Research On Algorithm For Traffic Event Detection In Complicated Environment Based On Image Feature

Video traffic incident detection system is the use of image processing technology, combined with artificial intelligence and machine learning academic knowledge of highway surveillance video analysis system to automatically detect abnormal events occur on the highway. It can quickly and accurately detect abnormal events occurring during operation of the highway. It also sends the appropriate information to highway management to minimize the impact of abnormal events on the highway and ensure smooth operation highway. It is an integral part of the highway emergency management platform and intelligent transportation system for the front information collectio. With China’s increasingly congested highways issues, video traffic incident detection system is gradually been widely used, and therefore research on this technology is very meaningful. Combined with the present situation of highway monitoring system, we analysis the existing problems in the monitoring and control system. We proposed to solve some common problems appeared in current video detection system. We focus on improving the detection rate and environment stability of the video detection system. Firstly, we need to solve is the environmental adaptability of the monitoring system. Because highway monitoring environment are mostly open-air setting, monitoring environment is always complicated and changable. This is a big chanlenge for our monitor and control system. Environment adaptability must be required. Through a number of highway-site environmental analysis and data collection, we propose a new classification algorithm to recongnition day-night switching, rainy surface water environment, rainy night street reflective environment and other harsh environments. Secondly, we use background subtraction technique to detect the traffic objectson the highway. In the context of differential binary target detection technique, we propose a new adaptive background subtraction threshold algorithm to adapt the highway complex monitoring environment. Meanwhile, we take special recognition algorithm for harsh monitoring environment to ensure the accuracy and stability of the detection system. Finally, we track and analysis the objects that acquired by the detection system. We constructed a Kalman predictor that adapt to track the moving objects appeared in the highway monitoring environment. We analysis event features or characteristics of the monitor screen of various types of abnormal traffic moving target. According to machine learning algorithms, we defined the characteristic parameters of various common traffic events to ensure accurate detection of all types of abnormal traffic incident.Through simulation tests conducted in a domestic highway, we tested the traffic incident detection algorithm of this paper has good performance and can accurately detect common types of abnormal events in several complicated environments.