基于改进分水岭的焊缝分割算法

Weld Segmentation Algorithm Based on Improved Watershed

针对传统分水岭算法处理背景复杂的焊缝图像时存在的过度分割问题,提出基于形态学重构的改进分水岭的焊缝分割算法。核心思想是对二值化的焊缝图像进行形态学处理,再用基于距离场的分水岭算法分割处理;将所得图像进行阈值分割,并与原二值化图像求逻辑"与"运算,之后搜索图中的颗粒并标记,利用颗粒的相关信息找出焊缝颗粒的标记,将标记依次转成ROI、掩模;最后利用掩模运算完成焊缝提取。经30组试验对比验证,该方法在处理复杂背景的平角焊缝图片时能显著提高焊缝分割的精确度,达96.7%。

Aiming at the over segmentation problem of traditional watershed algorithm in dealing with weld images with complex background,an improved watershed weld segmentation algorithm based on morphological reconstruction is proposed.The core idea is to carry out morphological processing on the binary weld image,and then segment it by watershed algorithm based on distance field.The obtained image is segmented by threshold,and the logical"and"operation is calculated with the original binary image.Then,the particles in the image are searched and marked,the marks of weld particles are found by using the relevant information of particles,and the marks are transformed into ROI and mask in turn.Finally,the mask operation is used to extract the weld.Compared with 30 groups of experiments,this method can significantly improve the accuracy of weld segmentation,up to 96.7%.