Three-Dimensional Vision Measurement Of Complex Shapes

Structured light vision measurement system uses a projector instead of a camera in binocular stereo vision system. The initiative light pattern is projected by a projector, the pattern is modulated by the scene, and captured by a camera. The captured pattern is decoded by some algorithms. Matching the decoded pattern with the projected pattern, each corresponding point is identified. Finally, the three-dimensional reconstruction is finished by using the triangulation principles.Structured light vision measurement system can effectively solve the difficult matching problems in binocular stereo vision measurement. Due to the projection is an active way. The features of projected pattern is already known. As long as the images pattern captured can be obtained by decoding. The corresponding relation and matching points between the two images can be accurately found using the features amount encoded.Structured light vision measurement is considered one of the most reliable three-dimensional measurement technology. This dissertation uses the time-coding method based on a sinusoidal grating stripe. Absolute phase of the sinusoidal grating stripe as the coding feature quantity. Since the absolute phase of the sinusoidal grating stripe is unique.The uniqueness can be served as the basis for the corresponding point. Three-step phase shifting method is used to project the grating stripe. And the grating stripe is projected in horizontal and vertical directions. The modulated image of the scene is captured by the camera. Then, wrapped phase map is obtained. Wrapped phase map is unwrapped by a phase unwrapping algorithm, finally, the absolute phase map is obtained. The corresponding points can be found by absolute phase of the two unwrapped phase maps.Due to the surface information of complex shapes are more complex and varied. Phase unwrapping errors appear easily. In this case, phase unwrapping is difficult. The accuracy and speed of phase unwrapping is directly related to the last measurement accuracy and speed. So this dissertation focuses on the phase unwrapping algorithms. The main contents of this dissertation is as follows:First, an overview of methods and principles used to measure the three-dimensional objects are introduced, and then the topics are extended to the phase measurement and phase unwrapping. The existing phase unwrapping algorithms are classified. Advantages and disadvantages of these algorithms are summed up to analyze the theoretical basis, the robustness and speed. This analysis paves the way for the new proposed algorithm.Secondly, quality-guided phase unwrapping algorithm is optimized. Quality-guided phase unwrapping is a relatively common and precision method among phase unwrapping algorithms,but it is very time consuming,especially dealing with a large image. In this dissertation priority queues, index linked lists,etc. are used to deal with the unwrapping order,which can effectively reduce the time complexity of the algorithm, and the processing time is reduced significantly.Again, we propose two new phase unwrapping algorithms. One is based on spiral-path following unwrapping method, and another is based on constructed weight-edges. The latter uses weight-edges to guide the path of phase unwrapping. Horizontal and vertical edges'weights are structured by using the quality value of adjacent pixels. Unwrapping process uses a two-point edge as a unit and follows a noncontinuous path. It is an effective solution to prevent the influences caused by the discontinuous areas, environmental noise block or lack of sampling.Finally, the uniqueness of the absolute phase is used to determine the location of the points in the projection pattern. The coordinate matrix of the matching points for three-dimensional reconstruction is obtained. Results of the experiments are compared to verify the above algorithm.