Vdieo Anomaly Identification Based On Pixel Analysis

As an important tool for modern traffic monitoring, the research and development of video abnormal identification technology has made significant progress, but the recognition accuracy and efficiency can not meet the requirements of many practical applications, there are still some problems to be resolved.This paper focuses on the algorithms of video anomaly identification and put forward a number of effective solutions. This paper first make a research on dynamic objects extraction in the video. Different from the research mainly rely on the background frame difference and dynamic extraction in the past. This paper proposed two-way frame difference method combined with the concept of the background updating. Solving the problem of high error caused by background frame difference and dynamic extraction method.During the stage of anomaly identification,the traditional methods first track the object (pedestrian and vehicle) and distinguish them,then make a judge that the track is normal or not. As the traffic volume growing in mordern city life,this method increases the computational efficiency of the computer and unable to meet the complex transport needs,because this method is too complex. This paper presents a method based on the background pixels. We call this method contexts and behavioral method.First, we present the method of background under the two models (Hidden Markov Model and Co-Occurences Model) and then we used two models to extraction video anomaly pixels.Finally, we introduced a imitation three-dimensional object recognition method,after extraction abnomaly pixels based on the contexts and behavioral methods,we use this method to detect the target object (vehicle).Compared with the trajectory method, the experiment result shows the time spending in this algorithm is shorter than other algorithm and this algorithm has a better application in transportation system.