ABSTRACT Drilling robot of rockburst prevention is a key equipment for underground rock burst relief in coal mines, and drill pipe positioning is the basis and premise for realizing unmanned pressure relief operation. Based on the analysis of the characteristics and defects of the current positioning methods, a drill pipe positioning method based on improved YOLOv8 is proposed in this paper. Firstly, a drill pipe image data set simulated coal mine working state is collected and established. A fusion of deformable convolution and CBAM attention mechanism is proposed to enhance the image feature extraction capability. Moreover, the rotation decoupling head (RDH) and DP‐YOLOv8 network structure are designed to predict the angle information of drill pipes with large aspect ratios. Finally, pixel‐wise alignment of depth and color images is performed, and three‐dimensional coordinates of the drill pipe are obtained through coordinate system transformation. Experimental results show that the proposed drill pipe positioning method achieves precision, recall, F 1‐score, and mAP50 of 96.19%, 96.47%, 96.33%, and 96.24%, respectively. The absolute error for drill pipe positioning is 0.015 m, with an average error of 0.009 m. The maximum angle error is 0.4°, with an average error of 0.225°.