Mobile Robot Platform Design And Implementation Based On μc/os-Ⅱ

In this paper, the development of mobile robot at home and abroad are analyzed by comparing the advantages and disadvantages in large number of robots, then a design of visual processing platform was proposed based on the ARM9 core for mobile robot. Then in the ADS integrated development environment, we completed the design of the BootLoader and improved system performance by the use of MMU and Cache, then completed the transplantation of theμC/OS-II system under S3C2440. Moving equipment of mobile robot in this article has used the drive system composed of L298N and DC motor. We finisheded the drive circuit design by using Protel, and adopted the two PWM outputs of ARM to control the mobile robot to go forward or backward, turn left or turn right. As video image acquisition and processing is a key part of the visual navigation, visual processing unit composed by the OV9650 and NEC LCD screen was adopted and the relevant drive design was completed in this paper. Experimental results show that it has the normal image acquisition and provide a solid foundation for the following visual processing and the LCD display also works well.Firstly, in the phase of target detection, one improved methods based on the maximum value of the color channels was adopted for the flaw of Background Difference, and the running results in ARM platform was given. Secondly a combination algorithm of the Background Difference and CamShift was used by analysing the initial window position of the target tracking algorithm. And the improved algorithm in the experiment was proposed, namely, the Regional Filtering Background Differential & CamShift algorithm. Finally, Self-Growth Frame Difference method was proposed by analysing the phenomenon of "Holes" in Frame Difference, and the method is verified effectiveness through its simplified algorithm. These algorithms effectively improved the effectiveness and accuracy of the target detection and tracking, the validation of which were proved in the circumstances of OpenCV and VC++ 6.0.