Rain Detection From X-Band Marine Radar Images

A method based on support vector machines (SVMs) is proposed to detect rain from X-band marine radar images. First, each image is labeled as rain-free or rain-contaminated according to the simultaneous rain rate data provided by a rain gauge. Then, feature vector is extracted from each image, which consists of the normalized histogram bin values. The SVM-based rain detection model is trained and tested using radar data collected in a sea trial in the vicinity of the east coast of Canada. Compared to the zero pixel percentage (ZPP) based method, the detection accuracy are shown to be improved by the proposed method. In addition, the method is robust for a wide range of bin number values even with limited training samples.