DEVELOPMENT AND EVALUATION OF A TRAFFIC MEASUREMENT SYSTEM USING HOG FEATURE AND SUPPORT VECTOR MACHINE

Abstract

 The purpose of this research is to develop a low-cost traffic flow measurement system. To achieve the objectives, we have utilized vehicle detection technology based on HOG and SVM. We have experimented traffic flow measurements on the multi-lane public road using the developed system. The results have been compared with the background subtraction method and YOLO (You Only Look Once) that is based on convolution neural network. Our system has demonstrated both real-time performance and better false detection and undetected rates.