排爆机器人双目立体视觉定位技术的研究

发布时间：2019-04-01 23:09

【摘要】：排爆机器人可以有效降低作业人员的风险,其排爆功能的实现很大程度上取决于精确的定位,双目立体视觉是计算机视觉定位方法中最重要的距离感知技术。因此,基于双目立体视觉的排爆机器人精确定位技术的研究具有重要的实际意义。本文首先分析了排爆机器人的作业工况及测量要求,确立了分别从排爆机器人的静态和动态场景进行定位测量的研究思路。根据双目立体视觉的定位测量原理和测量模型,推导出空间点的三维坐标公式,并完成了双目相机的标定工作。其次对图像灰度化和平滑处理,去除图像中冗余的信息和干扰噪声,为后续的匹配工作做好准备。在静态图像的定位测量过程中,采用ORB检测算法结合Flann匹配算法进行立体匹配,并使用PROSAC算法去除误匹配点,得到精准匹配的特征点进行三维坐标与距离的计算。在动态图像的定位测量过程中,采用CamShift跟踪算法,将其与Kalman预测器结合解决跟踪过程中颜色干扰的问题,引入双目立体视觉的极线约束,得到精确的匹配点进行三维重建,计算出该点的空间坐标与距离值。最后分别进行了双目立体相机的静态和动态场景的定位实验,结果表明达到了排爆机器人现场定位与测距的要求。
[Abstract]:Binocular stereo vision is the most important distance sensing technology in computer vision positioning method, which can effectively reduce the risk of operators, and the realization of its explosive removal function depends on the precise positioning to a large extent, and binocular stereoscopic vision is one of the most important distance sensing techniques in computer vision positioning methods. Therefore, the research of precise positioning technology based on binocular stereo vision is of great practical significance. In this paper, the working conditions and measurement requirements of the robot are analyzed firstly, and the research ideas of positioning measurement from the static and dynamic scenes of the robot are established. According to the positioning principle and measurement model of binocular stereo vision, the three-dimensional coordinate formula of space point is derived, and the calibration of binocular camera is completed. Secondly, image grayscale and smoothing are processed to remove redundant information and interference noise in the image, so as to prepare for the subsequent matching work. In the process of static image location and measurement, the ORB detection algorithm combined with the Flann matching algorithm is used for stereo matching, and the PROSAC algorithm is used to remove the mismatched points, and the accurate matching feature points are obtained to calculate the three-dimensional coordinates and distances. In the process of dynamic image location and measurement, the CamShift tracking algorithm is used, which is combined with the Kalman predictor to solve the problem of color interference in the tracking process, and the polar constraint of binocular stereo vision is introduced to obtain accurate matching points for 3D reconstruction. The spatial coordinates and distance values of the point are calculated. Finally, the static and dynamic scene positioning experiments of binocular stereo camera are carried out, and the results show that the localization and ranging requirements of the robot are met.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】