Research And Implementation Of Video Capture And Face Detection Based On ZedBoard

Face detection is an important branch of the technology of target detection and recognition.Its purpose is to detect faces and identify face regions in a given static or dynamic character scene,which provide the basis of face recognition and other technologies.With the increasing popularity of video technology,digital video capture technology is being widely used in the information collection on the front of related systems.Products that combine embedded video capture technology and face detection technology are also constantly infiltrating into people's daily life,such as intelligent access control system,intelligent monitoring system and so on.Based on the current popular embedded video capture technology and face detection technology,this paper focuses on the research of embedded video capture and video face detection.First of all,this paper introduces the current status of video capture technology and face detection technology,and based on this basis,two design schemes of USB video capture and face detection system based on embedded Linux have been given.Both of these programs are based on the ZYNQ development platform of Xilinx company,we focus on video capture technology and face detection technology,and then combine video frame capture and dynamic face detection into Zynq-7000 SoC.In this paper,ZedBoard is used as a hardware platform,and two different schemes are proposed for ZedBoard based video capture technology.The first plan is using the FFmpeg technology to call the image capture function in OpenCV for the video image data acquisition and displays the video images captured by OpenCV using the Qt GUI.The second way is to capture video data from the USB camera using V4L2(Video for Linux two)interface,then uses the Qt program to integrate the video capture and display programs into one application.Finally,both schemes combined with the OpenCV library using AdaBoost face detection algorithm to achieve dynamic face detection for input video.Through the test,the average time of face detection with these two schemes is 18 ms and 23 ms respectively,they can meet the requirement of real-time detection well.In this paper,two different schemes are proposed for the video acquisition of USB camera on ZedBoard.It has some reference and reference value in the fields of high definition video capture processing and target detection.