A novel algorithm of water region detection in SAR image based on bag of visual words and local pattern histogram

Water region detection based on SAR images is a difficult problem for its computing complexity. This paper proposes a novel water region detection method in SAR image of complex scenery. The algorithm takes advantages of Bag of visual words (BOV) to precisely describe the homogeneous region in complex scenery. Local pattern histogram (LPH) and single-class Support vector machine (SVM) are adopted to determine the edge information of water region precisely. The feature extraction is calculated block by block, which reduces computing workload and interference from noise. The experiments based on SAR images of real complex scenery show that the proposed method achieves higher accuracy and robustness.