Sketch Face Recognition With Multi-feature Fusion

In the 21st century,the artificial intelligence and Internet industries are developing rapidly,and machine vision is a branch of artificial intelligence that is rapidly developing.Its applications are in various fields,such as: industry,agriculture,medicine,military,aerospace,etc.Machine vision is the use of machines.Take measurements and judgments instead of human eyes.In recent years,face recognition has been one of the research hotspots in the fields of machine vision,pattern recognition and image processing.It is a biometric technology based on the feature information of face,in identity identification,monitoring system,enterprise and residential.Widespread applications in areas such as security and management,public security and judicial search for fugitives,and self-service.Sketch face recognition is face recognition based on sketching faces.Sketch face recognition is mainly applied in the detection of public security judicial cases.When there is no confirmed photo of the suspect in the case,the comparison between the sketch picture and the face picture can help the investigator to lock or narrow the scope of the suspect.At present,the research on sketch face recognition has received much attention.The early single feature sketch face recognition method uses a single global feature or a single local feature,which is obviously insufficient for the use of feature information.In recent years,the multi-feature sketch face recognition method has become a hot research direction,and great progress has been made in the recognition effect,but there are still many problems,such as the recognition accuracy still has room for improvement.This paper combines global features and local features,and uses multi-feature fusion to sketch face recognition.main tasks as follows:(1)The existing sketch face recognition based on multi-global feature extraction is not robust to changes in facial expressions and poses of images.Excessive dimensionality increases the probability of over-fitting during training.The lack of localinformation in the recognition process;the existing sketch face recognition based on multi-local feature extraction lacks the use of global feature information in the face recognition process.For these problems,a multi-feature fusion sketch face recognition method is proposed.Multi-feature fusion sketch face recognition method,MFSR).The directional gradient histogram features,the local binary pattern features and the block local binary pattern features are sequentially extracted from the image,and the three features are fused by weighting.By calculating the Euclidean distance between the sketch image and the face image,the face map closest to the sketch image is matched.The experimental results show that this method has a slight improvement in the correct recognition rate.(2)For the existing block-local partial binary mode,the same weight is used for all sub-blocks,and the problem of the prominent features of the sub-blocks cannot be highlighted.The partial local-valued features of different weights are used for sketch face recognition.Weighted multi-feature fusion sketch face recognition method(WMFSR1).According to the face recognition cognition theory,humans mainly rely on the structural information and feature information of face in the cognitive process of face recognition.The structural information is the global feature.The feature information is the local feature.The more obvious the local feature,the easier it is for people.Face recognition.Blocking local binary mode with different weights,by dividing the image into several blocks,assigns relatively low weights to most blocks that are background images,and relatively higher blocks corresponding to most human faces.The weights are assigned to different blocks to make it easier to identify between different images.The experimental results show that this method is superior in correct recognition rate.The first method is to extract the direction gradient histogram feature of the whole picture,and obtain the structural information of the whole picture,and does not highlight the structural information of the face part.Therefore,based on the first method,this paper proposes a multi-feature fusion sketch face recognition method.Secondly,the weighted multi-feature fusion sketch face recognition method(WMFSR2)extracts the direction gradient histogram feature,the local direction gradient histogram feature and the partial local binary pattern feature for the face image and the sketch image,and then the three features.Calculate the distance and finally get a matching picture based on the distance.Experimental results show that this method is superior to other methods inrecognition rate.