一种类圆形图像目标分割方法的研究

Research of Segmentation Algorithms for Quasi-circular Object

研究工业视觉识别系统与图像处理有密切关系。针对类圆形图像背景和目标在色彩和灰度上相近难以准确分割的问题,为了提供精确图像目标,提出了一种新型的类圆形目标分割方法。采用带方向的序贯滤波和方向边缘检测,有效去除背景干扰信息;边缘检测的结果填充后进行图像投影,据此得到待分割目标的大致区域,再次滤波得到较佳的预分割结果;采用中心扫描算法,接合投影曲线完成粗分割,得到边缘曲线;最后根据边缘曲率对边缘线进行矫正,实现了类圆形目标的完整提取。通过实验表明,对灰度相近的图像能进行有效提取,避免了阈值法的局限性,提高了类圆形目标分割的完整性和精确度。

Aiming at the images which background is similar as the objects on color and gray value.This paper proposed a new segmentation method aimed at accumulation of quasi-circular object.It firstly carries out a directional sequential filtering algorithm and edge detection to avoid the disturbance from background.Secondly the edges are filled with the result of edge detection and a filter processing is made to get a pretreatment result.Then the scan segmentation algorithm based on the primary result is use to get a cur-sory segmentation area curve of edge.And the regional border lines are rectified with the method which based on the curvature changes of the marginal line.Experiments prove that it effectively segments the images which backgrounds are close to the objects and improves the accuracy of segmentation.