Research On Feature Extraction And Recognition Of Face Image Based On Topological Structure

In recent years, with the deepening of research in the field of image processing and pattern recognition, The research related to face recognition also will be flourished. Topology of the face image does not change with the changing of distance, direction and other parameters of face image, therefore it is an important factor in studies of face recognition technology. This paper proposes a method of face recognition based on local structure feature and spatial topological relations. From the perspective of human cognition, aiming at the shortcomings of the triaditional method of face recognition which is time-consuming, based on the theory of topological pattern recognition,analyzing the local structural characteristics and spatial topological relations proposing a framework about pattern recognition which includes local structural features extracted from human face, the construction of overall topological relationships and modeling for local and overall features.There are many unique advantages on face recognition such as intuitive, natural, safe, reliable, and so on. As a kind of biological recognition technology, face recognition has a wide range of applications in the field of public security, banking, anti-terrorism and surveillance systems. integrated with other identification technologies, the market share increasing year by year, so the research on face recognition technology has important social significance.In this paper, choosing two-dimensional face image as the research object, using the method of sliding window to construct the overall spatial topological relations for the preprocessing face image, while changing the two-dimensional image into a chronological feature sequence of one-dimensional, treating the extracted features as local structure. The topological relationship is constructed by the order of window in the sliding process. Due to artificial neural networks and hidden Markov model has been applied in the field of face recognition. Meanwhile, in view of the artificial neural network with adaptive, self-organizing, self-learning and other ability such as the high efficiency of recognition, strong ability to resisting noise and so on, so this paper chooses Artificial Neural Network to make models for local structure of face image. Given the strong ability to deal with time-series by Hidden Markov Model, it is used to model the global topological relationship. Combining the artificial neural networks and hidden Markov models,we can achieve the classification of face images. A series of experiments have proved that the method has a capability of good anti-jamming and resistance to deformation and has achieved a better forecasting classification results.