Research On Effective Face Detection For RFID And Face Identification System

The performance of face identification technology is always limited by distance,pose, illumination conditions and image quality. To face identification technology, agood face image should be a clear frontal face image which have plenty pixels andacquired in normal illumination. We call these images effective face image.Most of the current face recognition techniques assume the face images acquiredby face detection are effective face images. However this assumption may not hold dueto the motion of face and environment conditions in reality. As a result, it is important todo effective face detection to images acquired by face detection.In this paper, we research an effective face detection algorithm, and use it in ourRFID and face identification system. It could determine whether the acquired images byface detection is helpful to the face recognition and discard the blur images which wereharmful. It insures the face recognition is steady and work well.The main research works are as follows:①We review some face pose estimation algorithms. Consider the need of oursystem, we choose the geometric methods algorithm to program, because it is fast andsimple. We test its performance on the Honda/UCSD Video Database and have somegood results.②We present a novel no-reference objective image sharpness/blurriness metricfor face recognition system. The proposed metric is able to measure the blurriness of aface image and help face recognition system to discard the blur images. It does notdepend on any assumption about distortion type or content. We test the performance ofproposed metric on the LIVE IQA Database, and compare it with some recentcompetitive IQA algorithms. We follow the suggestions of VQEG report to measure theperformance of our algorithm. Simulation results with well known methods show thatthe performance of the proposed metric outperforms conventional NR IQA metrics.③We design a RFID and face identification system to upgrade the performanceof tradition face identification system. Base on the research of face pose estimation andNR IQA, we find some way to do the effective face detection and use it to our system.This system has been used to Ji an garden in Jiangbei district of Chongqing and a Military Airport now.