Research On Three-dimensional Measurement Based On Structured Light Fusion Binocular Stereo Visio

With the rapid development of technology and the improvement of the automation industry,optical three-measurement and reconstruction technology has been widely applied in fields such as process manufacturing,precision inspection,space exploration,and biomedical engineering,which puts higher demands on fast and accurate three-measurement and reconstruction in the future.Optical three-measurement technology and binocular stereo vision technology are important branches of computer vision,and they have similar measurement applications in different scenarios.However,the accuracy of structured light three-measurement is usually limited by the level of the projector technology and calibration accuracy,while binocular stereo vision is sensitive to objects with discontinuous disparities or weak textures.Therefore,improving the three-measurement accuracy of structured light and binocular stereo vision is of great research significance.This paper optimizes the structured light three-measurement technology,integrates binocular stereo vision technology,eliminates the need for complex projector calibration,and uses absolute phase values as the regional matching elements,achieving efficient and high-precision three-measurement.The main work of this article includes:(1)building a three-dimensional measurement hardware system that combines structured light and binocular stereo vision,calibrating the binocular cameras with a checkerboard pattern,with a calibration error of approximately 0.04 pixels,and designing a corresponding software algorithm framework in a C++ development environment under the Windows operating system;(2)by using Bayer dithering twodimensional area modulation technology,encoding the structured light pattern and maintaining the good sinusoidal frequency spectrum characteristics of the encoded pattern.Even slight defocusing can obtain high-quality sinusoidal fringe patterns.The wrapped phase information of the measured object is obtained using the three-step phase-shifting method,and the phase is unwrapped to absolute phase using the multi-frequency heterodyne method;(3)An improvement is made to traditional region-based stereo matching algorithms by replacing the matching primitive based on regional grayscale values with region absolute phase values.This improvement can address the issues of repeated or weak textures while avoiding the problem of selecting matching region sizes.A search method that combines global and local search is employed to quickly locate corresponding points.Compared with the disparity constraint method,the search time is reduced by 70.4%.Finally,using a high-precision matte ball for measurement,the RMSE of the proposed method is 0.059 millimeters.