为了适应印刷电路板(PCB)上IC类贴装器件检测的实时性要求,提出了一种基于边缘积分投影和颜色统计特征的检测算法。分析了不同贴装质量的IC类器件在三色(红、绿、蓝)结构光下的特征,用Sobel算子提取图像的水平和垂直边缘,并对边缘做积分投影,从而避免了设置边缘阈值的困难。在此基础上,利用最大邻域梯度法和滑动定位窗算法定位引脚的水平和垂直边界,并检测缺件、错件、偏移、歪斜等缺陷,最后,引入HSI(色度、饱和度和亮度)空间的颜色统计特征,完成对翘脚、极性错误的检测。实验结果表明:该算法可有效地检测IC类器件常见的贴装缺陷,整体准确率为97.9%,检测单个IC器件的时间约为30 ms,能够满足在线检测系统高准确率和实时性的要求。

In order to satisfy the on-line inspection requirements of the surface mounted Integrated Circuit(IC) devices on Printed Circuit Boards (PCB),an inspection algorithm based on the edge integral projection and color statistical features was presented. The features of the IC devices with different mounted qualities under a three-color (red,greed,blue) structure light source were analyzed. Then,the horizontal and vertical edges were extracted with the Sobel arithmetic operator and the edges were projected for avoiding the difficulty of right thresholding. Based on the integral projections of the edges,the horizontal and vertical borders of the IC leads were obtained by the maximum neighbor gradient algorithm and the sliding location window algorithm,respectively. After location the leads,the defects such as missing devices,wrong devices,shifts and skews were inspected. Finally,the statistical color features in HSI (hue,saturation,intensity) space were extracted to detect the defects of lift leads and wrong polarity. Experiment results show that the proposed method can identify the defects of the IC devices effectively,the total success rate is 97.9% and the time consuming for single IC device is about 30 ms. It has verified the validity of this algorithm in terms of the recognition rate and speed.