Design Of Target Detection And Recognition System Based On Dual Camera Collaboration

Target detection and recognition algorithm is not only a middle-level video information extraction technology in computer vision technology,but also a prerequisite and basis for high-level behavior analysis and attitude recognition technology.It is widely used in weapon equipment,public monitoring,intelligent traffic control,bomb detection and other fields.the traditional single camera system can not consider clarity and field of view at the same time,foused on this problem,this thesis uses two core processing units,FPGA and GPU,thus fully combines both advantages to develop a target detection and recognition system with the dual-camera collaboration.it not only has the advantages of high integration,good real-time performance and low power consumption,but also can meet the requirements of monitoring wide field of view and obtaining high-definiton target images.The main contents of this thesis are as follows:(1)On the basis of comprehensive analysis of various mainstream processor strengths and weaknesses,in view of the disadvantages of the single camera working and the image processing system based on PC,this thesis adopts the dual-camera work cooperative mode which are telephoto camera and short focal camera,and also uses the FPGA + GPU dual processor hardware architecture to improve.It not only analyzes the significance of this system,but also introduces the research status of dual-camera cooperative working,target detection and recognition algorithm,and target detection and recognition system.(2)Starting from the difficulty of system design,the thesis clarifies the functional requirements and design indicators of the system,so it proposes an overall design scheme of the target detection and identification system with the dual-camera collaboration,and the overall working process is summarized.According to the overall system design scheme,the system is divided into four subsystems for independent design.the thesis introduces the functional design and device selection of each subunit module in detail,and completes the circuit schematic diagram and PCB design of the whole system.(3)By analyzing the various target detection and recognition algorithms,because of the double gaussian small target detection algorithm has simple principle and also easy to implement,so it can meet the requirements of real-time detection system.However,YOLO algorithm has fast speed and good recognition effect,so these two algorithms are used to for realize the function of this system.Considering the resources of the hardware platform,this thesis proposes the Calibration method and its cooperative strategy of the target detection and recognition system with the dual-camera collaboration.Furthermore,this thesis explains the principle and implementation steps of the dual-gaussian small-target detection algorithm for the short-focus part and the target recognition algorithm YOLO3-tiny for the long-focus part.(4)In order to ensure the stable operation of the system for a long time,the thesis measures the volume,weight and power consumption of the system and proposes a complete test scheme.The system which meets the perset functional requirements and design indexes is proved by the final test results.And it realizes the functions of real-time detection and identification of the image data with resolution of 1920×1080@60Hz.Through the actual scene test,in this thesis,the target detection and recognition system based on Dual-camera collaboration designed is proved to have the functions of monitoring largefield of view scene and target recognition,which has the advantages of low power consumption,portability and high integration.It can be applied to a variety of fields and has good theoretical and reference value.