Research On Face Recognition Based On Saliency In Complex Environment

Face recognition is one of the most representative methods in identification technology and has important research value.Face detection is performed first on the input image,and then face features are extracted of the detected face.Finally,the extracted features of the face are classified and identified,and the identity information is determined.The facial features vary greatly between people,and the whole process of face recognition can be done by non-contact method,so it has broad application prospects in various fields such as public security,aviation and finance.However,the face recognition method is susceptible to external disturbances.The proposed face recognition methods are mostly suitable for simple constrained environments,but not effective in the actual complex unconstrained environment.Therefore,it is necessary to further study and improve the method of face recognition to make it more suitable for real life.In this paper,we study the face recognition in complex environment and propose a method of extracting face invariant features by combining sparse coding and LBP operator.Using sparse coding algorithm to simulate the visual perception mechanism,and the algorithm is used to train the input image to get multi-scale and multi-directional feature extraction filter.Aiming at the shortcomings of traditional sparse coding algorithms which are computationally intensive and time-consuming,an improved fast sparse coding algorithm is adopted to improve the recognition speed.The resulting feature extraction filter extracts the overall appearance contour of the image,which is complementary to the local feature extracted by the LBP operator,so the two features are fused.Experimental results show that the proposed method has higher recognition rate than traditional methods in practice.The human eye vision system helps the brain filter out unimportant redundant information and focus on the features of the area of interest.According to the human visual attention mechanism,the significant sparse coding feature is calculated and analyzed,and the noticeable saliency map of features is constructed by using the theory of information maximization.It mainly analyzes the saliency map generated under each interference factor and the influence of introducing the saliency map on the recognition result.Comparing the saliency map with the saliency map obtained by the traditional Itti method,the result shows that the saliency map generated by this method is more suitable for the face recognition system.After the saliency map is introduced,the recognition effect is greatly improved.Finally,an online face recognition experiment system is built for face recognition in real-time video.The system uses the above-mentioned feature extraction and saliency analysis algorithm,which is suitable for static and video images.It has faster recognition speed and better recognition effect,and it can be applied in the actual scene.