Robust infrared small target detection via non-negativity constraint-based sparse representation.

Area of Science:

• Computer Vision
• Signal Processing
• Remote Sensing

Background:

• Infrared (IR) small target detection is crucial for military applications like remote sensing and guidance.
• Sparse representation using over-complete dictionaries effectively captures target features.
• Existing methods face challenges in diverse and complex scenes.

Purpose of the Study:

• To develop a robust and effective method for infrared small target detection.
• To improve detection accuracy and reduce computational load in various scenarios.
• To address limitations of current sparse representation techniques in IR target detection.

Main Methods:

• Frequency saliency detection for preprocessing to isolate suspected regions.
• Construction of a target over-complete dictionary using a 2D Gaussian model with feature constraints.
• Sparse representation with non-negativity constraint and l1-regularization solved via accelerated proximal gradient (APG).
• Introduction of a sparse rate (SR) index for adaptive target segmentation.

Main Results:

• The proposed method demonstrates high effectiveness in detecting small IR targets.
• Robustness is shown across a variety of complex background scenes.
• Preprocessing significantly reduces the computational burden for subsequent analysis.

Conclusions:

• The developed sparse representation-based method offers a robust solution for IR small target detection.
• The frequency saliency preprocessing and SR index contribute to improved accuracy and efficiency.
• This technique holds promise for applications in IR remote sensing, early warning, and precise guidance.