Research On Robot Vision And Target Recognition Technology

With the development of digital image processing technology,robot vision technology, target identification and detection technology has become a hot topic in the field of artificial intelligence.However, the current image vision technology has the characteristics of computational complexity,low real-time, inefficient implementation, and poor system portability. How to design a embedded target identification system with low-cost, high real-time, low-power has been put on the agenda.This paper propose a kind of embedded image processing system based on DSP+FPGA, can solve the above problems perfectly. This paper analyze the advantages of embedded systems, and propose a viable solution for embedded systems. Firstly analyze the basic knowledge of video signal, use a generic video image sampling chip SAA7113H, realize analog signal to digital signal conversion; Secondly, analyzes the advantages of the image processing algorithm based on FPGA and the FPGA implementation; Then, with a high performance DSP as the controller of TI company, to realize the whole system operate coordinately; Finally,through configuring the USB module, to realize data exchange between image processing system and PC system,and real-time processing.Through the monitor to get the ball coordinate information, when the ball moves a certain space coordinates, By using the ball coordinate information, control the camera tracking ball movement,to realize the purpose of real-time tracking.This article designed an embedded image acquisition card, having realized the real-time video data acquisition and processing, has the advantages of flexible structure, strong versatility suitable for modular design, can be widely used in the lab, warehouse, bank, bus station which need to record the real-time scene field.