Pedestrian Detection with Multiple Classifiers on Still Images

Abstract

In this work, an algorithm that detects pedestrians in still images using different classifiers is presented. HOG, which is frequently used in pedestrian detection, and support vector machine (SVM), K nearest neighbors (KNN) and AdaBoost algorithms were used as descriptors. It is decided whether the image is pedestrian by looking at the result of three different classifiers. In order to demonstrate the effectiveness of the method, the system is trained using the INRIA data set and tested by using Penn Fudan Pedestrian Dataset which is different dataset. Experimental results show that the proposed method detects higher accuracy than pedestrian detection using a single classifier.