Panorama is an important expression of virtual reality. Generally, the virtual panorama stitching process requires that the picture has obviously overlapping parts. Other locations need to have large differences in characteristics, otherwise, it will lead to obvious confusion stitching, this paper proposes a feature-matching image stitching method based on a virtual environment to solve this problem. At key frames of its stitching, the irregular color complex model is used to bind the scene to enhance the recognition of image features, and the results are greatly improved compared with the former.