[A New Spectral-Spatial Algorithm Method for Hyperspectral Image Target Detection].

Area of Science:

• Remote Sensing
• Image Analysis
• Geospatial Intelligence

Background:

• Hyperspectral remote sensing offers rich spectral and spatial data for target detection.
• Traditional methods struggle with sub-pixel targets and noise, or require significant spatial information.
• Increasing spatial resolution necessitates algorithms that integrate both spectral and spatial characteristics.

Purpose of the Study:

• To propose a novel hyperspectral target detection algorithm.
• To enhance target detection by integrating spectral and spatial features.
• To improve upon traditional target detection operators like CEM and ACE.

Main Methods:

• Combines traditional target detection operators with neighborhood clustering statistics.
• Segments images into potential target and background regions.
• Utilizes centroid-based clustering to refine target identification and exclude background noise.

Main Results:

• The proposed algorithm effectively detects targets by leveraging both spectral signatures and spatial context.
• It outperforms traditional methods like CEM and ACE, particularly for larger targets.
• The approach mitigates background interference in statistical feature extraction.

Conclusions:

• Integrating spectral and spatial characteristics offers a more robust approach to hyperspectral target detection.
• The novel algorithm provides improved performance over existing methods.
• This method is valuable for analyzing high-resolution hyperspectral imagery.