Design And Implementation Of Face Recognition System Based On Illumination Compensation Method

The performance of target recognition,especially face recognition system,is greatly affected by the change of illumination.Based on a low-light image enhancement method,the illumination compensation problem in face recognition technology was studied systematically,and a face recognition system based on convolutional neural network was designed to verify the proposed method.The main research contents are as follows in the dissertation:Firstly,the problems in the face recognition was analyzed systematically by the current research of face recognition,and the result is obtained that the illumination has a more significant impact on the performance of face recognition systems.Based on the summary of the current solutions to the above problem in face recognition,the method of image enhancement and the method of the face recognition based on convolutional neural network were studied systematically.Secondly,aiming at the problem of illumination image estimation in low-light image enhancement algorithm of the Retinex model,a low-light image enhancement method based on YCb Cr color space is proposed.The original low-light image is transformed from RGB?Red Green Blue?color space to YCb Cr color space.The Y component in YCb Cr color space is extracted and the initial illumination map ₁L?x,y?is constructed.The enhanced illumination image ₂L?x,y?is obtained by the gamma transformation of ₁L?x,y?,the enhanced image R?x,y?is obtained according to the Retinex model,and we use a multi-scale approach to boost the details of the image R?x,y?and obtain the final enhanced image _eR?x,y?.The experimental results show that,the method can not only effectively improve the brightness of the low-light images,enhance the details of the image,obtain a better visual effect with fewer color and lightness distortions.At the same time,an external light compensation system is built,which takes STM32 microcontroller as the control core and LED as the light source,and the dimming instruction of the face recognition system is received by wireless way to solve the problem of illumination change in face recognition system.Finally,a face recognition system based on convolutional neural network is built based on the above methods,and the face recognition experiments were conducted under different ambient illumination conditions at 2 meters,3 meters and 4 meters away from the camera respectively to study the influence of different distances and illumination on face recognition rate.In addition,the low-illumination Retinex image enhancement method and the external supplementary light system proposed above were applied to the system,and the face recognition rate was compared and analyzed.The experimental results show that,compared with the face recognition system without light compensation,when the illumination value is 10.9lx at 2 meters,the face recognition rate of the system is increased by 69%.When the illumination value is 15.6lx at 3 meters,the face recognition rate is increased by 41%.