In the current application of 3D laser projection technology, the standard processing of projection coordinates is ignored and the full-angle laser scanning on the surface cannot be realized, resulting in poor fusion of laser point cloud data and large scene reconstruction error. Therefore, a 3D reconstruction model of interior scene based on 3D laser projection application optimization is constructed. A voxelized raster sampling algorithm is constructed based on point cloud surface eigenvalues and split cells to denoise laser point cloud data. Using the world coordinate system to normalize the 3D laser projection coordinates, the 3D laser projection model is established and the optimal projection rotation angle is obtained. The 3D model of indoor scene is reconstructed by using calibration plate and rotation angle stitching fusion point cloud data. The experimental results show that the proposed method has a good indoor scene reconstruction effect under the application of 3D laser projection, the error of laser point cloud data is small, the peak signal-to-noise ratio is high, and the performance of point cloud can be better fused.