The Application Of Side Face Recognition Based On Minimum Spanning Tree

The biological recognition is regraded as hot research topic between pattern recognition and computer vision technology,it provides people with a highly stable and highly reliable authentication mode in the 21 st century information age.While the face recognition is known as the focus of bio-metrics technology research because of its advantages,for example it is easier to get,and has the uniqueness of the individual and not easily felt by people when its characteristics are collected.And with the rising of artificial intelligence technology,it makes the research on facial recognition technology be more and more valuable and the range of applications be wider,such as the system of control access intelligently and the system of intelligence proctored network exam and so on.However,there still is a problem need to be solved if the face recognition wants to be used widely for the network test system now,that is,the images that we acquired in the network exam not only have positive face images,also include side face images of various angles,because the monitoring cameras captured images of the candidates is random,and there are color and shape distortion.So a side face recognition method based on the minimum spanning tree combined with knowledge of graph theory and information theory was proposed in this paper,against the color and shape distortion,achieving identification of candidates.The main works are as follows:Firstly,random side face images could be acquired by the cameras(considering the side face angle of the candidates on online exam system could not be too large,otherwise they will be considered cheating,thus images obtained are smaller angle),and then these obtained images were pre-treated to improve quality.Then through the establishment of skin model,the color region segmentation was detected from a complex background,finally the right side face area was identified using an external matrix screening method based on the minimum pinpoint.Secondly,the image was divided into few blocks by facial features were located according to these features to describe the overall invariant structure of the side faces better,and then the sift(scale invariant feature transform)algorithm was used to find the extreme points of each region in the image.These points were filtered as that having an identification value as the feature points of side face.At last these feature points weremade use of building the minimum spanning tree.It would be integrated into gray scale information and location information of feature points to joint Renyi entropy,the minimum spanning tree was taken advantage of estimating Renyi entropy,finally the result was got by measuring the degree of difference between the two faces according to its value,thus achieving recognition of side faces.The sift algorithms is achieved in the field of face recognition algorithms,and the method and experimental results are compared to it at the side of the accuracy and recognition rate algorithms.That this algorithm has a good prospect is proved by all of these.