The Research On A Target Recognition And Tracking System Using Monocular Vision

The rapid development of machine vision technology in recent years gradually risesto a new study neighborhood-visual servoing. The main research discussed in this articleis how the visual servoing uses visual feedback information on the control of2-DOF PTZ,and make it to track a moving target.This paper firstly analyzes and makes contrast about the basic algorithm of the imagefiltering, image segmentation, object detection and tracking, and thus set up their owntracking algorithm. Specific processes include the establishment of target detectionalgorithm based on frame difference method, and utilize the improved OSTU method tosegment moving target, with the mathematical morphology, we remove interferenceinformation in binarization pictures, therefore we obtain the position deviation betweenthe centric position and the principal point of the image.Secondly, according to the proposed tracking algorithm, we build a camera-PTZservo control model, whose purpose is to control the PTZ rotation angle in real time, sothat the moving target is always keep at the center position of the camera. With the pinholeimaging principle, we turn the deviation mentioned before into the angle of the PTZ, andfeedback the information to control system to make the PTZ rotate, so that the camera cantrack the target. Meanwhile, according to the design requirements we develop our owninterface by C++secondary development in order to operate the images easily.Experimental results show that the proposed camera-PTZ control model can track themovement of the target ball.Finally, we summarize the work of the article, and point out the algorithm stability,real-time and other aspects could be improved. And prospect the work that the proposedcamera-PTZ model can be used to intelligent manufacture, online detection and real-timemonitoring of the work piece.