In order to realize the automatic detection of the grating surface defect, a grating surface defect detection system based on machine vision is developed. First, the hardware platform of the system is built, so that key detection difficulties are analyzed and solved. Then, on the basis of the analysis and comparison of several traditional edge detection algorithms, the edge drawing detection algorithm is introduced to detect the grating surface image. Finally, according to the characteristics of the surface defects of various gratings, a number of threshold values are selected to define the surface defects of a variety of gratings and defects were extracted and labeled, and the detection algorithm is developed. Experimental results indicate that the detection accuracy of the system reaching micron level can detect the surface defects of many kinds of gratings in real time, with the detection efficiency up to 30 seconds, the recognition rate more than 94% and the mistake examining rate lower than 4%.