Design And Implementation Of Target Tracking System Based On FPGA

As an important branch of robot vision,target tracking has gradually become a hot research topic.It is widely used in all aspects of production and life.In terms of military aspect,it includes missile guidance,UAV and so on.In civilian area,it includes medical diagnosis,virtual reality,intelligent driving system,human-machine exchange and so on.In the situation of high realtime requirement,such as unmanned aerial photography,intelligent driving and so on,the algorithm can't meet the real-time requirement with software.According to the advantages of FPGA parallel processing,the parallel processing of image processing algorithms on FPGA can greatly improve the speed of image processing and meet the requirements of real time.Therefore,image completion algorithm based on FPGA has gradually become a hot research topic.Aiming at the shortcomings of general computer in image processing,a target tracking system based on FPGA is designed in this paper.First,the target recognition and target tracking algorithms are introduced and compared respectively.The target recognition algorithm introduces the optical flow method,the background difference algorithm and the inter frame difference algorithm respectively.The target tracking algorithm is classified,including feature based target tracking algorithm,contour based target tracking algorithm,model based target tracking algorithm and template based target tracking algorithm.Through the summary and comparison of the various algorithms,the background difference algorithm is used to recognize the moving target,and the template based matching algorithm is used to track the target.In this paper,a target tracking effectiveness discrimination method based on absolute error and algorithm(Sum of Absolute Difference,SAD)is designed for the tracking algorithm of template matching which is easily disturbed by external factors.Based on the generated SAD matrix,the correlation between the position point of the target in the SAD matrix and the adjacent position points is used as the judgment basis,and the target tracking effectiveness is judged by the maximum correlation confidence degree(Maximum Close Distance,MCD).When the target tracking failure is determined,the new template of the weighted cumulative production is carried out through the optimal matching position of the target initial template and the current frame to continue the target tracking.The algorithm is simulated by Matlab software.The experimental results show that the algorithm can achieve continuous and effective tracking of target.In contrast,the original SAD algorithm is significantly improved in terms of anti occlusion ability.The result of ModelSim software simulation proves that the implementation of image algorithm by FPGA is faster than that of software implementation.In this paper,the whole system is divided into five parts by modular design,which are video image acquisition,tracking result display,image preprocessing,image storage and image processing.Through the hardware description language Verilog HDL,we design,compile and simulate the Quartus II software.Finally,we deploy the program to FPGA to achieve target tracking.Through many experiments,it is verified that this design can achieve accurate tracking of moving targets.