It is difficult for robot to move in wild environment because of pit areas, so a visual coping method was put forward to detect those pit areas. Firstly, according to project requirements, a part of suspected areas with small size were removed, as well as some the suspected areas with edge gradient. Secondly, the oval similarity was calculated to determine gray level segmentation threshold, and the similarity threshold was confirmed by analyzing the oval similarity curve, which was used to separate pit areas from the suspected pit areas. At last, the simulation results on 200 pictures with different angles, scenes and pit umbers show that the proposed method can be applied to extract pit area in complex environment, and is also not sensitive to outline regularity of pit area; besides, it can adapt to complex environment.