Identification And Behavior Analysis Of Children In The Wild

The main goal of this paper is to apply computer vision technology to the construction of smart kindergarten,to help preschool teachers understand children better,and to reduce the work burden for preschool teachers,so that teachers have more time to spend on teaching-research activities.According to the idea of identifying children first and extracting their activity information later,this paper carried out a series of studies on the identification and behavior analysis of children in the wild.The main contents and results of this paper are as follows:1.In view of the low accuracy of face recognition in natural scenes,this paper proposes three methods to improve the effect of face recognition.First,evaluate image sharpness by Fourier transform;second,filter side face by face orientation;third,optimize the image illumination by the image histogram;and the test is carried out on the activity data of children collected in wild,which improves the accuracy rate of face recognition from 65.34%to 91.43%.2.Based on the above,to solve the problem that identity recall factor is low when use face information alone in the wild,and the identification accuracy is not high when use body information alone in the wild,this paper proposes a multiple modal identification method,and test three different fusion methods.Finally,we propose a method that identification method based on dynamic updating benchmark.This method combines human face information with human body information and makes full use of the features of face recognition and pedestrian re-identification.Under the condition that the identity recall factor is close to 93.51%,the identification accuracy rate is improved to 89.85%.Compared with the use of face information alone,the identity recall factor of this method increased by about 81%,and the identification accuracy rate increased by about 26%when compared with the use of pedestrian re-identification alone.3.Based on the above results,we extract the activity information of each child and visually presented it to the preschool teachers to assist the teachers in teaching.