Research On Face Recognition System Based On ZYNQ

With the rapid popularization of video surveillance systems, many video surveillance applications have an urgent need for a rapid, non-contact identification technology, in order to quickly verify the status of the targets, to implement access authorization or to achieve early warning functions. Face recognition technology is undoubtedly the best choice. Fast face detection technology can find human faces in video image and compare them with the face database for rapid identification.In order to integrated the functions of video capture, face recognition and display in an embedded system, this face recognition system is designed on ZYNQ development platform of the Xilinx company. The system function is designed with both hardware and software. The preprocessing part is achieved by the FPGA processor and the video image acquisition, face recognition part are accomplished by the ARM processors. This implementation makes full use of the logic resources and the parallel computing ability of the FPGA. This implementation also takes the advantages of the ability to accomplish complex algorithm and building up operating system of the ARM processors. The works were done as follows:(1)Establish an embedded Linux system on Zedboard, which includs Linux kernel, the cross-compile environment, OpenCV and Qt library file.(2)Control logic such as video transmission with DMA and the HDMI display is realized on the programmable logic part of the ZYNQ chip.The image preprocessing part is accelerated with the high-level synthesis tool.(3)Software design can achieve functions such as image transmission, image preprocessing control, video capture and face recognition. The video capture and face detection function are implemented with OpenCV, Adaboost and PCA algorithm.This paper designs an embedded face recognition system with both hardware and software of the ZYNQ. The system has the characteristics of good real-time performance, simple operation and portable.