Spectral-Spatial Enhanced Local Contrast Strategy for Hyperspectral Small Air Target Detection

Area of Science:

• Remote Sensing
• Signal Processing
• Computer Vision

Background:

• Detecting small aerial targets is crucial for civil aviation but challenging due to weak target signatures.
• Hyperspectral imaging (HSI) offers a promising solution by capturing rich spatial and spectral information.

Purpose of the Study:

• To develop an effective hyperspectral small air target detection strategy.
• To enhance the ability to distinguish targets from background noise in HSI data.

Main Methods:

• Proposed a spectral-spatial enhanced local contrast strategy for HSI small air target detection.
• Developed an unsupervised band selection method based on local contrast (LC-UBSM).
• Introduced an improved RX detection algorithm incorporating combined spatial and spectral variance (CSSV-RX).

Main Results:

• The LC-UBSM method effectively selects bands with high target-background distinguishability.
• The CSSV-RX algorithm successfully detects targets while suppressing background and noise.
• Validated the method's effectiveness and robustness on GAOFEN-5 and EO-1 satellite datasets.

Conclusions:

• The proposed spectral-spatial enhanced local contrast strategy significantly improves hyperspectral small air target detection.
• The combined approach of advanced band selection and improved RX algorithm offers a robust solution for real-world applications.