STRAIGHT TO VIDEO? AUTOMATED VIDEO ANALYSIS AS A TOOL FOR ANALYZING ROAD USER BEHAVIOR

摘要

A university study that explores the issues surrounding how to extend the use of automated video analysis. Video analysis has become a rather common tool for data collection in traffic. Various instruments based on the technique, such as traffic counters, systems for detection of incidents (congestion, stand-still, appearance of a physical obstacle on the road, abnormal behavior such as driving in the wrong direction), or vehicle identification (license plate readers) are available on the market, or are being developed by traffic research centers for their particular purposes. There are, however, some common limitations of these systems. First of all, most of them are entirely focused on detecting motor vehicles, while other categories of road users are ignored. This is explained by the other limitation: often the systems are designed for operation in relatively simple traffic conditions, such as motorways or tunnels, where pedestrians and cyclists do not appear at all. In the trade-off between processing time and complexity of the analysis, short processing time is given priority, which results in a limited number of performed tasks that in addition are fairly simple.