基于双目视觉的目标运动参数高速实时测量方法研究

发布时间：2018-05-17 10:26

本文选题：立体视觉 + 高速实时测量　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：目标运动参数测量作为国内外测试技术研究的前沿领域,其可为试验测试和技术分析提供重要的数据校对来源,在国防军工、航空航天以及交通运输等领域,发挥着不可替代的作用。视觉测量作为应用最为广泛的非接触测量方法之一,具有良好的适应性和较高的测量精度,但是大多数的目标运动参数视觉测量系统都采用离线数据处理,不具备在线实时提供目标姿态和位置等运动参数的手段,只能在视觉传感器采集信息后进行数据处理。这种方式只能作为辅助验证手段,参考价值有限,无法满足实时测量的需要。因此,本课题对基于双目视觉的目标运动参数高速实时测量方法进行了研究,设计一套基于双目视觉的目标运动参数高速实时测量系统。本课题的主要研究内容如下:首先,针对双目视觉测量原理问题进行方法研究,详细阐述单目和双目摄像机成像模型和畸变模型,并根据双目摄像机成像原理和立体视觉标定方法,提出了基于双目摄像机标定重构误差的多步迭代误差补偿方法,提高双目摄像机标定精度,并根据本课题的双目摄像机系统的实际需要,提出适用于本课题的目标点三维重构解算方法。其次,针对目标运动参数高速实时测量问题进行方法研究,提出基于特征面的目标运动参数测量方法,利用三个独立非共线的配合目标构建空间特征面,根据特征面在左摄像机坐标系下的姿态和位置变化解算得到目标运动参数,并针对目标定位提取与跟踪匹配问题提出可行的解决方案。进一步结合实验室硬件完成系统设计,其中包括图像高速采集和处理系统以及双目摄像机同步外触发方案。最后,针对本课题的测量系统设计了双目摄像机立体视觉标定实验,完成了双目摄像机高精度立体视觉标定,验证本课题提出的多步迭代误差补偿方法的有效性;进一步设计并完成测量系统的性能评估实验,对前文中所提出的技术解决方案进行了实验验证,在测量系统精度检测实验中,利用运动模拟装置模拟目标在空间中二自由度的转动,分别将静态测量结果和动态测量结果与激光测距仪测量数据和运动模拟装置反馈数据进行比对和分析,实现基于双目视觉的目标运动参数高速实时测量。
[Abstract]:Target motion parameter measurement, as a frontier field of test technology research at home and abroad, can provide an important source of data proofreading for test and technical analysis. It can be used in the fields of national defense military industry, aviation and aerospace, transportation and so on. Play an irreplaceable role. As one of the most widely used non-contact measurement methods, visual measurement has good adaptability and high measurement accuracy. It does not have the means to provide real-time motion parameters such as target attitude and position, and can only be processed after the information is collected by the visual sensor. This method can only be used as an auxiliary verification method with limited reference value and can not meet the needs of real-time measurement. Therefore, this paper studies the method of high-speed real-time measurement of target motion parameters based on binocular vision, and designs a high-speed real-time measurement system of target motion parameters based on binocular vision. The main contents of this thesis are as follows: firstly, the method of binocular vision measurement is studied, and the imaging model and distortion model of monocular and binocular cameras are described in detail. According to the imaging principle of binocular camera and the method of stereo vision calibration, a multi-step iterative error compensation method based on calibration reconstruction error of binocular camera is proposed to improve the calibration accuracy of binocular camera. According to the practical needs of the binocular camera system, a method of 3D reconstruction of the target point is proposed. Secondly, aiming at the problem of high speed real-time measurement of target motion parameters, a method based on feature surface is proposed to measure target motion parameters, and three independent non-collinear cooperative targets are used to construct spatial feature surface. According to the change of attitude and position of the feature surface in the left camera coordinate system, the motion parameters of the target are calculated, and a feasible solution to the problem of target location extraction and tracking matching is proposed. Furthermore, the system is designed with laboratory hardware, including the high-speed image acquisition and processing system and the external trigger scheme of binocular camera synchronization. Finally, the binocular camera stereo vision calibration experiment is designed for the measurement system of this topic, and the high-precision stereo vision calibration of binocular camera is completed, which verifies the effectiveness of the multi-step iterative error compensation method proposed in this paper. Furthermore, the performance evaluation experiment of the measurement system is designed and completed, and the technical solution proposed in the previous paper is verified experimentally. The motion simulation device is used to simulate the rotation of the target in space. The static and dynamic measurement results are compared and analyzed with the laser rangefinder data and the motion simulation device feedback data, respectively. High speed and real-time measurement of target motion parameters based on binocular vision is realized.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】