HOG Based Pedestrian Detection System for Autonomous Vehicle Operated in Limited Area

Abstract

Research on autonomous vehicle is growing rapidly in recent years. Its ability to navigate without solely depending on a driver enables various applications from daily transportation to high risk expedition. In the navigation system of autonomous vehicle, pedestrian detection plays a fundamental role to avoid accident causing human fatalities. In this paper, we propose a pedestrian detection system using the Histogram of Oriented Gradient (HOG) and Support Vector Machine (SVM). This system is designed for use in limited campus area i.e. roads connecting campus buildings. We collected 2000 image samples of roads with or without people passing by. We used 90% of those samples for training the model, while another 10% was used for testing. The model is able to distinguish the number of people on the road in the field of view from zero to four people with the accuracy of 98.00%.