PCB Pad Detection Algorithm Based on Principal Component Analysis and Classification Regression Tree

Detection of a printed circuit board (PCB) pad is an important process to test the bottom circuit characteristics and plug-in soldering. Identification and accurate localization of circular and square pads in complex lighting environments are characterized by complexity, difficulty in extracting the main features, and determining the classification thresholds. In this context, a fast detection method based on principal component analysis (PCA) and a classification regression tree (CART) is proposed. Firstly, in order to accurately extract the connected domain and geometric features of the pad, the binary image is obtained by morphological preprocessing. Then, in order to screen out the main geometric features to increase the reliability and efficiency of the algorithm, PCA is used for the representation of the connected domain in the low-dimensional feature space. A CART model is trained to identify the connected domain. Finally, the coordinates of the target pad are extracted and transformed into the real coordinates by the hand-eye-calibration algorithm and transmitted to the mechanical probe system. Experimental results show that the method can accurately and quickly identify circular and square pads in a complex lighting environment. The average positioning accuracy is better than 1.5 pixels, and the frame rate can reach 12 fps, which ensures accuracy and real-time analysis of PCB detection, and effectively avoids the influence of random factors on the irregularly connected domain.