Binocuiar Structured Light Vision System And Three Dimensional Reconstruction Method

With the development of computer technology,as an important branch of robot vision,binocular stereo vision has drawn the attention of scholars.Firstly,images from different angles should be taken by two calibrated industrial cameras.Moreover,images from this pair of cameras need to be registered by key points.Then,using mathematical corresponding relationship between three coordinate systems,which are camera coordinate system 、 space coordinate system and image coordinate system,the 3D information of the scene can be obtained.Binocular stereo vision technique not only has its own academic value,but also is widely applied in industrial inspection,quality assessment,automatic detection,medical diagnosis and city security,etc.In this dissertation,structural light is added based on the foundation of binocular stereo vision technique.The key points on the object surface are also increased.Depending on the features of high precision and high efficiency of structural light,the three-dimensional reconstruction with high precision is realized.In this paper,the main work and achievements are as follows:1.A high-resolution color vertical stripe light template is designed and proposed.This template is constructed by five colors.These colors are picked from the RGB color range and which have great difference in value.2.A detection method composed by gradient edge detection and wavelet edge detection is proposed to improve the edge detection quality.Since gradient is sensitive to color change,the gradient map is computed in the three-channels respectively.R,G and B.By using the maximum value strategy to merge the detected gradient edges,the comprehensive edge detection result can be obtained.Edge detection method with wavelet maximum value can generate local featuresand grasp abrupt value change.After extracting edges and applying gradient edge detection and wavelet edge detection to connect breakpoints and process pseudo-edge,the edge map with high precision can be obtained.3.A matching method based on periodic color vertical stripes in this paper is proposed.In view of the possibility of misplaced matching of cycles,As the color distortion of the white stripes is not severe after shooting,this paper utilizes white as a factor to divide the stripe cycle,with white as a factor to divide the stripe cycle.Using the proposed edge detection method center line of the stripe could be found.Through adaptive window matching ideas,the window is set In the midline of two adjacent stripes.In this way,window misalignment in different regions could be avoided,and then use two factors to improve the adaptive weight performance.First,we add the color space factor,and change the computing space from CIELab color space to RGB color space.Then,the image gradient is included in the matching primitive,utilizing the color sensitivity of structural light to the greatest extent.Finally,noise removal and parallax optimization are performed to obtain a more accurate parallax value the triangular method and geometric calibration of the camera internal and external parameters are used to obtain accurate three-dimensional coordinates.