基于成对几何直方图的PCB标志检测与匹配

矩方法和模板匹配方法在PCB标志检测中长期被使用,但是这两种方法存在低速和匹配率低下等问题。由于图像的形状特征对各种干扰具有很好的稳健性,本文提出一种利用Freeman链码描述图像轮廓,然后利用顶点检测算法简化图像的顶点以减少计算量,最后利用这一图像轮廓生成成对几何直方图,利用其旋转不变性对图像进行基于轮廓的检测与匹配的方法。实验结果表明利用成对几何直方图方法进行的PCB标志检测与匹配的综合正确检测率高达96%。

Detection and Match of PCB Mark Based on Pair-wise Geometrical Histogram

Aiming at the low speed and match rate of the moment method and template match method,which have been used in the detection of PCB marks for a long time,and for the robustness of the image shapes,this article proposes a method that uses Freeman chain code to extract the contours of images and vertex detection algorithm to simplify the vertexes of the contours.Finally the pair-wise geometrical histograms whose rotation invariance is applied in the detection and match of PCB mark are generated by these vertexes.The experimental results using this method indicate that the comprehensive correct match rate is up to 96%.