Key Techniques For 3D Reconstruction Of Binocular Vision Based On Structured Light

The demand for obtaining 3D information of objects is growing,and 3D reconstruction technology is developing rapidly.Binocular vision and structured light are two representative directions in 3D reconstruction technology based on vision.Binocular vision has the advantages of simple and flexible structure and many suitable scenes,but it is too dependent on the texture features of the object surface.For weak texture and repeated structure areas,it is unable to carry out effective feature extraction and stereo matching.The structured light technology does not depend on the texture information of the object surface.Different structured light can be selected according to the actual application requirements.The measurement accuracy is high,but it is easy to be affected by the ambient light,and the calibration technology is complex.This paper compares the current mainstream three-dimensional vision measurement technology,and puts forward the combination of binocular vision and structured light to give full play to their respective advantages.The key technologies of binocular vision 3D reconstruction combined with randomly distributed blob structured light are studied as follows:Research on the detection algorithm of blobs.An improved algorithm of Lo G blob detection is proposed.By using Hessian matrix and FAST feature detection idea,the interference of stripe,edge and similar blob areas is effectively removed,and the correct extraction of structural light blob is realized.Research on subpixel localization of blobs.In the Lo G convolution image,the sub-pixel location of the blob center is realized by the method of quadric surface fitting.The accuracy of this method is higher than that of sub-pixel location in source image.Research on the blob matching technology between stereo image pairs.A sub-pixel matching method for the blobs between stereo image pairs is implemented.The first step of the method is to detect all structural light blobs in the image and locate the blob center of sub-pixel;the second step is to use the normalized variance matching algorithm with adaptive window size to realize the pixel level matching of blobs in the left and right images;finally,the sub-pixel data of the blob center obtained in the first step is used to obtain the sub-pixel level matching between the left and right image blobs,so that the disparity value is increased from pixel level to sub-pixel level.This method is simple and efficient,and the disparity value of sub-pixel can be obtained without complex global matching and optimization algorithm.