Sea-Land Clutter Segmentation Algorithm Based on Multi-measure Fusion with SVM Classifier

Abstract

Effectively segmenting sea clutter and land clutter in the sea-land junction area is of great significance for target detection and recognition on the sea surface. Existing sea-land clutter segmentation algorithms are mostly based on a single measure, of which the segmentation effect is not very satisfactory. In view of this problem, this paper proposes a novel sea-land clutter segmentation algorithm based on multi-measure fusion. Firstly, the characteristics of the clutter in the echo data collected by the sea detection radar are analyzed, and multiple appropriate segmentation measures are selected as feature vectors and fed into the Support Vector Machine (SVM) classifier. Then the classification result is converted into a binary image and processed by morphological filtering method to ensure the connectivity between the sea clutter area and the land clutter area. Finally, the feasibility and validity of the algorithm are verified by the real radar data.