Using Biometric Features on Long Distance Videos for Accurate Pedestrian Age Classification
Pedestrian detection is popular topic currently, and there are various studies on image and video based pedestrian detection. After detection, it can be more specified as classification, tracking, etc. Classification of child and adult is useful for the social security, because the number of social crimes is increasing. The goal of this study is classification of pedestrian in two categories as child and adult. In this study, Haar cascade classifiers are used. At first, the full body and head area of the pedestrian are detected. Next, biometric ratio between height and head of pedestrian is calculated. Then, age is classified by comparing the biometric ratio between height of head and whole body. Experimental results showed that proposed biometric feature has a significant discriminant value for distinguishing the child from the adult on long distance videos. The accuracy rates are 74.7% for adult, and 68.1% for child.
This work is licensed under a Creative Commons Attribution 4.0 License.