【作者】 刘永久；

【导师】 双丰； 石井抱；

【作者基本信息】 中国科学技术大学 ， 信息获取与控制， 2014， 博士

【摘要】 结构光三维形态测量技术已广泛应用于工业检测、模式识别以及逆向工程等领域,显示出了广泛的应用前景。随着工业生产自动化水平的提高,现有的基于标准帧率的三维形态测量技术已不能满足快速运动物体连续三维形态测量的需要,高速、实时三维测量正在成为三维测量技术的发展趋势。针对目前结构光三维测量方法无法兼顾运动物体测量与测量效率难题,本文基于格雷码编码结构光方法,对高速视觉三维测量、同步误差补偿算法、异构并行运算以及移动三维测量等关键技术进行了研究。以期实现运动物体的高速实时三维测量,并将该技术扩展至工业检测领域。本文的主要研究内容及取得的成果如下：1.采用了一种基于高速视觉的投影和同步图像获取方法,将格雷码结构光编码方法应用于运动物体三维测量中,利用缩短帧间投影与图像获取时间方法减小同步误差。该方法可以突破了标准帧率的限制,实现运动物体的连续三维测量；2.基于物体运动信息估计方法,提出了一种“运动补偿算法”,利用物体本身的运动速度信息补偿同步误差。实验结果表明该算法可通过物体运动速度信息预测帧间像素位置,实现同步误差补偿,进而获得更加准确的三维形态测量结果；3.针对高速视觉中的大量图像数据处理问题,本文采用CPU+GPU异构并行计算模型,充分利用GPU的多核并行计算能力,加速三维形态测量算法。在当前实验中,实现了512×512像素分辨率下500帧/秒的运动物体实时三维形态测量；4.对移动三维形态测量技术进行了研究,开发了一种机器人搭载用的高集成度实时测量三维系统,利用机器人实时反馈的运动信息补偿同步误差,实验结果表明该方法能够克服基于质心跟踪方法中的估计盲区问题,进一步扩大了三维形态测量范围。论文在运动物体高速三维形态测量、同步误差补偿算法、异构并行运算以及移动三维形态测量方面取得了一定的研究成果。研究成果扩展了格雷码编码结构光三维形态测量适用范围。更多还原

【Abstract】 Structured light three-dimensional(3D) shape measurement technique has been widely used in industrial inspection, pattern recognition and reverse en-gineering and showed its perspective application. With the development of indus-trial production and automation, current3D shape measurement techniques based on standard frame rate can not meet the requirement for3D shape measurement of fast moving object. High speed, real-time and moving object measurement technology is becoming the development trend of3D shape measurement tech-nology. Currently,3D shape measurement methods based on structured light can not take into account3D shape measurement of moving object and measure-ment efficiency. Based on gray-coded structured light method, this thesis focused research on high-speed3D shape measurement, motion-compensated synchroniza-tion errors reduction algorithm, heterogeneous parallel computing algorithm and mobile3D shape measurement system. We aim to realize the real-time3D shape measurement of moving object and expand its application in industrial inspection.The major content and research results are summarized as follows:1. Gray-coded structured light method was used for3D shape measurement of moving object by introducing a high-speed projection and real-time images ac-quisition method. In the proposed method, synchronization errors are reduced by shortening the projection time. The proposed method can overcome the re-striction of standard frame rate and achieve continuous3D shape measurement of moving object.2. Based on motion information estimation methods of moving object, we proposed a "motion-compensated algorithm". In this algorithm, the moving velocity information of the object is used to compensate the synchronization er-rors. Experimental results showed that the proposed method can further reduce synchronization errors by estimating the pixel coordinates among captured frames and obtain more accurate3D shape measurement results.3. To solve the processing problem of massive image data, a heterogeneous parallel computing model "CPU+GPU" was introduced. Computing abilities of multiple cores on GPU are utilized to accelerate the3D shape measurement algo- rithm. Current experimental results showed that the proposed method can realize real-time3D shape measurement of moving object at500fps with a resolution of512x512pixels.4. On the basis of study on mobile3D shape measurement techniques, we developed a robot-mounted highly integrated real-time3D shape measurement system, which is capable of compensating the synchronization errors by using the feedback information of the robot. Experimental results showed that the devel-oped real-time3D shape measurement system can overcome the blind estimation problem in centroid-based tracking method, and further expand the3D shape measurement scope.This thesis has conducted work on3D shape measurement of moving object based on high-speed vision, synchronization error reduction algorithm, hetero-geneous parallel computing and mobile3D shape measurement system, which extends the application field of gray-coded structured light method.更多还原