Image orientation rectification method is an important part of PCB defect detection method based on machine vision, which can effectively improve the automation of defect detection system. Aiming at the low accuracy of traditional positioning rectification methods, a PCB image orientation rectification method based on multi-directional improved Sobel operator was proposed. Firstly, multi-directional improved Sobel operator was adopted to extract the G channel edge information of PCB image. Then, the edge image was binarized and the image contour was extracted through the median filter to realize the orientation of PCB region. Finally, the whole image was corrected by affine transformation based on the region of interest. The experimental results showed that the multi-directional improved Sobel operator proposed had better edge detection effect. Using the proposed evaluation function for image orientation rectification proposed, it was found that the PCB image orientation rectification algorithm was superior to the existing methods, and the accuracy of image orientation rectification improved effectively.