PCB显微图像缺陷圆孔检测方法研究

Study on the method of defect circle hole detection in PCB microscope image

分析了随机Hough变换(RHT)与基于最小二乘法圆拟合(CBL)的圆检测算法的基本原理,讨论了他们在圆检测算法中具体实现方法。针对RHT与CBL各自的优点,提出了RHT与CBL相结合的圆检测新方法:首先对PCB图像进行预处理,用RHT原理确定所求一定数量的初候选圆参量;然后由这些参量求出边缘图像中各点到初候选圆边界的距离,从而确定每个初候选圆附近的点集,剔除干扰点与噪声;最后利用最小二乘法(LSM)对各点集进行圆拟合得到各个点集较精确的圆参量,用简单模式聚类方法提取出最后真实的缺陷圆孔边缘。将该方法应用于包含模糊与噪声的PCB显微图像缺陷圆孔检测,尽可能真实地提取出了缺陷圆孔边缘。

The basic principles of circle detection of Randomized Hough Transform(RHT) and circle fitting based on least square method(CBL) were analyzed,and their specific methods in the algorithm of circle detection were discussed.In accordance with the advantages of RHT and CBL,a new method of circle detection was presented:At first,the preprocess of PCB image was executed,and the parameters of some original possible circle were confirmed by the principle of RHT.And then,the distance was computed from every point in the edge image to each original possible circle detected by these parameters.Based on the distance,the point sets in the vicinity of each original possible circle were extracted and the disturbance or noise far from the original possible circle was deleted.At last,the exact parameters of circle were obtained after fitting the points within each sets with LSM,the last real defect circle edge was extracted by simple method of pattern clustering.The real edge of defect circle hole of PCB microscope image was obtained as far as possible when we applied the method to detect defect circle holes which include fuzzy and noise.