The binocular structured light 3D imaging technology can obtain the height distribution of the measured object by projecting structured light and then obtaining 3D information about the object. It can be widely used in industrial production, medical research, and other fields. However, due to the interference of unstable factors such as ambient light and optical lens errors, the widely used three-dimensional imaging technology based on Gray phase shift technology still needs fixing, such as decoding errors and inaccurate matching. This paper proposes an improved algorithm based on existing technology. The algorithm uses mutually exclusive complementary Gray codes for projection, uses a bidirectional averaging decoding method to unwrap the phase, and generates high-quality point cloud data through mask-constrained stereo matching. After experimental analysis and verification, the algorithm in this paper has improved its accuracy by 11.49% compared to traditional algorithms, and the average number of point sets has increased by 14.82%. This proves that the algorithm can effectively suppress optical noise interference and improve the accuracy of 3D imaging.