Sidescan Sonar Image Target Extraction Method Based on Variable Initial Signed Distance Function-Based Active Contour CV Model

Xue Du,Qiuting Gong,Renfeng Jia,Hongzhou Liao

DOI: https://doi.org/10.1109/icma.2018.8484332

2018-01-01

Abstract:UUV uses side scan sonar to perform search tasks. Due to the low contrast, blurring edges, color loss, and serious noise of sonar images, the UUV will produce noise gray values that are similar to the target and randomly distributed along the target and even cover the target. Due to the characteristics, the edge of the search target of the side scan sonar is blurred, and it is difficult to effectively detect and recognize it. The geometric active contour model is based on the curve evolution theory and the level set method. The curve evolution only depends on its inherent geometric characteristics. By implicitly representing the closed curve as the zero level set of the level set function in the high dimensional surface, the level set is used. The evolution of the function to obtain the evolution of the curve solves the problem of topological changes. This paper proposes a self-seeking method for level set initial contour curve. The CV model using global variables is used to extract the contours of sonar images, and the irregular blocky noises appearing in the vast majority of split sonar image objects are filtered out. Suppresses the possibility of irregular block noise around the target being segmented, and achieves the segmentation effect of the contour-fitting target, clear boundaries, and minimal noise.