Research On Face Recognition Algorithm Based On D-S Evidence Theory And Local Neighborhood Pattern

With the research and development of artificial intelligence in multiple disciplines and fields,digital image processing technology has also become a hot topic,and has gradually penetrated into people's production and life,applied to various fields,and has become an important research direction in the current computer field.In order to achieve efficient face recognition,improve the anti-interference of related algorithms,and ensure the integrity and authenticity of the face,this paper studies the face recognition algorithm based on D-S evidence theory and local domain patterns.The main research contents are as follows:Local binary mode(LBP)can better measure the texture characteristics of local images,and its recognition effect is better even under the influence of light sensitivity and different postures.This article first summarizes the local binary mode and its improved algorithm,and then proposes the local neighborhood patterns(LNP)on this basis.First,divide the image code in the face database into some area modules in a certain order,then take the pixel value of each center point and the pixel value of the surrounding neighborhood as a whole,and calculate the average vector of each area,The distance between the neighboring pixel point vectors.Finally,the distance between the vectors of adjacent pixels is used,and finally the method of extracting the ring feature region is adopted,and the nearest neighbor classifier is used as the experimental object.In order to make the LNP algorithm insensitive to light changes,this articl combines the D-S proof theory with the LNP face detection algorithm,convolves the preprocessed image to obtain the horizontal and vertical edge images,and then uses the LNP method to process of extract useful feature vectors,calculate the distance between the sample and each category,and then use the D-S proof theory to merge and make the correct decision.Experiments using the Extended Yale B show that this method can not only improve the robustness under the influence of light,posture and expression changes,but also has a good detection rate.