Traffic Monitoring without single Car Detection from optical airborne Images

Abstract

This article describes several methods for traffic monitoring from airborne optical remote sensing data. These methods classify the traffic into free flowing traffic, traffic congestion and traffic jam. Furthermore a method is explained, which provides information about the average speed of dense traffic on a defined part of the road. All methods gather the information directly from image features, without the use of single vehicle detection. The classification of the traffic is done by stacking at least three overlapping images on top of each other and calculating the standard deviation of the gray values of each overlying pixel. In addition to that a texture analysis is implemented to differentiate the traffic. The average speed of dense traffic is calculated employing disparities of two following images of the same scene. All methods, which were presented in this article were tested on various data sets and compared with interactive measured reference data. These methods will be applicable in combination with methods using single car detection in the ARGOS-Project (AiRborne wide area hiGh altitude mOnitoring System).