Research On 3D Measurement Based On Analytic Solution And Image Processing Method

With the rapid development of artificial intelligence and robotics,real-time three-dimensional measurement techniques are becoming more and more important.The 3D robot vision is actually dependent on the development of real-time three dimensional sensing or measurement methods.Real-time and high-accuracy 3D shape measurement is required or highly desired by both scientists and engineers in so many fields such as medical surgery,biomedical engineering,intelligent manufacturing and entertainment industry.Based on the used devices and working principles,real-time three-dimensional measurement techniques could be classified into three categories: structured light(SL)technique,stereo vision(SV)technique and time of flight(TOF)technique.Up to date,all these techniques have achieved great success in both academics and commerce.Aiming at the technical defects of stereo vision method and structured light scanning method,in this article,the stereo vision method and the structured light scanning method are combined to obtain their respective advantages,and a structured light active stereo vision based on dual camera 3D reconstruction method are proposed.The effectiveness of the method is verified by multi-group and multi-scene experiments,and the dynamic target is reconstructed with low computational complexity.Based on structured light and active stereo vision,the main points are as flow:(1)A new structured light lattice is proposed,which is composed of red,green and blue spots with equal spacing.The center of the dot matrix is a red marked point,and the positions of each color point are restricted to each other.The proposed structure not only improves the robustness of spot matching,but also can use the matching relation of any two colors to fit the matching relation of the third color point.Using the above position characteristics,the optimal and stable matching relationship is established through iterative fitting;(2)HSV domain color image and slope difference automatic threshold segmentation(SDD)extraction method is adopted to solve the problem of RGB structured light extraction in multicolor object surface;(3)Aiming at the computational complexity and accuracy of binocular vision matching,a binocular matching strategy based on human eyes is proposed.By using this strategy,the target object contour,different color blocks(the center point),color blocks(the center point corresponds to the lattice)are matched one by one,and the matching relation of binocular structure spots is established.The binocular matching strategy reduces the computational complexity and improves the precision of matching;(4)Aiming at the problem of missing or mis-extraction of structural light caused by partial reflection of the target object or other reasons,a model restoration method based on the dot matrix of basic structural light is proposed.This method mainly realizes the effective correction of missing structured light points through probability calculation and position fitting recovery,and establishes a complete structured light lattice based on the basic structured light lattice model,which improves the matching precision and system stability;Through a series of qualitative and quantitative comparisons with the reconstruction results of the current mainstream 3D reconstruction equipment,the advancement and robustness of the proposed method are highlighted,and the value of the proposed 3D reconstruction method is fully demonstrated.