基于水平集改进的水下目标轮廓提取方法

An improved method of underwater objects contours extraction based on level set

摘要:

为以更快的计算速度提取到更加精确的轮廓,提出一种基于水平集改进的水下目标轮廓提取方法.即利用水下目标检测结果确定目标演化子区域,缩小目标区域范围,同时在目标检测结果中,根据目标高亮区和阴影区的位置,确定各个目标演化子区域初始闭合曲线的中心坐标,通过Vese-Chan分段常量四相水平集方法的演化函数进行目标高亮区和阴影区的轮廓提取.对不同原始声纳图像的实验比较分析表明,提出的水下目标轮廓提取方法具有较高的适应性和较快的计算速度,能精确地提取到目标高亮区和阴影区的轮廓.

Abstract:

In order to improve the accuracy of extracted contours and raise calculation speed,this paper presents an improved method based on level set for underwater objects contours extraction.Object evolution subregion,which is determined by the underwater objects detection results,is proposed to dwindle the search region.At the same time,the centers coordinates of initial closed curves of each object evolution sub-region are determined by the location of object-highlight and shadow regions.The contours of object-highlight and shadow regions are extracted by the four-phase piecewise constant Vese-Chan level set evolution functions.Experimental results and analysis of different kinds of real sonar images demonstrate that the method proposed here is highly adaptable and the calculation speed is faster,moreover,it can more accurately extract object-highlight and shadow contours.