The Research And Application Of Markov Random Field In Machine Learning

As a major branch of Probabilistic Graphical Models,Markov Random Field has been widely applied in machine learning.In this theory system,the Hidden Markov Model is a powerful tool for many great applications in Intelligent Transportation System,such as the video-based traffic anomaly detection.This paper contains two tasks.The first task is to design a framework of solving machine learning problem using HMM,combining the theory of MRF and the requirement of machine learning task.Secondly,the framework is applied into the Intelligent Transportation System,and a traffic anomaly detection method based on Hidden Markov Models is presented,which is capable of real-time detection of abnormal events in traffic surveillance.We propose a traffic anomaly detection algorithm based on the Continuous Hidden Markov Model.Our method consists of three parts: First,feature extraction.We extract the optical flows in each frame of the video that is segmented into various parts.Then we perform coordinate transform on the optical flows to get the standard features.Second,CHMM training.The parameters of the HMM-GMM model is trained via training samples.Third,anomaly detection.We compares the likelihood of the test sample with a threshold,which is derived from the trained CHMM model.The experiment proves the effectiveness of our proposed method,meets the requirements of high accuracy,real-time performance,adaptability of multiple scenarios,low training cost,etc.