Research On Novelty Detection For High-Speed Railway

Railway is the main transportation in our country. To improve the capacity of railway transport, high-speed railway has been greatly developed in recent years. As a result of the imperfection of the railway infrastructure, there are many railway accidents at present. How to improve the situation has been a public concern.In this thesis, we will study the method of novelty detection. With the monitoring video, we can process it and figure out the models of background, finally detect the novelty in the railway track. The main contributions of this thesis include the followings:1, We designed a new color feature based the RGB color space, and used Gaussian Mixture Model and Bayesian Model with this color feature to detect novelty detection. Comparing these solutions with traditional color features, our method achieved a better performance.2, We improved the texture-base background model. If we just use the texture-base background model, there will be too much noise, and high computational complexity. After dividing the image into several regions and using background models, our solution can easily and quickly detect the novelty and be insensitive to noise.3, We designed and implemented the prototype system, and it has a very good performance.