Ship Detection from Ocean SAR Image Based on Local Contrast Variance Weighted Information Entropy

Area of Science:

• Remote Sensing
• Maritime Surveillance
• Image Processing

Background:

• Ship detection in Synthetic Aperture Radar (SAR) images is vital for maritime surveillance.
• Challenges include varying ocean waves and strong sea surface echoes, creating heterogeneous and cluttered backgrounds.
• Existing methods struggle with complex backgrounds and high computational costs.

Purpose of the Study:

• To propose an innovative and computationally efficient method for ship detection in SAR images.
• To effectively distinguish vessels from complex backgrounds and strong clutter.
• To achieve stable detection performance in challenging maritime environments.

Main Methods:

• Pre-screening SAR images using the Maximally Stable Extremal Region (MSER) method to identify ship candidate regions.
• Employing a novel Local Contrast Variance Weighted Information Entropy (LCVWIE) metric to assess candidate region complexity and dissimilarity.
• Utilizing an adaptive threshold for final ship detection based on LCVWIE values.

Main Results:

• The proposed method demonstrates stable ship detection performance in both strong clutter and heterogeneous backgrounds.
• Experimental results confirm the effectiveness of the LCVWIE metric in evaluating candidate regions.
• The method achieves low computational complexity compared to existing ship detection techniques.

Conclusions:

• The developed ship detection method offers a robust solution for maritime surveillance using SAR imagery.
• The LCVWIE metric provides an effective way to differentiate ships from complex sea clutter.
• The approach balances detection accuracy with computational efficiency.