Abstract: To enhance the automation level and safety of anchor drilling technology in underground coal mines, a method for automatic detection of steel strip rock bolt boreholes is proposed. The detection network utilizes the SSD algorithm, balancing detection accuracy and speed, and a lightweight MobileNetV2 backbone feature extraction network is used to establish the rock bolt borehole detection model. To address the issue of small rock bolt borehole targets in underground coal mines, the prior box is optimized to match the effective receptive field, improving recognition accuracy. The rock bolt borehole detection model achieves 94.24% accuracy and 94.08% AP for rock bolt borehole target identification in the self-built dataset, with a detection speed of 84.73FPS and a model size of only 14.3 MB. The model is optimized using TensorRT and deployed on the NVIDIA Jetson Xavier NX hardware platform to demonstrate its practicality.