Background model based on fuzzy ART with forget mechanism and traffic states recognition

Traffic state is a key parameter for traffic guidance system, which has been considered as a promise way to relieve the traffic congestion. This paper presented a method to acquire traffic state information covering large road net by analyzing traffic surveillance video. A new background model was proposed which is based on fuzzy ART and forgetting mechanism to solve the problem in this domain. A learning quantization vector (LVQ) was used as the classifier to map the traffic parameters into a certain class. According to the test results, the accuracy produced by the suggested method is 91.5%, and this means that the method is feasible and practical.