结构光三维测量技术具有测量速度快、非接触式等特点，目前已被广泛应用于工业制造、科研教育、文物扫描、医疗卫生、数字娱乐等领域。基于条纹投影的相移轮廓术是结构光测量技术中最受欢迎的方法之一，因为其高精度、高点云密度的优势。系统标定是三维结构光系统中的关键技术，更好的标定结果意味着更高的测量精度。在本文中，提出了一种新的基于斑点检测的投影仪标定方法，以找到标定目标上的特征点相应的投影仪像素坐标，实现从相机图像到投影仪图像的亚像素级映射。该方法独立于相机标定的结果，因此不会受到相机二次误差的影响。实验结果也显示本文提出的投影仪标定方法的有效性，其重投影误差比现有的标定方法小约10%，并且该标定结果下的重建精度也更高。为了实现高精度、高鲁棒性的三维测量系统，本文对编码方案进行了设计，最终选择使用格雷码+相移码的编码方案，并对解码过程中可能出现的周期错位点提出了矫正方法。另外，本文选择将编码图案预畸变处理的方法，以便适用于投影仪的非线性模型，提高测量精度。同时该方法可以将投影图案数量减半，避免在解码端进行畸变矫正，降低计算复杂度，从而便于实时系统的设计。然而，基于条纹投影的相移轮廓术由于需要投影多帧图像常常被用于静态物体的三维测量。在动态场景中，例如胸腔的呼吸运动等，物体的运动会引入相移误差，从而导致测量误差。本文提出了一种基于四步相移轮廓术的新型补偿方法，以减少物体在匀速或匀加速运动时产生的运动误差。该方法利用条纹图案的相位相关性和周期性特征，仅从四个相移图案中估计相位误差，并实现逐像素的误差补偿。这种方法也可以应用于非刚性变形对象，并有助于恢复高质量的纹理。仿真和实验都表明，对于标准单目结构光系统，所提出的补偿方法可以有效地提高测量精度，并减少由运动引入的表面波纹。最后，本文对搭建的三维系统进行了实时化设计，一方面提出将格雷码和相移码交叉投影的方式来提高三维测量帧率；另一方面，提出基于CPU+GPU的异构并行加速模型，并将三维测量算法部署在GPU上进行加速。

Structured light three-dimensional measurement technology has the characteristics of fast measurement speed and non-contact, and has been widely used in industrial manufacturing, scientific research and education, cultural relic scanning, medical and health care, digital entertainment and other fields. Phase-shift profilometry based on fringe projection is one of the most popular methods in structured light measurement technology because of its advantages of high precision and high point cloud density.System calibration is a key technology in 3D structured light systems, and better calibration results mean higher measurement accuracy. In this paper, a new projector calibration method based on blob detection is proposed to find the corresponding projection pixel coordinates of the calibration target marker points, and achieve sub-pixel level mapping from camera image to projector image. This method is independent of camera calibration results, and therefore is not affected by camera secondary errors. Experimental results show that the proposed projector calibration method has a reprojection error about 10% smaller than existing calibration methods, and the measurement accuracy of the calibration result is also higher.In order to achieve a high-precision and robust 3D measurement system, this paper designs an encoding scheme and finally chooses the gray code combing with phase-shifting code encoding scheme, and proposes a correction method for possible periodic misalignment points during decoding. In addition, this paper chooses the method of pre-distorting the coding pattern, so as to be suitable for the nonlinear model of the projector and improve the measurement accuracy. At the same time, this method can halve the number of projection patterns, avoid distortion correction at the decoding end, reduce computational complexity, and facilitate the design of real-time systems.However, the phase-shifting profilometry based on fringe projection is often used for the 3D measurement of static objects because the object‘s motion, such as respiratory motion in the chest, introduces phase errors and measurement errors. This paper proposes a new compensation method based on four-step phase-shifting profilometry to reduce the motion errors generated by uniform or uniformly accelerated motion of the object. Using the phase correlation and periodicity of the stripe pattern, phase errors are estimated only from the four phase-shifted patterns, and pixel-by-pixel error compensation is achieved. This method can also be applied to non-rigid deformation objects and can help recover high-quality textures. Simulation and experimental results show that the proposed compensation method can effectively improve measurement accuracy and reduce surface ripple introduced by motion for standard single-camera structured light systems.However, phase-shifting profilometry based on fringe projection is often used for 3D measurement of static objects because of the need to project multiple frames of images. In dynamic scenarios, such as the breathing movement of the thorax, the motion of objects introduces phase shift errors and thus leads to measurement errors. In this paper, a novel compensation method based on four-step phase-shifting profilometry is proposed to reduce motion errors generated by objects moving with uniform or uniform acceleration. In this paper, the phase error is estimated from only four phase-shifted patterns by exploiting the periodicity of the fringe pattern, and a pixel-by-pixel error compensation is realized. This method can also be applied to non-rigid deforming objects and helps to recover high-quality textures. Both simulation and experiments show that for a standard monocular structured light system, the proposed compensation method can effectively improve the measurement accuracy and reduce the surface ripples introduced by motion.