GUO Guankai，LIU Wei，YU Lingling，XU Haowen，WANG Wenjing

(School of Electrical and Electronic Engineering, Shandong University of Technology, Zibo 255049, China)

Abstract: In order to segment touching particles accurately and efficiently, an improved FAST feature point detection algorithm was proposed, which was combined with the watershed algorithm based on h-maxima transformation to realize the segmentation of touching particles. First, the candidate segmentation points and lines were determined by the h-maxima watershed algorithm. Next, the feature points of the edge image were obtained by traversing edges of the binary image based on the improved FAST feature point detection algorithm. Then, the method of non-maximum suppression was used to gain pending segmentation points from all the candidate ones according to the distribution of feature points. The candidate segmentation lines were selected by the pending segmentation points to get the pending segmentation lines. Finally, the pseudo segmentation points in the pending segmentation lines were identified according to the edge feature points and the segmentation lines connected with the pseudo segmentation points were removed, and the correct segmentation lines were determined. Segmentation experiments on images of representative particles were carried out, compared with the three existing image segmentation algorithms. Experiment results show that computation time of the improved algorithm is the shortest and the segmentation accuracy is greater than 95%.

Keywords: touching particles; image segmentation; edge feature point; edge segmentation point; segmentation line

中图分类号：TP391

文献标志码：A

收稿日期： 2018-10-11，修回日期：2018-10-31，在线出版时间：2019-01-04 08:44。

基金项目：山东省自然科学基金项目，编号：ZR2017LF026。

第一作者简介：郭观凯(1986—)，男，硕士研究生，研究方向为颗粒图像处理及检测仪器研发。E-mail：guoguankai@163.com。

通信作者简介：刘伟(1971—)，男，博士，副教授，硕士生导师，研究方向为颗粒测量技术。E-mail：weikey@sdut.edu.cn。