A5059495707- Advanced Image Fusion Techniques
- Remote-Sensing Image Classification
- Remote Sensing and Land Use
- Image and Signal Denoising Methods
- Image Retrieval and Classification Techniques
- Geochemistry and Geologic Mapping
- Advanced Data Compression Techniques
China University of Petroleum, East China
2022-2024
Huzhou Vocational and Technical College
2009
Autoencoders (AEs) are commonly utilized for acquiring low-dimensional data representations and performing reconstruction, which makes them suitable hyperspectral unmixing. However, AE networks trained pixel by those employing localized convolutional filters disregard the global material distribution distant interdependencies, resulting in loss of necessary spatial feature information essential unmixing process. To overcome this limitation, we propose an innovative deep neural network model...
Deep learning (DL) has gained popularity in hyperspectral unmixing (HU) applications recently due to its powerful and data-fitting capabilities. As an baseline network, the autoencoder (AE) framework performs well HU by automatically low-dimensional embeddings reconstructing data. Nevertheless, there are spectral variability nonlinear mixing problems highly mixed region of images, which can cause interference structures using only AE. Therefore, inspired effectiveness mask modeling, we...
Autoencoders are widely utilized in hyperspectral unmixing as an unsupervised end-to-end learning model. In particular, convolutional autoencoder networks popular for processing multidimensional features. Nonetheless, the traditional Autoencoder network's receptive field is constrained task, and establishing connection between local spatial neighborhood spectrum fails to improve performance significantly. To address these limitations, a bilateral global attention network based on both...
Due to the low resolution of hyperspectral images, problem mixed pixels is common, and unmixing a crucial technology solve pixels. Among them, nonnegative matrix factorization (NMF) widely used because it can simultaneously perform endmember abundance estimations. As variant NMF, archetype analysis (AA) find most representative sample in dataset, which has strong interpretability compared with NMF. However, traditional AA-based methods consider only one spectral curve represent class,...
Hyperspectral unmixing is a key technology in the development of remote sensing applications. However, since both endmembers and abundances are unknown, non-convex problem with large solution space. To solve this, existing methods usually impose same strength sparsity constraint. this often does not hold practice. Because purer regions generally sparse, while distribution more mixed should be smoother. Temperature scaling technique introducing temperature parameter T into softmax activation...
Hyperspectral unmixing (HU) is widely used to process mixed pixels as an essential technology. Among them, the nonnegative matrix factorization (NMF)-based approach one typical of blind techniques, which can achieve endmembers and abundances simultaneously. Considering physical meaning extracted endmembers, archetypal analysis (AA) method constructs a new decomposition structure with stronger interpretability than NMF. However, AA ignores significant sparse property abundance in unmixing....
Paper presents a method for color image compression coding by using correlation vectors combined with SPIHT Algorithm. In this method, the Green component of is coded Algorithm, and other two components are segmented into multi-subregions quadtree partitioning according to non-correlation degree three components. Based on calculated between in same subregion, becomes vectors. Experiment results show that high decoding rates, quite good signal noise ratio, rate visual quality.