As part of its continuing task of monitoring UK road traffic conditions, the Department of Transport performs a traffic census, referred to as the Core Census, at a number of representative points on the UK road network. The automatic system collects statistics relating to vehicle flow, speed and type of vehicle by using a configuration of inductive loop, tribo-electric and piezo-electric sensors. However, under dense traffic conditions, the performance of the automatic system deteriorates, because the presence of successive vehicles over the inductive loop sensor can overlap. It is therefore worthwhile investigating the use of video information, processed automatically by computer, to supplement or replace the surface sensors. This report reviews the various ways in which image processing techniques have been applied to traffic monitoring and establishes the state-of-the-art in each case. The approaches considered have been classified into: loop emulation; incident and congestion monitoring; vehicle tracing; and number plate reading. The report contains a detailed discussion on vehicle recognition and classification by means of image processing, mentioning the use of neural network and artificial intelligence techniques. In particular, the use of three-dimensional models is discussed. Recommendations are made for particular approaches to allow such image processing techniques to be developed for application within the UK road traffic core census programme. This research was carried out under contract for the Traffic Operation Division of the Traffic Group, Transport and Road Research Laboratory TRRL.
Abstract