Researches On 3D Reconstruction Technique Based On Binocular Structured Light

In recent years,the development of optical three-dimensional measurement technology with its unique non-contact,high-precision,rapid measurement advantages are gradually replacing the traditional measurement methods,widely used in product design,industrial testing,biomedical and virtual reality and other fields.Binocular stereo vision technology and structured light measurement technology are two kinds of techniques commonly used in optical three-dimensional measurement.The principle of binocular stereo vision is similar to the principle of biological vision system in nature,its implementation is simple and requires low measurement environment,but the method has high matching cost and poor accuracy in calculating the depth information of non-textured surface.The structure light measurement technology can increase the texture information of the measured object by the projection structure light,the measurement precision is high,the disadvantage is that the process of calibrating and correcting the non-linear error is more complicated.According to the advantages and disadvantages of binocular stereo vision technology and structured light technology,this paper proposes a binocular structured light three-dimensional imaging scheme combining the two technologies.Based on the two cameras,this program adds another projector for projecting structured light information.In the system proposed in this paper,the coded structure light controllably increases the texture of the surface of the measured object so that the feature points of the left and right images match accurately and quickly,and because the system is based on the binocular vision imaging principle,the calibration of the projector is avoided.The main work of this article includes the following aspects:1?The optical three-dimensional reconstruction method is studied,combined with the characteristics of binocular stereo vision technology and structure light measurement technology,the binocular structure optical three-dimensional imaging system is designed,the hardware platform of binocular structure light three-dimensional reconstruction system is built,and the software module with visual interface is developed.2?The structural optical coding technology is studied and a non-repeating De Bruijn color stripe coded pattern is generated.According to the principle of the color camera and the characteristics of the grayscale camera,a method of obtaining color images by a grayscale camera is adopted.For the problem of color distortion in the acquired color image,a color clustering method based on HSI space was used to solve the color stripe label.According to the window uniqueness of the De Bruijn sequence,the center points of the left and right image stripes are matched.3?According to the stripe gray distribution map of the acquired image,a sub-pixel stripe centerline extraction algorithm was proposed.The algorithm first uses the parabola-based mathematical model to extract the sub-pixel edge points of the fringe.According to the search domain established by the geometric centerline,the noise edge points can be filtered out,and then the sub-pixel fringe centerline can be extracted.Combined with the calibration parameters that have been obtained before,the three-dimensional coordinates of the feature points are calculated.Experiments show that the proposed three-dimensional reconstruction scheme can effectively improve the accuracy of object reconstruction.The disparity data are calculated with the existing matching information.4?The camera calibration and stereo correction technology are studied,and the calibration of binocular structure light three-dimensional reconstruction system is completed,and the calibration results show that the system is stable and the calibration error is small.Based on the disparity data,the three-dimensional coordinates of the feature point are calculated.Experiments show that the proposed three-dimensional reconstruction scheme can effectively improve the accuracy of object reconstruction.