Design And Implementation Of Vehicle Tracking Systems Based On Intelligent Transportation

Now the motion and automatic detection and tracking of targets have manydifficulty that needs to be resolved in precision strike military targets and intelligenttransportation systems, it also is the research hotspot in the field of computer vision.Mean-shift is a non-parametric estimation of the density gradient, which was first usedfor cluster analysis in pattern recognition, and has been widely used in target tracking inrecent years.Video pre-processing stage of moving vehicles in MS tracking algorithm includesimage smoothing filter, background update Moving target detection and coarsepositioning and so on. Then on the basis of the traditional MS tracking algorithm, thepaper focuses on the vehicle target tracking under dynamic scene. MS algorithmidentify the vehicle using SVM, and merge the bandwidth of kernel function andtemplates target update mechanism, This paper presents the new algorithm in dynamicenvironment, which weights target template using the background and candidate targetarea using kernel-weighted, remove the influence of t background pixel of targettemplate histogram, the algorithm achieves more accurate target tracking.To the targetof Large scale changes,only when bandwidth and scale of the target are matched canmake tracking more accurate and much steadier, so it requires update of the kernelfunction bandwidth. the template can reflect latest features of the target, while at thesame time retain inherent characteristics, so it requires template update. Due tointerference and environmental factors, there will be a loss of vehicle target tracking,using linear prediction to predict the location of the target candidate, it enhance therobustness of designed system.This paper designed vehicle tracking system uses a dual-core technology based onARM and DSP, this system has wide application prospect. The dual-core features basedon HPI interface can improve the real-time of system, and completed the design of theperipheral circuits and driver, while conversion technology with timing and logic of theCPLD control system, to design the video capture circuit, it Make a simple hardwarestructure. FIFO is used as a buffer between image sensor and DSP, this method cansolve the inconsistency of operating frequency of camera and DSP processor, it canprove real-time. MS algorithm is eventually transplanted to the design of the trackingsystem. After the optimization of test and debug, and this paper ultimately complete a miniaturized optical tracking system based on vehicle tracking.