Abstract: The positioning accuracy of the existing underground positioning method fluctuates greatly and is difficult to be further improved. In order to solve the problem, a target identification and precise positioning method based on underground moving image collection is proposed. The environmental images are collected by using the camera carried by the positioning target, and the collected raw images are pre-processed by the adaptive histogram equalization method. The deep learning technology SSD algorithm and data enhancement SSD algorithm are used to identify the underground mark target, and the monocular distance measuring method based on the pinhole imaging principle is applied for ranging and positioning. The experimental results show that compared with two traditional algorithms of gray image matching algorithm and characteristic image matching algorithm, the SSD algorithm has better adaptability to distance and angle changes, and the effective detection rate still reaches 89.2% at 4.5 m. The data enhancement SSD algorithm improves the robustness and the detection accuracy rate is 1.7% higher than that of the SSD algorithm. The algorithm can better adapt to the complex environment. The results of underground application show that the target identification and precise positioning method based on underground moving image collection can achieve satisfactory results in the range of 2-10 m. The measurement accuracy decreases as the distance increases.