To solve the problem of digital light processing (DLP) light curing 3D printer image distortion and aberration, a optical micro-projection distortion correction method based on moving least squares method (MLS) is proposed. This method identifies error regions based on pixel-level images captured by a plate camera, and selects control points, and calculates target moving points. By using these two sets of pixel coordinate points to determine the image error, the MLS algorithm, which combines area control and compute unified device architecture (CUDA) acceleration, maps the control point coordinates to the target moving point coordinates, producing an inverse aberration map. This inverse aberration map is then used to correcte the aberration. Experimental results show that this method can correct micro-projection distortion in DLP optical machine. The final correction error reaches 0.5 pixel in accuracy, which improves the projection quality. This advancement significantly benefits the light-curing 3D printing industry.