Dual-camera-based Active3D Measurement

The measurement which expresses the geometrical shape, position, gesture, movement andother information within the scope of the space is called the3D measurement. The computerhas the ability to obtain3D information of objects by using the3D measurement. Theseinformation are widely used, including design and analysis based on physical objects,detection and monitoring the production process, surveying and mapping of civic architectureand terrains, deformation monitoring, design of large-scale construction and piping,manufacture of airplane and ship, construction of road and bridge, electronic display ofvaluables, preservation of cultural relics, plastic surgery and virtual reality, etc.Combination binocular vision technology with structured light technology can calculate thedata of point clouds in this paper. By researching the problems of image pre-processing,stereo matching of images and the processing of point clouds, the system will establish a goodfoundation of obtain high precision point clouds.Background information is removed of each image as well as eliminating noise of all imagesbased on wavelet transform in the process of measuring point clouds. From the effect ofcalculate point clouds in the following articles can be seen that these image preprocessingoperations effectively reduce the error of point clouds which comes from image problems.An algorithm of mutual and self-adapting method in sub-pixel level is proposed for imagematching. Based on Gray code images, the algorithm calculates the matching point locationby use of the comparable condition that has the same number of Gray code value in each row.The experimental results are provided that this algorithm matching effect works well andimproves the matching accuracy. Point cloud registration is one of the important steps to make up point cloud informationmissing. In this paper, we use the ICP algorithm combined with the marked-point to splicepoint clouds after denoising and simplifying them. The registration method transform two setsof coordinates of point clouds at marked-point first, and then use the ICP algorithm iterativecomputation of two sets point clouds with the nearest point,. Experimental results are verifiedthat this method can improve the speed of two sets of point cloud registration and has goodeffect.