基于视觉信息处理的移动机器人旋转检测和目标跟踪研究

发布时间：2018-01-18 14:42

本文关键词：基于视觉信息处理的移动机器人旋转检测和目标跟踪研究　出处：《湖北工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着机器人科技的发展,现在机器人基本可以进行较高难度的动作,像随时调转方向、实时探测目标、驻地旋转等。但是,在地面崎岖不平或者沙质土壤的情况下,机器人极易出现检测器和传感器失敏的情况,这样就会无法准确的测算其旋转行为能力。对于这些现象,笔者将采用视觉信息处理方式来进一步探究机器人的旋转行为和追踪目标能力。在机器人上安装摄像头,通过摄像头采集相应的信息,再借用通信网络将采集的信息传送到机器人内部服务器;接着,行为人确定将要追踪的目标,将选定的数据传送给机器人的核心控制系统;核心控制系统对旋转角度以及目标距离进行快速计算,再将换算后的数据传送给及其热的运动控制系统,机器人便可以进行相应的行动。本篇论文主要探讨了两方面内容:第一个就是探讨机器人旋转的能力。首先在选用摄像头时,为了防止图片变形情况的出现,本文选用了黑白交错的方格对摄像头进行固定。然后,当机器人进行旋转时,会出现标准图和旋转图,通过明确的距离长度来确定两者图的固定点,再通过测量二者固定点之间的角度,便可以检测出机器人的旋转角度。本文探讨的第二个内容是研究机器人追踪目标的能力。全程测试中使用了TLD科学检测技术,一方面检测了机器人在目标物被掩盖以及目标物运动时追踪能力,最终得到了很好的结果;另一方面,控制机器人对路面上的随意车辆进行追踪,最终结果是该机器人能够对确定的目标表进行精确追踪,完成要求。通过多种实验最终获得了一些很好的结果:第一,在检测机器人的旋转角度时,全程采用了SIFT科学技术。通过检测发现,机器人无论是先进行平面移动再进行旋转还是先进行旋转再进行平面移动,所获得的旋转角度都是相同的,从而得到结论:平面移动和旋转行为的顺序并不会引起旋转角度的变化;通过对SIFT科学技术的进一步优化,使得机器人控制系统的运行速度增快了10倍;通过检测出的机器人旋转角度以及旋转时间等数据便于计算出旋转的角速度。第二,在检测机器人的追踪目标能力时,全程采用了TLD科学技术,并且通过区域划分、样本精简等方式来优化TLD科学技术,最终使得控制系统的运行速度增快了2倍左右,即按照原科学方法处理每一帧图像需要花费70Ms,但是采用优化后的科学技术仅花费30Ms。通过均衡图像精确度和花费时间,最终采用了320*240区域的追踪图像。
[Abstract]:With the development of robot technology, the robot now basically high difficulty movement, like any direction, real-time detection, resident rotation. However, in the rough terrain or sandy soil under the condition of the robot, prone to detector and sensor desensitization, so it will not accurately measure the rotation behavior ability. For these phenomena, the rotational behavior and target tracking ability the author will use the visual information processing way to further explore the robot. The installation of cameras on the robot, collecting relevant information through the camera, and then use communication network to transmit the information to the robot internal server acquisition; then the behavior of the people will determine to track the target, will the selected data is transmitted to the robot core control system; the core control system of the rotating angle and the target distance for fast calculation Then, the converted data is transmitted to the heat and motion control system, the robot can act accordingly. This thesis mainly discusses two aspects: the first is to explore the ability of robot. First, in the selection of the camera, in order to prevent the deformation of the picture, the black and White Checkered the camera was fixed. Then, when the robot is rotated, standard map and rotation graph, to determine the fixed points between the two graphs through the clear distance between two fixed point and measuring angle by, can detect the rotation angle of the robot. This paper discusses the second aspects is the ability to study the robot target tracking. The TLD detection technology used in the whole scientific test, hand robot in target detection and target tracking motion masking ability, finally obtained A very good result; on the other hand, the control of robot on the road at the vehicle tracking, the end result is the robot can accurately track, to determine the target table to complete the requirements. In a variety of experiments to obtain some good results: first, the rotation angle detecting robot, full use of the SIFT of science and technology. Through the detection, whether it is the first mobile robot plane of rotation or rotation of plane movement, rotation angle obtained are the same, in order to get the conclusion: the plane movement and rotational behavior of the sequence and not change the rotation angle; through further optimization of SIFT science and technology the running speed of the robot control system is increased by 10 times; the robot rotation angle detected and rotation time data for calculating the angular velocity of rotation . second, in the target detection capability of robot, full use of TLD technology, and through the regional division, such as the reduced sample method to optimize the TLD of science and technology, finally makes the running speed control system is increased by about 2 times, which is in accordance with the original scientific method of processing every frame image takes 70Ms. But using the optimized technology only costs 30Ms. through a balanced image precision and time, the image tracking 320*240 region.

【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】