The traditional earth rock transportation business in mining areas usually takes loading, transportation and unloading as a process. In this process, the loading task is to load coal, iron ore and other mining materials; Transportation is to complete the transportation of mining materials in the designated road section; Unloading is the last step, and the mining products are released to the designated location. The unloading site in the mining area is often a dump surrounded by a retaining wall. The truck needs to be accurately parked at the front of the retaining wall before it can be unloaded outside the wall. Therefore, the accurate detection of the retaining wall is particularly important in the unloading process. In this paper, through the workflow of automatically updating the unloading point of the retaining wall based on the detection of the retaining wall, the unloading point is selected only at the first stop, and then the unloading point is automatically updated according to the detection results of the retaining wall, which greatly improves the unloading efficiency. In the detection process, the retaining wall point cloud filtering method based on the mining area retaining wall rules is used to filter the noise and uneven points in the retaining wall point cloud. Finally, the posture of the unloading stop of the retaining wall is obtained by the least square fitting method to guide the unmanned mine car to accurately dock to the completion of the unloading of the retaining wall.