3D Topography Measurement And Application Based On Convergent Structured Light Binocular

3D reconstruction techniques are important for obtaining the geometric shape of objects.Among these techniques,structured light-based 3D reconstruction has gained widespread attention and research interest due to its non-contact nature,high accuracy,and high speed.However,common parallel structured light stereo vision systems are limited in practical applications due to restrictions in the field of view,requiring visual sensors with larger field of view angles and longer object distances for measurement.Larger field of view angles result in greater lens distortion,while longer object distances decrease the image size of the captured object,both of which reduce the measurement accuracy of the system.Therefore,the study and application of convergent binocular vision play a crucial role in improving measurement accuracy.This thesis focuses on the research of 3D shape measurement technology based on a convergent binocular structured light stereo vision system,using the Gray code-assisted phase-shifting method as an example.The implementation of 3D shape measurement involves several steps: precise calibration of the system’s structural parameters,generation of projected fringe patterns based on structured light encoding rules,sequential capture of the fringe images of the measurement object using visual sensors,phase unwrapping,stereo rectification based on the system’s mathematical model,feature matching of corresponding points in the left and right images,and finally,reconstruction of a high-precision dense point cloud using the principles of triangulation to achieve 3D shape measurement of the captured object.The main content and achievements of the thesis are as follows.1.Improvement of calibration accuracy: The accuracy of calibration parameters plays a crucial role in 3D shape measurement.The Zhang calibration algorithm is used to calibrate the structural parameters of the binocular stereo vision system based on the characteristics of planar calibration objects.In this thesis,a calibration parameter optimization algorithm based on multidimensional mixed constraints is proposed,which achieves approximately 50% improvement in reconstruction accuracy compared to traditional calibration algorithms.2.Automatic optimal display for stereo rectification: This thesis proposes an automatic optimal display algorithm for stereo rectification with the objective of simultaneously reducing the display area and minimizing the introduction of redundant pixels.3.Based on the image features of convergent binocular structured light,a stereo matching algorithm is proposed that focuses on phase information and incorporates epipolar and phase constraints.The algorithm introduces curve fitting methods and fully utilizes the continuous characteristics of phase values in phase-measurement profilometry to achieve sub-pixel level stereo matching of particle pixels.4.The distribution of errors in the convergent binocular structured light 3D shape measurement system is studied,and an experimental setup is designed to verify the distribution of system measurement errors under fixed structure parameters.The experimental results are in good agreement with theoretical analysis.