摘要
针对目前轴承钢球表面缺陷提取方法的不足,设计了一种通过图像来提取钢球产品表面缺陷的算法。该算法首先利用分段线性灰度算法对钢球表面微小缺陷进行增强,再结合最大熵来实现对钢球表面缺陷的自动分割,最后采用投影原理和二维联合统计算法,完成对缺陷的快速提取和区域归类。实验表明本文算法对钢球表面五类缺陷的提取可以达到很好的效果,在basler工业相机,900×560分辨率的条件下,算法耗时小于30ms,能够满足钢球表面缺陷检测的实时性要求。
Aiming at the analysing of the characteristics of steel ball surface flaw, an extraction algorithm,ex-tracting and classifying the surface flaw by the projection principle and two-dimension-combined statistical algo-rithm, improving the slight flaw analysis effect by a piecewise linear gray level algorithm and achieving automatic segmentation by the largest maximum entropy, is designed in this paper to discover the steel ball surface flaw by vi-sion. According to the experiment results, five kinds of steel ball surface flaw can be detected effectively and the al-gorithm computation time consuming lasts less than 30ms by a basler industrial camera in the 900 ＆#215; 560 resolution . The algorithm can meet the requirement for real-time in industrial vision detection.
出处
《激光杂志》
CAS
CSCD
北大核心
2014年第9期58-61,共4页
Laser Journal
基金
国家自然科学基金(No.61075007)
关键词
缺陷检测
机器视觉
图像分割
区域归类
Defect detection
Machine vision
Image segmentation
Region clustering
