A binary image connected domain labeling method for industrial product surface defect detection is provided, comprising: firstly, acquiring a gray level image of a product surface and performing denoising processing, and converting the gray level image into a binary image; then, performing first scanning on the binary image in a manner of scanning once for every three rows from left to right and from up to down, labeling three pixels each time, at the same time of labeling, combining equivalent connected domains and recording representation labels of the connected domains at any time; and finally, performing second scanning on the labeled image, and at the time of performing the second image scanning, duplicating each pixel in a corresponding connected domain by using a representation label of the corresponding connected domain. In a whole process, the number of repeatedly checking pixels is minimized, so that efficiency of a binary image connected domain labeling algorithm is improved; efficiency of detecting a product surface defect is improved, and a requirement of industrial product surface defect detection for a speed is satisfied. Therefore, the method has characteristics of being high in efficiency and strong in practicability.