Realizing Target Detection in SAR Images Based on Multiscale Superpixel Fusion

Area of Science:

• Remote Sensing
• Image Processing
• Signal Processing

Background:

• Synthetic Aperture Radar (SAR) imagery presents challenges for target detection due to false alarms in complex land areas, particularly near coastlines.
• Existing methods struggle to effectively differentiate targets from land-based clutter in SAR images.

Purpose of the Study:

• To develop an advanced algorithm for accurate SAR target detection in complex scenes.
• To specifically address and mitigate the issue of false alarms in land areas of SAR images.

Main Methods:

• Proposed an algorithm utilizing the fusion of multiscale superpixel segmentations for SAR image analysis.
• Implemented land-sea segmentation based on statistical properties of superpixels at different scales.
• Integrated land-sea segmentation results with Constant False Alarm Rate (CFAR) detection to eliminate false alarms.
• Fused detection results from multiple scales to enhance overall detection robustness.

Main Results:

• The proposed algorithm demonstrated significant effectiveness in reducing false alarms in land areas of SAR images.
• Experimental results on real SAR data confirmed the algorithm's capability for reliable target detection.
• The fusion of multiscale segmentations improved the robustness of the detection process.

Conclusions:

• The developed algorithm successfully overcomes the challenge of land-based false alarms in SAR target detection.
• The fusion of multiscale superpixel segmentations offers a robust approach for enhancing SAR image analysis.
• This method provides a valuable tool for accurate target identification in complex SAR environments.