Research On Feature Extraction And Classification Algorithm For Face Recognition Of Coal Mine

Video monitoring system of coal mine overcomes the defect that the show of punch card is not the same people, plays an important role in practical application problems such as post disaster emergency rescue and miners identity authentication. The existing face recognition methods have shown good performance in controlled environment, but the recognition effect is not significant under the conditions of non controllable environment, such as when occlusion and illumination is not sufficient enough. The thesis aims at the interference exists underground coal dust, dust, sound and other kinds of noises, increases the difficulty of monitoring identification, the key technology and process of face recognition are studied,which have great application value.For that the underground coal face image is susceptible to interference and general feature extraction methods is sensitive to noise, this paper proposed coal mine image of difference characteristics extraction method based on the Shearlet transform. The method uses with shearlet transform the direction choice characteristics and image feature representation capability, and uses the Shannon entropy theory giving different weights and feature encoding fusion different feature subgraph, effectively reducing the dimension. The experimental results show that the face image affected by noise pollution and dust that feature extraction and anti noise ability is good.In view of the problem of the original sparse description method that have high complexity and the number of training samples, this method presents a fast sparse description method based on different features of Shearlet transform for face recognition. First of all, this paper proposes the matching score fusion method used the advantages of sparse description method, which is considered again the "current" test sample, which has "the advantage of sparse"; Secondly, according to the matching score selected a training subset to description test samples, excluding the interference caused by the training samples, which are not similar to the test samples; Finally, the fast sparse description method is used to solve the optimization problem of L0 norm. The experimental results show that the proposed face recognition algorithm ensure high recognition rate and reduce the computational complexity.Taking the coal mine video monitoring as the background, this paper focuses on the process of image feature extraction and classification in the process of face recognition, which has obtained some research progress. A more accurate relationship between the number of scale direction in Shearlet transform and the recognition results will be the next focus of the study. The next plan would be the combination between location technology of face recognition technology and accurate personnel to improve the effectiveness of disaster relief greatly.