous areas are easily extracted. To solve this problem, a new geothermal anomaly detection method is proposed in this article and implemented in the Jiaonan uplift in the Yishu fault zone. A temperature inversion experiment is carried out on TIR remote sensing images based on the radiation transfer equation. Then, a gradient operator is used to extract high-temperature regions in various periods, and geothermal anomaly areas are selected through the spatio-temporal analysis method proposed in this article after excluding the influence of impervious surfaces, water bodies, and vegetation. The temperature anomaly points, which are all high-temperature points in each inversion result and geothermal anomaly areas extracted by the proposed method, are compared with the five geothermal anomaly points, which are determined based on 150 geothermal wells and the geological structure in the study area. The spatial locations of the temperature anomaly points and geothermal anomaly areas are close to those of the geothermal anomaly points. Compared with the mean grad method, the proposed method is found to effectively delete some pseudo-anomaly areas under the premise of ensuring the extraction accuracy of geothermal anomaly areas.

3D mesh classification deep neural network (3D DNN) has been widely applied in many safety-critical domains. Backdoor attack is a serious threat that occurs during the training stage. Previous backdoor attacks from 2D image and 3D point cloud domains are not suitable for 3D mesh due to data structure restrictions. Therefore, in a pioneering effort, this paper presents two types of backdoor attacks on 3D mesh. Specifically, the first attack is a Mesh Geometrical Feature guided 3D Mesh Backdoor Attack named MGF-MBA . Most 3D DNNs have to convert 3D mesh to a regular matrix (mesh geometrical feature), which is a refinement of the input 3D mesh. The 3D DNN directly learns the 3D shape from the mesh geometrical feature, which enables attackers to implant backdoor through it. Hence, the proposed MGF-MBA generates a backdoored 3D mesh under the guidance of mesh geometrical feature. The second attack is a Remeshing based 3D Mesh Backdoor Attack named ReMBA . The quality of samples backdoored by exiting backdoor attacks always decrease. Although many efforts have been made to reduce the descent in quality in return for stealthiness, the descent persists. For better stealthiness, we regard the backdoor implantation process as a way to increase the quality of backdoored sample rather than a way to reduce it. Specifically, ReMBA designs a new isotropic remeshing method that attempts to represent a 3D mesh by equilateral triangles while keeping the number of vertices, edges and faces unchanged. Numerous experimental results show that both MGF-MBA and ReMBA achieve guaranteed attack performance on 3D DNNs. Furthermore, transferability experiments demonstrate that ReMBA can even attack 3D point cloud networks with an increased ability to resist defenses.

With the improvement of spatial resolution of hyperspectral remote sensing images, the influence of spectral variability is gradually appearing in hyperspectral unmixing. The shortcomings of endmember extraction methods using a single spectrum to represent one type of material are revealed. To address spectral variability for hyperspectral unmixing, a multiscale resampling endmember bundle extraction (MSREBE) method is proposed in this paper. There are four steps in the proposed endmember bundle extraction method: (1) boundary detection; (2) sub-images in multiscale generation; (3) endmember extraction from each sub-image; (4) stepwise most similar collection (SMSC) clustering. The SMSC clustering method is aimed at solving the problem in determining which endmember bundle the extracted endmembers belong to. Experiments carried on both a simulated dataset and real hyperspectral datasets show that the endmembers extracted by the proposed method are superior to those extracted by the compared methods, and the optimal results in abundance estimation are maintained.

Hyperspectral unmixing (HU) is widely used to process mixed pixels as an essential technology. Among them, the nonnegative matrix factorization (NMF)-based approach is one typical of the blind unmixing techniques, which can achieve endmembers and abundances simultaneously. Considering the physical meaning of the extracted endmembers, the archetypal analysis (AA) method constructs a new matrix decomposition structure with stronger interpretability than NMF. However, AA ignores the significant sparse property of abundance in unmixing. Therefore, we propose the L ₁ sparsity-constrained AA algorithm for HU. To solve the new optimization problem, we explore a new optimization method for optimizing abundance. The alternating direction method of multipliers (ADMM) is used to increase the strong convexity and convergence of the problem. Then the fast gradient method (FGM) instead of traditional gradient descent is used to speed up algorithm convergence. The experimental results in both the synthesized and real datasets show that the proposed method outperforms several sparse NMF-based and AA-based methods.

Mixed pixels are the main reason for the low accuracy of traditional remote sensing applications. Hyperspectral image unmixing can explore the sub-pixel information of mixed pixels, which is an effective measure to solve the problems of low precision. Current unmixing methods only consider the correlation of local or global pixels and do not fully utilize images information. Based on the above problems, we proposed a novel sparse unmixing model named combining low-rank constrain for similar superpixels and total variation sparse unmixing (CLRSS-TV) in this current research paper. The adjacent similar pixels are clustered first into superpixels, then the superpixels with the same semantic information are merged into a collection. The spatial information of the image is fully considered by twice clustering. Weighted low-rank constraint is imposed on each collection to consider the spectral correlation of pixels. Moreover, the TV regularization and l_2,1 norm are utilized to improve the smoothness and row sparsity, respectively. Experimental results reveal that the proposed method is not only competitive with the four state-of-the-art algorithms in simulated and actual datasets, but also has good anti-noise ability and universality. However, this paper still fails to overcome the problem of parameters selection in sparse unmixing, so the future work study intends to adopt the intelligent optimization algorithm to obtain the optimal parameters.