Implementation Of Moving Object Detection Base On FPGA

Moving object detection is an important branch of computer vision technology.It is the basis of subsequent video processing,such as moving object tracking,recognition and behavior understanding.It has a wide range of applications in medical,aviation and surveillance.With the continuous advancement of human exploration in the fields of medical imaging,artificial intelligence and virtual reality,people have higher and higher requirements for the accuracy and real-time performance of moving object detection systems.Simple software processing methods cannot process high-definition images in real time.Field Programmable Gate Array(FPGA)hardware platform can perform parallel processing of data,process data quickly,and has the flexibility of function programmable.Therefore,it is of great significance to realize moving object detection system on FPGA platform.In this paper,several moving object detection algorithms are analyzed in detail,and the advantages and disadvantages of each algorithm are introduced.Finally,the ViBe algorithm is selected as the main research object,and the shortcomings of the existing ViBe algorithm are improved.The simulation results are carried out under the Matlab platform.The preliminary results of the improved algorithm are obtained.The results show that the improved method is effective and feasible.Then the algorithm is designed on the SDSoc software platform and improved by the software and hardware coordination method.The algorithm improves the real-time performance of the detection system.In addition,this paper also introduces the image processing algorithms used in moving object detection systems,including median filtering and image morphology processing,and we carries out hardware design and implementation,which improves the accuracy of the system detection results.In this paper,we built test platform on the Xilinx Zynq Ultra Scale+ MPSo C ZCU104 development kit.The improved ViBe algorithm is implement as the core to realize the moving object detection system.Then the CWD2014 datasheet is used to test the performance of the implemented moving object detection system.The results show that the hardware processing results of the improved algorithm are higher than the original ViBe algorithm,GMM algorithm and PBAS algorithm in terms of recall rate and F1 parameters.The detection accuracy is higher than the original ViBe algorithm and GMM algorithm,and the Percentage of Wrong Classifications(PWC).Compared with the original ViBe algorithm,GMM algorithm and PBAS algorithm,the improved algorithm has better detection performance,and the processing speed of hardware implementation is 5.8 times that of software implementation.It can process 1080 p 30 frames of video stream in real time.This paper also evaluates and analyzes the overall power consumption of the moving object detection system.The results show that the on-chip power of the system is 7.7W,which is much lower than the power of the CPU platform.It is very suitable for embedded applications.The system has a high real-time and good detection effect and has practical application value.