Dual-Modal Approach for Ship Detection: Fusing Synthetic Aperture Radar and Optical Satellite Imagery.

Area of Science:

• Remote Sensing
• Computer Vision
• Maritime Surveillance

Background:

• Optical imagery offers high resolution but is limited by weather and light conditions.
• Synthetic Aperture Radar (SAR) imagery performs well in adverse conditions but suffers from noise.
• Existing fusion methods often fail to exploit the complementary strengths of SAR and optical data for ship detection.

Purpose of the Study:

• To develop a novel fusion framework for enhanced ship detection using both SAR and optical satellite imagery.
• To improve the accuracy and processing speed of ship detection in diverse maritime scenarios.
• To effectively integrate the complementary strengths of SAR and optical data.

Main Methods:

• Optical image detection: Contrast Limited Adaptive Histogram Equalization (CLAHE) with YOLOv7.
• SAR image detection: Customized Detection Transformer (SAR-EDT) with denoising and optimized pooling.
• Fusion module: Intersection over Union (IoU) for bounding box overlap, confidence score averaging, and duplicate elimination.

Main Results:

• The proposed framework significantly improves ship detection accuracy.
• Enhanced performance across various maritime scenarios, overcoming limitations of individual modalities.
• Optimized processing speed for optical image detection using CLAHE and YOLOv7.

Conclusions:

• The novel fusion framework effectively leverages complementary SAR and optical data for superior ship detection.
• The integration of specialized models and a robust fusion module addresses limitations of previous approaches.
• This research advances maritime surveillance capabilities through improved automated ship detection.