Vehicle Traffic Analysis Using Yolo

Abstract

Traffic density specifically in the crowded urban areas is at an all-time high. It
requires highly accurate and fast traffic analysis systems for capturing data to produce
insights and for surveillance purposes. The data of vehicle traffic collected over a time
period can be used to find traffic density patterns and procure insights which can be used
for improving the traffic management. Existing hardware-based techniques for traffic
analysis include magnet based loop detectors embedded inside the road provide useful
data, but also has a significant downside: physical damage over a period of time, which
reduces their functionality and accuracy. Even most of the software based techniques
perform well to an extent, however they can only detect moving vehicles. To solve this
issue, in this paper proposes to use a convolutional neural networks based algorithm
known as You Only Look Once (YOLO). This paper proposes to create an end-to-end traffic
analysis system which can take video as the input, process the video using YOLO algorithm
and produce the output report using which insightful analysis can be obtained. The data is
obtained from a surveillance camera to evaluate this model.