基于机器视觉的非平整物体表面凸起异物检测方法

A detection method of saddle-backing body on uneven surface based on machine vision

摘要:

在换向器质量检测中,其端面槽口附近凸起异物属于严重质量问题,目前仍采用人工检测方式,存在检测速度慢和漏检率高,且换向器端面的非平整特性给检测带来挑战。针对此问题,运用机器视觉技术,提出一种基于光度立体视觉的换向器非平整面凸起异物检测的方法。该方法通过四方位打光方式,获得不同方位光照图,对光照图进行计算得到其反照率图,再进行高斯卷积处理,然后针对特定区域采用极坐标转换用于提取缺陷特征并对其进行凸起异物识别与检测。实验结果表明,该方法能够快速有效检测换向器非平整表面存在的凸起异物问题,检测精度达到99.8%,能够满足对换向器质量的在线检测需求。

Abstract:

In a process of commutator quality inspection, there are serious quality problems in a convex foreign body near an end notch. So far a manual detecting mode is used to solve this problem. But there are detection inefficiencies, high leakages and other issues. Then an uneven characteristic of a commutator end have more challenges to the detection problem. Aiming at this problem, a method of detecting the surface foreign body of non convex of commutator based on a photometric stereo vision by means of an machine vision technology is proposed in this paper. A different range of illumination image is obtained by a four-way lighting method. An albedo image is reckoned up the lighting image, and then a Gauss disconsolation is applied to treat the images. A defection features is extracted by a polar coordinate transformation used in a particular area, as well as a convex body recognition and detection are carried on. Experimental results have shown that the proposed method can detect the raised foreign bodies on the uneven surface of commutator rapidly and effectively, and the detection accuracy is up to 99.8%. This method can satisfy the requirements of on-line inspection of commutator quality.