The Video Sequence Face Target Tracking Technology Based On Fpga

With the constant development of FPGA, its parallelism, high speed and reusability have reflected its advantage of flexible application in control domain. Using the IP core has greatly shortened the development cycle of the researchers, ensuring the stability, easy-maintenance and portability of the module. In the mean while, face tracking, as an important branch of video target tracking, has gained more and more attention. It has been widely applied in security monitoring and intelligent human-computer interaction, with the trend of rapid development.The subject of this paper is a study on the technique of face tracking on the basis of FPGA. In this paper, the designed system, based on FPGA hardware platform, has the property of completing the final tracking task by constructing particle filter with the eigenvector of face color information. This paper firstly states the subject meaning, face detection and tracking technique briefly. Then it presents color system, FPGA development process and conventional target method. At last, it introduces the concept of particle filter.This paper researched on face tracking by particle filter based on the FPGA.This paper consists of two parts:hardware circuit and algorithm design. Hardware circuit includes SAA7113H decoding chip, SAA7121coding chip, Xilinx Spartan3E main control chip and off-chip memory.And schematic diagram design, debug,placement and routing are completed by designing PCB circuit with Altium Designer.In the part of algorithm design, face detection and tracking is based on FPGA. And particle filter algorithm, based on face color characteristics, is constructed with skin color difference signal as the feature.The calculation of a set of weighted sampling particles can represent the posterior probability density. ITU-656video format and YCbCr4:2:2sampling format are analyzed.Detecting the color difference signal and calculating the weights.Finally,particle state reflected by the biggest weights gets the result of detection.The system is written by Verilog HDL and developed on Xilinx ISE platform with the call of related IP core resources and the debugging by simulating tools and oscilloscope. The final system realizes face tracking, achieving the expected effect.