摘要
圆周表面缺陷检测是圆柱形金属工件质量自动化检测的必要环节,针对圆柱面划伤、凹坑和凸起等凹凸性缺陷难以检测的问题,论文提出了一种基于改进光度立体的金属圆柱面缺陷检测方法。首先,针对线扫相机难以在相同位置处采集多角度光图像的问题,提出了圆柱面多角度光图像采集方法,利用触发信号协调相机采集与光源有序变化,拆分采集数据以获得不同照明情况的多张图像;其次,针对金属表面非朗伯反射特性引起的光源标定误差问题,提出了基于导向滤波和Retinex的镜面反射分量抑制方法,极大地提高了表面法向量的计算精度。同时,充分利用光源照射方向的先验知识加快光度立体的计算速度。最后,使用物体形貌信息对缺陷区域进行提取与检测。试验结果表明,论文优化方法能够有效提高光度立体计算的实时性和准确性,增强了圆柱面划伤、凹坑和凸起等凹凸性缺陷的检测能力。
Automatic surface defects detection of metal cylinder is urgently demanded in industrial production.In order to solve the problem that it is difficult to detect uneven defects such as scratches,pits and bumps on cylinder surface,a surface de⁃fects detection method of metal cylinder surface based on improved photometric stereo is proposed in this paper.Firstly,aiming at the problem that line scan camera is difficult to acquire images illuminated in multi-angle at the same position.A cylindrical multi-angle illuminating image acquisition method is proposed.Trigger signals are used to coordinate the camera acquisition with the orderly change of illuminant.Different lighting conditions are separated from line scan data to form multiple images.Secondly,in view of the calibration error of light source caused by non-Lambert reflection of metal surface,a method based on bilateral filter⁃ing and Retinex is proposed to suppress specular reflection component,which greatly improves the calculation accuracy of the sur⁃face normal vector.Meanwhile,the prior knowledge of illuminant irradiation direction is fully used to speed up calculation.Finally,defect areas are extracted and detected based on the object topography information.The experiments show that optimized method in this paper can effectively improve the speed and accuracy of photometric stereo,as well as increase the capability of detecting scratches,pits and bumps defects on cylindrical surface.
作者
陈星寰
吴静静
CHEN Xinghuan;WU Jingjing(School of Mechanical Engineering,Jiangnan University,Wuxi 214122;Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment and Technology,Wuxi 214122)
出处
《计算机与数字工程》
2024年第7期2167-2175,2188,共10页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号:61873246,62072416)资助。
关键词
缺陷检测
圆柱面
光度立体
反射量估计
defect detection
cylinder surface
photometric stereo
reflection estimation
