A Raster-Image-Based Approach for Understanding Associations of Urban Sensing Data

Torrential rains, the complicated network of roads, and the high density of vehicles contribute partly the number of traffic accidents. In order to understand the correlation between these factors towards building a risk map that can alert drivers of dangerous zones, the visual patterns reasoning system is proposed. By converting sensing data collected from different resources to raster images, the accident patterns can be treated as visual patterns that can conserve spatiotemporal information of events. Deep learning techniques are utilized to build a model based on these raster images towards detecting places with a high probability of accidents. Image clustering is applied to learn a representation of each type of correlations. Thus, the visual pattern of high-probability traffic accidents can be reasoned in the natural language format. The very first result of the ongoing project shows the promising research direction. Currently, only spatial information is stored in raster images and no comparison to related methods is made. In the future, temporal information will be integrated with spatial information to create raster videos towards understanding a correlation of environment, congestion, and human reaction before and after accidents happened. This could help to predict time and place which has a high probability of accidents, and a situation of congestion after accidents happened.