It’s difficult to quantifying and unifying the grain quality and test standard in explosives grain appearance detection, put forward a machine vision detection method for optimal design of explosive grain appearance detection. According to the principle of artificial vision pattern detection, the system model is established by image recognition technology, the length is detected by image detection, the appearance defect is detected by high-speed camera scanning, and the physical verification is carried out by building a drug grain image detection model. The experimental results show that this method is feasible and effective in detecting the appearance of explosive grain and has high practical value.