A New Descriptor for Pedestrian Detection Based on Feature Fusion

Abstract

Pedestrian detection is a significant task with many applications including smart vehicles, surveillance and robotics. Due to the variations in appearance, pose, color, illumination and interference, accuracy and robustness in pedestrian detection are desired to be improved. This paper proposes a new method that introduces Choquet integral to fuse the Histogram of Oriented Gradients (HOG) and local binary pattern (LBP) descriptor in parallel. The fusion descriptor is used as a detector for pedestrian detection with Support Vector Machine (SVM) algorithm. Experiments are carried out on the INRIA Person Dataset and the results validate the efficiency of the proposed method.