Anomaly detection for intelligent Video surveillance using machine mLearning techniques

Abstract

The video-based surveillance system is playing an efficient role in transportation system. Well organized association is required between human and machine in order to decipher valuable information from big, multidimensional and diverse data set. These systems capture traffic scenes, evaluate the information obtained and finally differentiate between normal and abnormal activities. The abnormal activities are difficult to find even by a trained human operator and the task of watching screens continuously becomes burdensome. As a result, performance of human operator degrades significantly. To assist human operators, the well-defined automatic anomalies detection system has been developed for making the system effective in the recent years. However, the accurate detection of abnormal events remains challenging problem in traffic surveillance due to high traffic, occasional background change, overtaking by vehicles, etc. The main scope of this research work is to provide a way of better detecting video surveillance anomaly for classification of the Vehicles movement. To improve the classifier accuracy, a new novel combination of classification technique is proposed for anomaly detection in intelligent video surveillance. So, in the initial proposal a Markov-Modulated Poisson Process (MMPP) for anomaly detection and classification in intelligent video surveillance is proposed. In addition, a discrete-time model with the poison process has been introduced to observe the superposition of normal behavior and the event behavior which may increase or decrease the number of counts observed. The event behavior is detected using a Markov chain model in order to capture the idea of event persistence. After the anomaly detection, the genetic algorithm has been utilized in this work to improve the classification newline