Study On Traffic Flow Detection And Prediction Based On Neural Network

Aiming at the existing problems of traffic detection and prediction,this article does many detection methods are studied for moving targets, selected real-time strong, fast and high practicability background subtraction method for the detection of moving targets. This article proposed a new method of background updating and reconstruction, which uses frame difference and background methods combined approach to the reconstruction of the background, using pixel-level update Frame-level multi-level background method, experimental results show that the method can achieve fast and accurate reconstruction of the background and updates. This article proposed a new method for the removal of shadow detection,which is based on HSI's,a quick way----fast, accurate and effective method, and real time. This article used the virtual induction coil method counts detected in the movement of vehicles.the result shows this method can provide a more accurate reliable traffic data for the detection of the back, in which the forecast traffic, Using BP neural network forecast traffic intersection traffic flow.And the results are satisfactory meeting the demand for transport induced to do some theory and practice.The results of this study indicate that video-based flow detection method has lots of advantages,such as high accuracy, rapid,real time and so on. Traffic prediction based on BP neural network has high accuracy to meet the application requirements.