Fusion-based spectral matching method for hyperspectral target detection

Abstract

Target detection using hyperspectral images is one of the most important applications in the field of both civilian and military. Some spectral similarity metric methods have been used in those applications, i.e. spectral angle mapper and spectral gradient angle and so on. However, those methods only consider the similarity between two spectral in one aspect, and many information hidden in the spectrum is ignored. Then, here, the authors propose a new fusion-based spectral-matching method for hyperspectral target-detection task. The proposed Euclidean-angle matching method combines the magnitude and the shape difference together to determine the similarity between the test and target spectra samples. Authors’ results indicate that the proposed fusion-based method exhibits a better performance over the other spectral-matching methods.

Keywords

• image matching
• image fusion
• object detection
• geophysical image processing
• hyperspectral imaging
• feature extraction
• remote sensing
• image classification
• spectral analysis
• hyperspectral target detection
• fusion-based spectral matching method
• spectral-matching methods
• fusion-based method
• target spectra samples
• euclidean-angle matching method
• hyperspectral target-detection task
• fusion-based spectral-matching method
• spectral gradient angle
• spectral similarity metric methods
• hyperspectral images