Design And Implementation Of Face Recognition System Based On Compressed Sensing

With the rapid development of computer science and technology,face recognition technology has gradually become a research hotspot in computer vision and biological information.Face recognition technology is widely used in the fields of identification,public security monitoring,etc.due to its low cost,non-invasiveness,and ease of operation.How to better deal with the limitations of high-dimensional images and sample size has always been a key and difficult point in face recognition algorithms.The theory of compressed sensing breaks the limitation of traditional Nyquist sampling frequency,which greatly saves the storage space.This paper studies sparse representation classification algorithm(SRC).A residual compressive sensing reconstruction algorithm is proposed by using compressive sensing theory,and a face recognition system is designed and implemented based on compressed sensing.The main work is as follows:(1)Studying the basic theory of recognition,including the face image detection,image preprocessing,facial feature extraction and classification algorithm.The focus is on the research and simulation of the image gray level,histogram equalization,denoising and the principal component analysis in feature extraction.(2)This study apply sparse representation in compressed sensing theory to face recognition technology,and conduct intensive study in the core theoretical knowledge including sparse representation,measurement coding and reconstruction algorithms.In the design of the reconstruction algorithm,the discrete wavelet transform base is sparse base and the Gaussian random matrix is the measurement matrix.Simulation and result analysis of image reconstruction are carried out on two algorithms: basis pursuit and orthogonal matching pursuit.(3)The paper proposes a compressive sensing reconstruction algorithm based on residuals and combines it with uniform block SRC algorithm.Using the local sparsity within the image and the non-local similarity between image blocks to improve the reconstruction performance,the problem of solving the sparse coefficient by the minimal norm method in the traditional SRC algorithm is optimized.The experimental results show that the improved uniform segmented SRC algorithm has a significant improvement in face recognition rate and robustness.(4)A web-based face recognition system is designed and implemented by using B/S mode.The system includes five modules: user registration,face detection,image preprocessing,face feature extraction and face recognition.The system is tested and analyzed in a self-built face database.The recognition success rate up to 87.5%.