In the deep space exploration missions, the detector needs to land in complex terrain areas. Therefore, the rapid detection of on-orbit obstacles is very important, and image segmentation is one of the key processes of on-orbit detection. In view of this, a multi-level threshold image segmentation algorithm based on particle swarm and gray wolf hybrid optimization is proposed. In the optimization process, the proposed algorithm defines the threshold series for different scenes by changing the initial population conditions considering the image energy distribution. In the process of location update, the proposed algorithm increases the perturbation operators to expand the scope of global search, and introduces dynamic weights to balance the global search ability and local search ability of the group, thereby improving the speed and accuracy of optimization and completing image segmentation. The experimental results show that compared with the traditional swarm intelligence algorithm, the proposed algorithm shows better search ability, and it has obvious improvement in dealing with the problem of complex images where the gray histogram does not show bimodal peaks.