Research On Target Detection And Positioning And Path Control Based On Machine Vision

ABSTRACT:With the development of the computer technology, the mobile robot control technology has become a hot research topic. The mobile robot technology has been widely used in the military, shopping malls, hospitals, homes and other fields. It has entered people’s daily life. Positioning technology is one of the important functions of the mobile robot. The machine vision can provide abundant information, and visual positioning can have high precision. In this thesis, research on target positioning is carried out based on global vision. Based on this, path planning is carried out. And the result of target tracking is used as the feedback to control the motion of a robotic car.Firstly, to overcome the disadvantage of two methods commonly used in the moving target detection--the frame-difference method and the background-difference method, a method of moving target detection is proposed based on the combination of the three-frame-difference method and the background-difference method with Gauss model, so it can effectively combine the advantages of the two methods. It improves the effect of moving target detection by incorporating shadow processing.Secondly, in view of the disadvantage of the moving target tracking method based on the traditional CamShift algorithm, three improvements have been carried on:the results of moving object detection are used as input to the target tracking problem for automatic selection; the extended search window is used to adjust the dynamic search window so it can solve the robotic car tracking failure problem caused by the excessive instantaneous velocity; an AND logical operation is performed between the color histogram calculated by the CamShift algorithm and the color histogram of the area detected by the moving target detection, which improves its anti-interference ability. Using the improved CamShift algorithm and the Kalman filtering to trace the robotic car improves the performance of the target tracking method, and it can feed back the robotic car’s position information in real time.Then the camera calibration is realized through the coordinate conversion from the image pixel coordinates to the world coordinates, and the robot visual positioning is realized. Then a safe and effective path is planned by using the obtained visual information with the grid method and the A*path search algorithm.At last, a robotic car control test system is designed. The robotic car’s path is controlled by the visual feedback information of the robotic car. It verifies the validity of the methods proposed in this thesis for moving object detection, target tracking and positioning.