基于Beaglebone Black轮式倒立摆机器人视觉控制系统研究

发布时间：2018-11-15 07:42

【摘要】：轮式倒立摆视觉机器人拥有视觉识别目标的功能并可长时间跟踪目标,在实际生活中有着广泛的应用前景。机器人可以用两个共轴轮保持车身平衡、前进、转向,并利用车载的摄像头获取图像序列在空间域中分离运动对象,对连续时间中的目标序列识别与跟踪。本文自主开发了轮式倒立摆机器人平台,并实现一种鲁棒性好的视觉跟踪系统。轮式倒立摆视觉机器人基于Beaglebone Black开发板设计而成,并扩展大量的传感器元件,包括MPU6050姿态传感器、光电编码器等。根据BeagleboneBlack自身嵌入式系统的实际情况,利用Beaglebone Black的PRU单元设计了运动控制器,包括PD平衡、PI速度、PD转向控制器,实现轮式倒立摆视觉机器人动态平衡、前进以及转向控制。利用一种基于核相关滤波的改进Tracking-Learning-Detection(TLD)跟踪算法,解决目标跟踪问题。利用基于空间上下文算法判断跟踪过程的异常状态,增强跟踪效果。针对实际情况为视觉跟踪系统设计了单目测距方法,并利用开源VLC和Opencv库在PC机上完成视觉系统。PC机视觉系统将轮式倒立摆视觉机器人传输回来的图像数据进行处理,并将处理结果利用WIFI发送给轮式倒立摆视觉机器人。对于改进的跟踪算法进行大量场景测试,证实算法的可用性,并对结果进行分析。通过对软硬件进行整体调试,最终轮式倒立摆视觉机器人达到预期效果,完成设定的功能目标。
[Abstract]:The wheeled inverted pendulum vision robot has the function of visual recognition and can track the target for a long time, so it has a wide application prospect in real life. The robot can use two coaxial wheels to keep the body balance, advance and turn, and use the camera of the vehicle to obtain the image sequence to separate the moving object in the space domain, and to recognize and track the target sequence in the continuous time. In this paper, a wheel inverted pendulum robot platform is developed and a robust visual tracking system is implemented. The wheeled inverted pendulum vision robot is designed based on the Beaglebone Black development board, and expands a large number of sensor elements, including MPU6050 attitude sensor, photoelectric encoder and so on. According to the actual situation of BeagleboneBlack embedded system, the motion controller is designed by using PRU unit of Beaglebone Black, including PD balance, PI speed, PD steering controller, to realize dynamic balance, forward and steering control of wheeled inverted pendulum visual robot. An improved Tracking-Learning-Detection (TLD) tracking algorithm based on kernel correlation filter is used to solve the target tracking problem. The abnormal state of tracking process is judged by spatial context algorithm, and the tracking effect is enhanced. A single visual measurement method is designed for the visual tracking system in view of the actual situation, and the image data transferred from the wheeled inverted pendulum vision robot is processed by the PC vision system on the PC computer using the open source VLC and Opencv library. The processing results are sent to the wheeled inverted pendulum vision robot using WIFI. A large number of scene tests on the improved tracking algorithm are carried out to verify the availability of the algorithm, and the results are analyzed. Through the overall debugging of software and hardware, the final wheeled inverted pendulum vision robot achieves the desired effect and accomplishes the set function goal.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242

【参考文献】