Artificial Structure Detection Based on MSER in SAR Images

In this study, a maximally stable extremal region (MSER)-based detection algorithm is proposed for efficient artificial structure detection in synthetic aperture radar images. The algorithm consists of three stages: First, the MSER algorithm identifies candidate regions of artificial structures and labels them as superpixels to enhance the processing efficiency. This stage reliably detects structures of various sizes and shapes. Second, analyzing the pixel count and average intensity within superpixels distinguishes natural clutter from artificial structures, with the OTSU method setting an optimal threshold for removing clutter. Finally, fuzzy C-means clustering refines the separation between artificial structures and clutter, flexibly handling ambiguous boundaries. The experimental results confirm that the proposed algorithm reduces false alarms and improves detection compared with traditional constant false alarm rate (CFAR) methods.