针对印刷电路板(PCB)的CT图像存在灰度不均匀、导线形状多变等特点导致的导线难以有效检测的问题, 提出了一种基于超像素分割的PCB导线自动检测方法。该方法使用基于引导滤波的类顶帽变换对图像预处理, 提高不同类别区域的类间差异, 改善后续的超像素分割结果; 然后选择graph-based超像素分割算法对导线定位; 最后, 采用导线几何形状、灰度分布等特征判断识别导线区域, 实现导线检测。对存在灰度不均匀、多条导线、多尺度的PCB CT图像进行了实际实验。结果显示: 该算法取得了较好的导线检测结果, 在实验测试图像上检测率达到了90%以上, 基本满足导线自动检测对精度和抗干扰能力的要求, 具有较高的应用价值。

The CT image of Printed Circuit Board (PCB) exists problems in grey inhomogeneity, changeable and irregular wire shapes, so it is difficult to be detected efficiently. This paper proposes an automatic PCB wire detecting method based on superpixel segmentation. The comparably top-hat transform based on a guided filtering was used to preprocess images and to improve the interclass difference of different regions and the subsequent superpixel segmentation results. Then, the graph-based segmentation algorithm was selected to achieve the wire positioning. Finally, the wire region was identified by using the geometry and grayscale distribution features of the wire to implement the wire detection. The experiments for the PCB CT images with inhomogeneity grey, multi-wire and multi-scales were performed. The results show that the algorithm is able to overcome the intensity inhomogeneity of PCB CT image and achieves a better result with a detection rate more than 90% . It concludes that the algorithm satisfies higher precision and strong anti-jamming requirements for automatic detection of the wires of PCBs and has high application values.