Research On Face Recognition In The Situation Of Loss Of Local Information

In non-limited face recognition, due to occlusion, illumination and otherexternal factors interfere, information for face collected from facial images is oftendisturbed. Faced with this situation, PCA, LDA, LBPH, support vector machines andother traditional face recognition algorithms’ recognition rate substantial decline, andcan not meet the application requirements.In recent years, face recognition based on sparse representation has beensuccessfully applied to face recognition, and shows robustness to partial occlusionand the noise, but large areas of continuous cover still make it difficult to reach ahigher recognition rate. On this basis, a variety of sparse representation classificationbased on improved face recognition algorithms have been proposed to deal with theproblem of local information missing due to partial occlusion and light, but thepresence of both time-consuming and accuracy limitations.The main work of this paper is divided into two aspects: First, the proposedface recognition method based on weighted iterative sparse coding, missing weightby reducing the area of information, reduce environmental interference and improverecognition rate. In continuity with the prior occlusion condition, the calculationmade based on the weight field analysis, to obtain a more accurate effectiveocclusion weights FIG.Secondly, lack of sparse representation classification common dictionarysparsity, the proposed formula to identify the credibility of the residuals definedaccording to the classification. We use the criteria of credibility as a screenedsparsity inadequate test samples, and using local features with spatial information forsecondary identification.Finally, experiments show that the proposed algorithm in the case for generalrecognition of the situation, with a high block, light robustness. And compared withsimilar algorithms, has the ability to identify more efficient.