Fujitsu Laboratories Ltd. and Fujitsu Research and Development Center Co., Ltd. are developing new technology to analyze traffic video in order to provide real-time information on congestion, accidents and crime violations.
Using machine learning and image processing, the technology analyzes the images from surveillance cameras installed along highways and streets, and groups characteristics that can lead to recognition errors, such as changes in lighting or environmental factors like nighttime and fog. The technology also analyzes moving objects, such as vehicles, bicycles and people, to identify accidents.
A comparison of the previous traffic camera technology (left) and a sample of the application Fujitsu is testing (right). Source: Fujitsu The goal is to improve the way surveillance cameras can be used to improve traffic safety, reduce pollution and reduce congestion. In China in particular, the quick rise in the rate of car ownership has increased the number of traffic violations and fatal accidents. This car ownership has also led to higher levels of congestion, chronic traffic jams, noise, air pollution and other issues. Fujitsu believes video camera analysis is one way to help solve some of these issues.
Fujitsu ran field trials on the technology in cities in China in collaboration with the Tsingha University Suzhou Automobile Research Institute. In these trials, 11 types of incidents of interest—traffic accidents and violations—were recognized with accuracy levels of 90% to 95%.
Fujitsu says that even in cameras that do not have image correction features, the technology can be used to deliver a low-cost monitoring system that can assess traffic conditions, apply traffic-flow controls and produce analysis to reduce congestion or rapidly respond to accidents.
