基于机器视觉的电子元器件检测系统设计

Design of Electronic Component Detection System Based on Machine Vision

摘要:

为了提高宇航、军工电子产品检测的效率和准确性，设计基于机器视觉的电子元器件检测系统，通过研究有效的定位与检测算法，满足不同规格的印制电路板（Printed Circuit Board，PCB）元器件焊装正误检测。采用改进的Hough圆检测算法识别定位点，结合电装工艺文件定位元器件所在位置，绘制感兴趣区域（Region of Interest，ROI）。最后采用加速鲁棒特征（Speeded-Up Robust Features，SURF）算法进行特征提取，利用基于最近邻与次近邻比值的方法完成匹配，元器件特征点匹配准确率达到85%。并对元器件在光照和仿射变换下的匹配效果进行测试，结果表明，SURF算法在不同实验条件下，包括位移、角度以及光照变换等，匹配准确率达到90%。系统实现了宇航、军工PCB板上元器件安装正确性检测，具有较好的精确性和稳定性。

Abstract:

In order to improve the detection efficiency and accuracy of aerospace and military electronic products, an electronic component detection system based on machine vision is designed. By researching effective positioning and detection algorithms, the function of detecting the wedding accuracy of components with different specification and models on Printed Circuit Board (PCB) is realized. The improved Hough detection algorithm is employed to identify the position of marking point, which combined with process documents can be used to search the location of components. And according to dimensions, the single component image is segmented by drawing Region of Interest (ROI) boxes. Finally, Speed Up Robust Features (SURF) algorithm is employed to complete feature extraction and method based on the ratio of nearest neighbor to next nearest neighbor is employed to do image matching，the accuracy is 85%. Besides, the matching results are tested when illumination changes and affine transformation of electronic components occur. Results indicate that in different experimental conditions such as displacement transformation, angle transformation, scaling transformation, illumination transformation, and so on, the algorithm can match the template image in the target region, which has good matching effect, the accuracy is 90%. The correctness detection of components is realized with good accuracy and stability.