研究目的
To propose a novel constrained detection algorithm, referred to as Spatial-Sparse CEM, to simultaneously force the sparsity of the output and piecewise continuity via proper regularizations for target detection in hyperspectral images.
研究成果
The proposed Spatial-Sparse CEM integrates a spatial regularization to the CEM detection problem, while forcing the output to be sparse via the sparsity regularization term. The problem is solved by ADMM method. Improved performance of the Spatial-Sparse CEM algorithm is illustrated by testing with several hyperspectral images. In the future, other spatial regularization methods and adaptive parameter selection methods for the algorithm will be considered.
研究不足
The technical and application constraints of the experiments, as well as potential areas for optimization, were not explicitly mentioned in the paper.
1：Experimental Design and Method Selection:
The proposed Spatial-Sparse CEM algorithm integrates a spatial regularization to the CEM detection problem, while forcing the output to be sparse via the sparsity regularization term. The problem is solved by ADMM method.
2：Sample Selection and Data Sources:
Synthetic hyperspectral data and real hyperspectral dataset (Samson) were used.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
The algorithm was applied to both synthetic and real hyperspectral images, and its performance was compared with classical CEM and Sparse CEM.
5：Data Analysis Methods:
Receiver operating characteristics (ROC) curves representing detection probability versus false-alarm rates were generated for quantitative performance comparison.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
