Research And Circuit Implementation Of Face Detection Based On FPGA

With the development of information technology and the popularization of e-commerce,the information security is becoming a very serious problem.Traditional authentication methods can't meet the needs of the application,so biometric recognition methods emerge on the stage.Face recognition is one of the most promising applications in biometrics,while face detection has a critical influence on the performance of face recognition.However,there are still many problems in the research of face detection.One is that the face detection rate isn't high when the illumination condition isn't good;another is that the most of algorithms are based on the PC platform which can't meet the needs of real-time.In order to solve these two problems and achieve the goal of real time face detection under weak illumination,this paper studies the main algorithms of face detection and digital image processing technology,improves skin color algorithm and designs a face detection hardware system based on FPGA.Firstly,this paper makes a series of improvements to the traditional skin color algorithm,adopts the Gauss model of YCbCr to do the skin segmentation,uses the logarithmic transformation method for weak illumination compensation and combines with morphological and spatial filtering to remove the noises.Then the hardware framework is designed which is divided into three parts: image input and output,image cache and image processing.And the sub modules are designed according to the function of each part.In this paper,the improved algorithm is verified by Matlab.The detection rates of single and multiple face reach 97.75% and 90%;and their average detection time are about 0.3s and 2.1s.Then each sub module is verified by Modelsim.Finally,this paper realizes the face detection system based on DE2 and D5 M by Quartus II which detection rate is 97.13% and speed is about 30 fps.The results show that the improved algorithm and the hardware system are feasible and can achieve real-time detection with high detection rate under weak illumination.