Face Detection System Based On ARM+FPGA Heterogeneous Platform

With the development of science and technology,there is a variety of convenient in our life.At the same time,the ascending of the safety consciousness results in face recognition and face tracking technology is more and more widely used.Face detection as an important front-end processing part of face tracking and face recognition attracting more and more attention absolutely.Face detection is a process of determining whether a face is present in an input image.If it exists,the position the face will be given.After research and comparison of various of face detection algorithm,this design choose the Ada Boost cascaded detector as the face detection method in accordance with Open CV Open Source Library.Ada Boost algorithm has a good result of face detection,but it is not widely available because of the high complexity and high computation.Most of the face detection algorithms are implemented in PC by software recently,this kind of method is unable to meet the engineering requirement for miniaturization in practical application.In addition,with the development of video monitoring area,the high-definition camera is gradually replacing the lower resolution camera which was widely used in public.Therefore,it is necessary to speed up the face detection of high-resolution images.Nowadays many face detection algorithms are difficult balance in accuracy and speed.By measuring the time consumption of different functions,this design analyzes which module need to be accelerated by FPGA.Some designs have speeded up the whole detecting time to some extent by accelerating the image preprocessing based on hardware platform.The main innovation of this design is the hardware acceleration of integral figure calculation which is the key point of face detection algorithm,in addition to the hardware acceleration of image preprocessing.The calculation of the integral figure is implemented in parallel with FPGA in this design.The detection speed of face detection system has been greatly improved in this design.The hardware platform in this design involved the ARM and FPGA.Software platform refer to transplantation of Linux OS,setting up cross-compiling environment,transplantation and compilation of Open CV Library,the module division of hardware and software collaboration.This design relate to a wide range of knowledge and multiple dimensions.There is a certain reference significance for the whole machine learning platform development.