全文下载： 
202206015.pdf
文章编号： 1672-6987（2022）06-0109-10； DOI： 10.16351/j.1672-6987.2022.06.015
隋鹏飞， 滕腾， 樊春玲*(青岛科技大学 自动化与电子工程学院，山东 青岛 266061)
摘要： 由于矿山爆破不充分，需要对残留岩石进行识别敲击，采用人工处理费时费力、且存在人身安全隐患，因此本文提出基于机器视觉的矿井岩石裂缝识别算法对裂缝进行识别以便后续裂缝清除。首先结合粒子群算法对矿井岩石裂缝图像进行阈值分割，然后通过模板匹配法对Zhang算法进行改进形成单一光滑像素骨架来寻找边缘断点。结合裂缝断处部分占整条裂缝比例较小，形态学操作可对断点产生影响，对断点处利用生长结构元素进行多次生长迭代实现连接，还原裂缝的连续性。本工作样本均从矿井实地采集，并挑选出具有代表性的两类裂缝图片样本，一类为规则裂缝（线条型裂缝），另一类为不规则裂缝（网状型裂缝），最后通过试验分析可得规则裂缝识别率在93%，不规则裂缝识别率79%，虽然不规则细小裂缝（宽度小于1 mm）的识别较难，但是由于此种裂缝危险系数较小可不进行处理，因此文中所提算法能够满足工程应用需求。
关键词： 裂缝识别； 阈值分割； 骨架提取； 模板匹配； 生长操作
中图分类号： TP 751.1文献标志码： A
引用格式：隋鹏飞，滕腾，樊春玲. 基于机器视觉的矿井岩石裂缝识别算法［J］. 青岛科技大学学报(自然科学版), 2022, 43(6): 109-118.
SUI Pengfei, TENG Teng, FAN Chunling. Mine rock crack recognition algorithm based on machine vision［J］. Journal of Qingdao University of Science and Technology（Natural Science Edition), 2022， 43（6）： 109-118.
Mine Rock Crack Recognition Algorithm Based on Machine Vision
SUI Pengfei, TENG Teng, FAN Chunling
(College of Automation and Electronic Engineering, Qingdao University of Science and Technology, Qingdao 266061, China）
Abstract: Due to insufficient blasting in mines, it is necessary to identify and knock the remaining rocks. Manual processing is time-consuming and laborious, and there are hidden dangers to personal safety. Therefore, this paper proposes a mine rock crack recognition algorithm based on machine vision to identify rock cracks in mines to facilitate subsequent crack knocking. First, the particle swarm algorithm is used to threshold the mine rock fracture image, and then the Zhang algorithm is improved by the template matching method to form a single smooth pixel skeleton to find edge breakpoints. Combining the fact that the part of the fractured part occupies a small proportion of the entire fracture, the morphological operation can have an impact on the fractured point, and the continuity of the connection and reduction cracks can be achieved by multiple growth iterations using the growth structure element at the fractured point. In this paper, samples are collected from the mine field, and two representative types of fracture image samples are selected, one is regular fracture (linear fracture), the other is irregular fracture (reticular fracture). Finally, through experimental analysis, the recognition rate of regular cracks is 93%, and the recognition rate of irregular cracks is 79%. Although the recognition of irregular and small cracks(width less than 1 mm)is difficult, because the risk factor of such cracks is small, it can not be processed, so the algorithm proposed in the article Able to meet the needs of engineering applications.
Key words: crack identification; threshold segmentation; skeleton extraction; template matching; growth operation
收稿日期： 2021-12-04
基金项目： 山东省自然科学基金项目（ZR2019MEE071）.
作者简介： 隋鹏飞(1996—)，男，硕士研究生.*通信联系人.
