Research On The Key Techniques Of Three-dimensional Foot Reconstruction From Multiple Perspectives

With the rapid development of the field of computer science,people's judgment and measurement of objects and the realization of three-dimensional reconstruction of object information have been replaced by machine vision to complete the human eye.This has also become a hot research direction,and scholars at home and abroad have now paid extensive attention to three-dimensional reconstruction of the real world for rapid and accurate digital modeling of this field.Since the foot is the second heart of the human body,the person's standing,walking,etc.must have feet to complete,and its deformity or abnormality will affect the supporting performance and athletic performance of the foot.When people are not standing still and are in motion,the footwear products have played a great role in effectively avoiding injuries to the feet.In the prior art,the common three-dimensional foot scanning technology is based on laser and structured light,and most of them are measured in an enclosed space,and the cost is high.In the background of the above research,this paper designs a set of multi-view foot acquisition system framework based on four cameras.Aiming at this system,the multi-camera joint calibration device and method are designed and developed first.The calibration method of this paper can be automatically run b y the program,which can realize the automatic calibration of the depth camera better,improve the calibration efficiency and reduce the cost.And solve the problem that the plane calibration can only be two-two calibration,the calibration process is complicated,it is not convenient for the user to operate and the image distortion caused by the angle of the adjacent camera optical axis is too large;secondly,based on the point cloud data,the use of particles Group algorithm and ICP(iterative closest point)method for reconstructing 3D foot shape:This paper uses particle swarm optimization algorithm to coarsely register the position,firstly reduce the positional deviation,and then use ICP algorithm to accurately register,so as to solve the ICP algorithm at two points.There is a problem of large position translation or rotation angle failure in the cloud.Finally,a multi-view three-dimensional foot scanning measurement system using four cameras is presented,which reduces the complexity and cost of three-dimensional foot measurement.The above algorithms are implemented in the Visual Studio 2015 set environment under WIN10 system,and the validity of this algorithm is verified by using the point cloud Database of Stanford University as experimental data.