Pedestrian behaviour and exposure to risk

Traffic conflict studies are generally performed using a combination of manual surveillance and laboratory analysis using video imagery. A comprehensive assessment of the potential causes of road incidents requires a large quantity of data to be captured, enabling site-specific pedestrian and vehicle interactions to be monitored during different periods of activity. Therefore, there is a demand for the development of an automatic conflict identification system capable of extracting image sequences of interest. Such a system will provide transport engineers with a superior means of data capture, enabling a more effective and efficient form of manual conflict analysis. The authors outline the ASSET-2 system, developed by the Defence Research Agency, for the automatic assessment of pedestrian movement. ASSET-2 system can track numerous, variably moving objects in real-time and handle inevitable occlusion effects effectively. ASSET-2 (scene segmenter establishing tracking) system, extracts object movement parameters using the principle of feature-based image motion estimation. ASSET-2 derives feature information using the SUSAN algorithm which locates corners by deriving regional measures of grey scale homogeneity