Generating selfie images on the surface of a celestial body poses several challenges, including the position of the robotic arm, camera field of view, and limited shooting time. To address these challenges, the PCMIS (3D Point Cloud Matching Based Image Stitching) algorithm is designed, along with a corresponding shooting plan. This algorithm establishes a correspondence between depth and color information, enabling the generation of stitching views under any given view parameter. Furthermore, the algorithm is accelerated using GPU processing, resulting in a significant reduction in stitching time. The algorithm is successfully applied to generate selfie images for the Chang’e-5 mission.