Design Of Automatic Target Tracking System Based On DM642

Automatic target tracking technology based on video image is a hot spot in thefield of military, security monitoring, and biomedical research. At present most ofautomatic target tracking technology is still built on the basis of the algorithm modelsimulation by computer, not in the specific system. With the rapid development andwide application of DSP, making the combination of embedded systems, imageprocessing and computer vision a lot of development. From the angle of systemimplementation, this thesis designs an automatic target tracking system with DM642asthe core, that can adapt to complex environment. The main research contents are asfollows:①The design of the hardware platform is the basis of the system. ResearchDM642chip features, as the core construct hardware platform, complete the interfacesdesign based on the system design requirements, mainly includes: according to theshortcomings of the software codec, use codec chips to implement the sampling andsynthesis of analog video signal, and complete the design of image capture and displayinterfaces; complete the design of network interface and I2C expansion interface.②Divide the system software into five tasks from function: image capture, imageprocessing, image display, PTZ control and network communication, and constructmulti-task communication system software architecture based on DSP/BIOS, then eachtask as a module complete the system software design. The details are as follows: studyFVID driver and develop video applications to implement video capture and display;image processing is the core of the system software design, according to the actualscene and the characteristics of the hardware platform, analyze and select appropriatealgorithms and transplant to the hardware platform, image processing is divided intotarget detection and target tracking in two stages, by the inter-frame difference methodto obtain the ADI image and extract the template features to implement moving targetdetection, based on linear prediction algorithm to estimate the location coordinates oftarget next frame may appear, then use correlation tracking algorithm combined withrevise measures such as template adaptive update, re-capture of the target whenloss-of-lock, to positioning the focus of target in successive image frames; research PIDcontrol principles, combined with position prediction proposed the PTZ control strategyto implement vision-guided automatic tracking of the target; implement network communication between PC and DM642, and design the human-computer interactiveinterface on PC.By optimization and debugging, the system runs as the expected results, that canadapt to relatively complex environment such as target deformation, distance changes,partial occlusion, similar target interference, with a certain degree of stability andreal-time.