Based ZYNQ Realization Real-time Face Detection Technology

Face detection has become a key technology in the pattern recognition and computer vision, Which is also a bottleneck restricting the development of machine vision. Meanwhile, the hardware and software co-design based ZYNQ platform for computer vision problems in the complex mode detection has opened up the idea, has become a hot topic, with a very broad application prospects. So the research on spread ZYNQ based real-time face detection system has important significanceThe paper firstly introduces the basic method of face detection and performance evaluation criteria, focusing on the traditional AdaBooat detection algorithm, and studies the architecture of the ZYNQ processor chip. On this basis, the two detection algorithm is designed, which is based on image preprocessing and experimental analysis of two algorithms designed by comparison, finally, select the serial program as paper’s detection algorithm. This improved the shortcoming of traditional algorithms in real-time and accurate detection of areas. The image pre-processing which in algorithms includes image enhancement, color segmentation, image filtering, edge detection and connectivity domain calibration and other modules. The AdaBoost cascade classifier contains AdaBoost algorithm implementation, Cascade classifier training modules. And then the VC test ystem of the real-time face detection has been builded, using simulation to verify the feasibility and performance. Simulation algorithm mainly includes image acquisition algorithms, image preprocessing, face detection based AdaBoost cascade classifier algorithms and real-time display algorithm. At last, the paper studies the ZYNQ implementation of face detection system based on ARM+FPGA architecture, he feasibility and correctness of the design and detection of face detection system based ZYNQ processor architecture is verified by experimentsIn addition, in order to ensure real-time detection, in this paper, the data source is detected from the image inserted in ZedBoard panel USB digital camera, and then the test results is displayed by the PC, which is connected to the VGA display interfaces where on the ZedBoard via a data cable. At the last, the accuracy of the algorithm designed in this paper is verified by1754different types of face image which is selected collection.