Research Of Intelligent Digital Monitor System Based On Face Detection And Recognition

The dissertation discusses in detail the application of fractional calculus in information science, and the developments and tends of wavelet transform theory, face detection and recognition technology, digital monitor system. Then it discusses the feasibility of applying the above technology to build a full new intelligent face detection and recognition system, and puts forwards methods for solving intelligent digital monitor system based on face detection and recognition. In the first, it puts forwards the theory and definitions of fractional wavelet transform and digital algorithm of fractional calculus of fractional wavelet transform, which is proven to be correct and high efficient in theory and practice. The theory and definitions of fractional wavelet transform, extents traditional integral wavelet transform to fractional one. Regard and implement wavelet transform in fractional world is the development to traditional wavelet theory. From view of fractional calculus theory, noise can be regard as isolated oddity point and texture detail of digital image has some highly self-resembling structure that has some oddity of fractional calculus. In order to reduce texture losing in texture processing for digital image by integral method, it puts forwards a recovery model based on the combine of integral and fractional calculus, which improves a lot the correctness and efficient of face marginal profile picking-up in pretreatment of face digital image. The model is that doing fractional calculus to face digital image, picking-up fracture information and taking it as texture detail of face digital image to compensate the deficiency in oddity signal picking-up by integral calculus methods. Applying fractional calculus in face profile picking-up strengthens marginal oddity information of traditional digital image, and develops picking-up algorithm. The theory and application is a pioneering to just arisen area that fractional calculus applied in information science. In the second, it puts forwards a novel high-efficient face detection algorithm in complex background. It is well-known that the veracity of face detecting has an important role for later face recognition. By studying the fundamental theory of face skin color model, it puts forward a method that automatically detecting multi color face in complex background. It firstly divides up skin regions from non-skin ones in YCrCb and HSV color space, clusters the detected skin pixels into several centers in CrCb space, applies morphological operation in every cluster center to remove the small background area, efficiently picks up skin from complex background by morphological filter algorithm based on fractional calculus, and forms choosing face region by region-combining. By repeating threshold method in choosing face region, it will detect pairs of eyes. In the last, it affirms by BP neural network. Experimental result shows that the correctness of the method can come up to 90%. At the same time, the dissertation puts forward and discusses ISOMAP based multi-pose face location recognition and analog model of non-linear dimension-reducing The model standardizes all kinds of lighting environment parameters by Adaptive dynamic adjustment algorithm to improve the Adaptive of the system to the environment. Compress original data space to eigenspace by ISOMAP algorithm in order to make data gathering in small efficient space, classify by improved network to improve self-learning ability, inspection efficient and real time monitoring. It greatly raises inspection and recognition speed of multi-pose face from different angles and reduces the storage of vast data in monitor system. ISOMAP is a dimensional-reducing method by using geodesic distances. By compare with other non-linear method, ISOMAP inverse transform can quick simulate face elevation image according to certain face information or known ISOMAP parameters. The experiment shows that the method reduces calculating greatly and roses anchor veracity. In the last, it puts forward intelligent digital monitor system model based onface detection and recognition, and studies the concerning theory problems. Firstly, bases on analyzing characteristic of the server group load balance, it puts forward LTI and LTI+ algorithm considering the traits of the move of concentralized load balance and data transfer of long-distance digital monitor system. Secondly, it combines the technology of the fractional digital image processing, face detection and recognition technology and LTI and LTI+ algorithms to reconstruct the traditional monitor system. Approving by theory and practice, it is a great technology improvement from traditional digital monitor system to artificial intelligent digital monitor.