Research on Grating 3D Reconstruction Method Based on Multi-view Shadow Segmentation
Single view structured light often encounters shadow problems caused by height when measuring complex objects,resulting in many noise points in the shadow area and affecting the measurement results.Therefore,this article proposes a raster 3D reconstruction method based on multi view shadow segmentation.By using the improved MobileViT as the backbone network and combining it with the APSA decoder module proposed in this paper to build a multi-scale self attention segmentation network,it is possible to accurately and quickly complete shadow area segmentation.The paper uses phase-shifting profilometry combined with multi frequency heterodyne method,combined with shadow mask,to obtain absolute phase maps from various perspectives.By us-ing Gaussian functions to process reprojection errors and obtain weights for each viewpoint,high-precision phase maps are fused and the complete point cloud information is finally calculated through system calibration parameters.The experimental results show that using this method for measurement achieves an absolute accuracy of 0.04 millimeters,effectively solving the problem of occlu-sion shadows while maintaining a high level of accuracy and reconstruction quality.
万方数据
维普
