Research And Implementation Of An Embedded System Of Human Face Detection And Face Recognition

Face Detection and Face Recognition are two of the research focuses in thecomputer vision and artificial intelligence field. Facial feature for authenticationcompared to other human biological characteristics such as Fingerprint, iris, palmprints, voice, etc. has simple, accurate, user-friendly and easily accessible features.Face recognition technology perfect blend of safe and convenient. Once developed, itwill greatly change people’s lifestyle.Embedded system with its cost-effective, low power consumption, small sizefeatures widely used in the field of consumer electronics. The face detection andrecognition algorithm on the embedded system will bring a lot of effectiveness. But theembedded system hardware resources, the operation speed is limited, it can`t becomparable to PC. So, face detection and recognition algorithms still can`t meet theapplication standards. There are the slow speed of face detection, high false detectionrate, low recognition rate and other shortcomings. On the basis of reading a lot ofreference, the paper focus on how to improve the speed of face detection, facedetection rate and reduce the memory, do the following:Firstly, a new embedded face detection system is proposed to solve the problem ofslow speed in real-time video face detection. In this paper, at first, do Adaboost facedetection on high color image base on skin detection algorithm. Then, morphologicalprocessing detection results, Set region of interest (ROI) on the next frame, and doAdaboost face detection on ROI. Experimental results show that the method hasimproves the detection accuracy and the speed.Secondly, for the slow recognition speed and the registration characteristics of thesample differences week problem of PCA face recognition algorithm, proposed aautomatically remove similar registered photos during face register and dual-thresholdmethod to improve the recognition speed, reducing storage space and the falserecognition rate. Lastly, optimized embedded software design, effective use of the embeddedsystem resources, multi-threaded approach to improve software efficiency.