Multi-resolution Dynamic 3-D Imaging Theory And Experimentation

The 3-D sensing of object surface is one of the core techniques in the 3-D digital imaging and modeling. We are now carrying out a research project supported by National Natural Science Foundation of China (NSFC) under the contract No. 60275012. The project is entitled "Multi-resolution Dynamic 3-D Imaging and Modeling". Several important issues for acquiring high quality range data will be systematically investigated and discussed in this thesis. The major works of this thesis include the following contents:We start with studying the technique proposed by Huntley and saldner (H-S algorithm), which is so called 'temporal phase unwrapping. Then, we propose an enhanced scheme for implementing the temporal unwrapping with the aid of wrapped maps in order to increase the efficiency of the H-S algorithm. We also present a clear geometric picture and physics insight of such an enhanced scheme. In addition, we evaluate the computational complexity of H-S algorithm and that of our approach to demonstrate the efficiency of the scheme proposed in this thesis. The initial condition for recursive scheme proposed in this thesis is theoretically deduced, leading to a deduction of the numbers of frames of encoded fringe patterns. With derived initial condition as a starting point, we are able to further increase the possibility of real time phase unwrapping. A detail analysis is given to prove that the enhance scheme for temporal unwrapping can be realized in a way of multi-resolution and level-of-detail, which would be a new feature for being able to achieve variable sensitivity. Based on the enhanced unwrapping algorithm proposed in this thesis, we build up a corresponding experimental setup for 3-D digital imaging. Two typical kinds of object surfaces, namely, a free form surface and a step-like surface are tested respectively to show the effectiveness of our approach.Alternative 3-D imaging method based on spatially binary coding is also studied. The technique of encoding and decoding with spatially 2-D patterns are investigated carefully. The necessary condition of two-dimensional patterns for binary encoding and decoding is discussed in detail. Again, we build up a corresponding experiment configuration for 3-D imaging using spatially binary code. Experiment results show good agreement with theoretical analysis. It is proved that this method can be applied for the 3-D acquisition for object surface with complex geometry and that of with deep surface slope. In this thesis, we also investigate the key issues involved in stereo vision in conjunction with spatial phase mapping. Several important aspects with respect to stereo vision are carefully studied in both theoretical analysis and experimental demonstration with free form surface as testing target. We also touch upon the technique of camera calibration in this thesis. This includes the modeling of the camera and the calibration techniques for off-the-shelf camera. We focus our attention on the technique of planar calibration and analyze its implementation step in detail. Finally, we give out the experiment of camera calibration for stereo vision.