Design Of Ultra High Speed Target Tracking System Based On SOC FPGA

The field of computer vision includes a variety of image analysis and processing techniques,and object tracking is an important part of it.In the existing object tracking algorithm,there is a contradiction between accuracy and speed.The algorithm with high precision often has a slower speed,while the algorithm with faster speed has lower precision.For the requirements in different scenarios,we need to weigh the speed and accuracy.In the ultrahigh-speed object tracking scenario,the tracking speed must be extremely fast and the accuracy should not be too low.Therefore,it is important to study an object tracking algorithm with extremely fast calculation speed and acceptable accuracy.The missile-borne application scenario has high requirements on the size and power consumption of the object tracking platform.The So C FPGA platform can meet the needs of this scenario.It has both ARM system control capability and FPGA high-speed parallel computing capability,and through appropriate architecture design can further improve the operation speed of the object tracking algorithm.Based on the analysis of the existing object tracking algorithm,this paper improves the algorithm and structure based on a higher speed algorithm,and improves the calculation speed while maintaining the tracking accuracy.After getting the improved algorithm,the architecture of the algorithm is designed for the So C FPGA platform,and it is transplanted to the XCZU7 EV platform,and finally an object tracking system with ultra-high-speed object tracking capability is obtained.Firstly,this paper compares the existing tracking algorithms horizontally,determines the basic algorithm used in this study as KCF algorithm,and then improves the KCF algorithm according to the structure and operation characteristics of FPGA,using simpler HSV features and linear kernel.Compared with the original KCF algorithm,it improves the tracking speed without significant loss of accuracy,so the improved object tracking algorithm is used as the algorithm used in the final tracking system.Secondly,according to the structure of the So C FPGA platform,the functions and architectures that the processor and FPGA need to complete are designed.For the processor,the embedded Linux operating system is run thereon,and the storage of the video image sequence,the extraction and transmission of the object image block,the reception of the object position offset,the calculation of the object actual position,and the sequential control of the above operations are completed.For FPGA,it uses deep pipelined and highly parallelized architectures,while using Fast Fourier Transforms and on-chip memory resources to further accelerate the algorithm.The object tracking system architecture designed in this paper can simultaneously exploit the computing characteristics of the processor and FPGA,and meet the requirements of the tracking speed and flexibility of the tracking system.In this paper,the accuracy and speed of the ultra-high-speed object tracking system are tested.The results show that there is no difference between the overall tracking system and the improved object tracking algorithm in terms of accuracy.Compared with the original KCF algorithm,the accuracy is reduced by about 8%.In terms of average speed,the improved object tracking algorithm implemented by software can reach 1241 fps,the original KCF algorithm is 287 fps.Does not include the time spent reading the test sequence from the SD card,the tracking speed of So C FPGA platform can reach 2503 fps.There are more significant improvements.Therefore,the So C FPGA object tracking platform developed in this paper has ultra-high-speed computing,portable,flexible and scalable features,which can meet the application requirements for tracking ultra-high-speed objects and achieve research goals.