Abstract: his article proposed a mine-used electric locomotive pedestrian safety warning system based on image processing. The system used TMS320DM642 as the core to complete hardware design which combined with video capture and other related peripheral circuits. It adopted the pedestrian recognition algorithm of gradient orientation histogram (HOG) as well as support vector machine (SVM), and realized electric locomotive pedestrian warning on the forward track in OPENCV platform. The experimental results showed that this system had the high accuracy of pedestrian recognition and could prevent locomotive accident injuries effectively.