Multi-camera Cooperation Design And Implementation Based On FPGA

With the rapid development of sensor technology,digital image processing technology and the embedded processor,cameras are widely used in video surveillance,human computer interaction,traffic control,machine vision,etc.Most applications systems use a single camera.As the application systems are developing more complicated and intelligent,multi-camera application researches come into being.In the intelligent video surveillance system,objects are easy to cover each other and miss in the single camera system,but the problems can be solved by multiple cameras system.Multi-camera cooperation is very important,which is an advantage but hard to research.Nowadays most application scenarios become more mobile and portable,which need an excellent real-time hardware processor.Based on the characteristics of programmable,real-time and low power in FPGA,this thesis puts forward a design of multi-camera cooperation target tracking based on FPGA,and test the cooperation algorithm and object tracking algorithm by Matlab;then implement the design,including the summary design and detailed design.In detailed design,some simple designs are used to replace the complex mathematical operations,and improve the processing efficiency.The main work: 1)cameras calibration,which describes the steps of calibration;2)camera synchronization,which through data synchronization by the external storage;3)cooperation control,to control the cameras and data association;4)particle filter on target tracking,improved the standard algorithm and make a simplify implementation on FPGA.ML605 development platform is used as a debugging environment,which has a Xilinx FPGA and three MT9P031 cameras to acquire image data in real time.In the indoor scene,pedestrians tracking tests are implemented under the system with the speed of 640x480@30fps.Test results show that the pedestrian with normal walking speed can be tracked accurately,and the system can record the location in real time.