Research On Sub-pixel Edge And Center Detection Technology Of Coded Structured Light Stripe

Edge is the most basic characteristic of image. In image processing, edge is the most important feature which the image segmentation depended on, it is also the important information source of texture feature and the foundation of shape feature. Edge detection is the key technology of measurement by using image processing, the precision of edge detection is always one of the determinant factors to measuring system precision.In the process of 3D measurement based on coded structured light, if using stripe edge to decode, the precision of edge detection directly influences measuring system precision; if using stripe center to decode, the precision of center detection directly influences measuring system precision. So we compare the existent methods of edge and center detection, study methods of sub-pixel stripe edge and center detection which aim at 3D measurement technology based on structured light by combining gray code with line-shift. The precision of this method can reach sub-pixel, and the method can get less influence of reflectivity, slope, roughness and defocus.The main research contents are as follows:1. Study stripe sub-pixel edge detection method based on forward and inverse gray code pattern, project forward and inverse gray code pattern to measured object, then take the intersection point of two gray curve, this method can reduce the influence of stripe scatter to measuring precision.2. Analyses influence of reflectivity, slope, roughness and defocus to stripe edge, use edge detection method to detect edge of image which gained in those case.3. Analyses center detection method of laser stripe, then adopt a center detection method which aims at line-shift, this method employs barycenter method which based on edge detection, the precision of the method can reach sub-pixel.4. Compile edge and center detection program using Matlab, the experimental result is presented in this thesis.Project gray code pattern to plane, pyramid and sphere, then detect the stripe edge of intensity image which are gained by camera. The experimental result indicates that the edge detection error less than 0.3 pixel and the center detection error is 0.3 pixel.