PCB走线检测的预处理算法

Preprocess algorithm of PCB line detection

介绍了一种PCB待测图像预处理算法,它包括对待测图像的去噪和二值化分割两个部分。该方法使用Harr小波对待测图像的灰度直方图进行平滑处理来消除其中的“伪谷值点”,再充分利用标准图像的先验信息来确定一个二值化分割阈值的大致范围,最后在这个范围内找到直方图的谷值,这即是所需的阈值。实验表明,该预处理算法能够得到一个满意的图像二值化阈值,处理后的待测图像目标边界清晰,很好地保留了所有走线缺陷,同时消除了大部分可能影响后续匹配的噪声,处理时间也较短,可以达到实时检测的要求。

A preprocess algorithm consisting of denoising and binarization for PCB test image is proposed. The algorithm smooths the gray histogram of the test image with Harr wavelet to remove the ＂pseudo valley＂, and fixes a range of the threshold according to the prior information in the standard image, and then finds out the valley of the histogram in the range and that is the threshold. The experimental results show that the preprocess algorithm is able to obtain a satisfactory threshold for the binarization of the test image. After the preprocess, the objects in the binarized test image have clear boundaries and all the defects are preserved, while the noise which could influence on subsequent processes has been almost removed. In addition, the process speed has met the requirement of the real-time inspection.