A High-Resolution 3-D Shape Measurement System

In this thesis, we first introduce the basic applications of three-dimensional measurement systems in aeronautics, medical systems, entertainment, and robotics. Then, it introduces various methods used for three-dimensional shape measurement. Among those techniques, the structured light technique is widely used due to its high-speed and high-resolution capabilities. This research is based on a new structured light method—combined stereovision and phase shifting method. In this research, a new three-dimensional measurement system is set up with two high-resolution cameras and one LED projector. A sub-pixel stereo matching method is proposed to improve the system resolution. Factors that may influence system resolution are analyzed. A method is proposed to reduce the random noise in the system by averaging multiple sets of samples. After the system accuracy and resolution are evaluated, an error compensation method is also proposed to reduce system error. Experimental results are shown to demonstrate the method's effectiveness.The principle of the combined stereovision and phase shifting method is introduced along with its unique advantages as compared with conventional methods. The combined stereovision and phase shifting method uses two cameras and one projector. This method first calculates the horizontal and vertical phase maps based on the fringe images from the two cameras. It then uses these phase maps for stereo matching at the pixel level. The reconstruction of 3-D models is similar as that in a typical stereovision system. Since this method uses the phase values only in the process of stereo matching, phase measurement errors do not influence measurement accuracy directly. Therefore, this method can significantly reduce errors caused by phase measurement inaccuracy. In the mean time, it makes projector calibration unnecessary. The principle of nonlinear calibration of the measurement system is also introduced.The hardware system of this high-resolution three-dimensional measurement system is introduced, including the characteristics of the cameras and projector used in the system. The system uses two high-resolution cameras and one LED projector. Two cameras and one projector are fixed on a metal frame and the whole system is very compact. The control of the cameras and projector and the method to synchronize the cameras and projector are also discussed.We modify the original combined stereovision and phase shifting method and propose a sub-pixel stereo matching method. With both the left and right absolute phase maps available, pixels in the right image can be matched to pixels in the left image based on the vertical and horizontal phase values. However, because the pixels in the image are discrete and limited, it is almost impossible to find two pixels with exactly the same horizontal and vertical phase values. This leads to some digital errors at the pixel level. In order to solve this problem, a sub-pixel stereo matching method is proposed. We use a two-dimensional interpolation method to interpolate the pixels in the left image to find a point where the horizontal and vertical phase values are exactly the same as those at a pixel in the right image. This method can match the left and right images at the sub-pixel level, which greatly reduces the system error. In the research, we adopt two different interpolation methods: two-dimensional linear interpolation method and two-dimensional cubic Hermite interpolation method. Two-dimensional linear interpolation is simple and easy to use, but its first derivative is not continuous, which makes the result not smooth. Two- dimensional cubic Hermite interpolation method realizes the continuity of the first derivative and makes the result smooth. The experimental results show that the sub-pixel stereo matching method effectively enhances the system resolution.We also propose to reduce the system random noise by averaging multiple sets of images. By analyzing the experimental results, we choose the proper number of image sets that we should collect.We discuss about the methods to enhance system accuracy and resolution and provide the experimental results. We propose an error compensation method to reduce the system error. We construct an error map and do three-dimensional linear interpolation to estimate the errors at the measured points. Subtracting the estimated errors from the measured coordinates of these points, we obtain their three-dimensional coordinates with better accuracy. Experimental results show that this method effectively reduces the system error. Finally factors which may influence the system resolution and accuracy are analyzed.