Target Recognition And Tracking Algorithm For Ground Mobile Robot Based On Binocular Vision

With the development of science and technology,more and more companies have begun to use the related technologies of computer vision to enhance their product's function.More and more scholars are investing in the field of image processing and trying to solve practical problems in many aspects.Based on the National College Student RoboMasters Robot Competition,this thesis focuses on the recognition and tracking technology of the ground mobile robots.Recognition algorithm and tracking algorithm are realized in the embedded computer.And this thesis designs a hardware architecture and software architecture for the RoboMasters robot system.According to the competition tasks,the recognition and tracking algoritlhm of RoboMasters robot system is designed on stereo vision technology.The main researches are as follows.Firstly,this thesis introduces the principle of binocular stereo vision.In order to solve the problem of the distortion of binocular camera,the method for camera calibration is designed in this thesis.The matching algorithm of binocular image is optimized for real-time ranging on the embedded computer.Secondly,this thesis designs the recognition algorithm for armor plate according to the features of robot.Multi-feature information of the robot is utilized to design the recognition algorithm which can improve the accuracy of the vision recognition system.Then,the target tracking algorithm is designed in this thesis.Kalman filter is utilized to filter the target position information.Therefore,the recognition precision of the target position can be effectively improved.Meanwhile,this thesis also analyzes and tests the TLD tracking algorithms and ASMS tracking algorithms.The test results show that the proposed algorithm can obtain higher tracking precision than TLD and ASMS.Finally,the architecture of the system hardware and software is designed in this thesis.TX2 embedded computer is utilized to process vision algorithm.STM32F427 microcontroller is utilized to process control task.The ROS robot operating system is run on the TX2 computer platform.The FreeRTOS multi-task embedded operating system is run on the STM32F427 embedded platform.In the laboratory environment,this thesis tests the target recognition and tracking algorithm of the ground mobile robot.And the overall test of system shows that the system can accurately identify and track the target of the ground mobile robot in real time.