Camera Tamper Detection Model using Codebook Model
| Autor: | Che-Wei Kuo, 郭哲瑋 |
|---|---|
| Rok vydání: | 2012 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 100 During daily life, people can ensure own safety and property by monitoring system for anytime and anywhere. At present, the most important function of Intelligent Video Surveillance not only can be record scene but also can be analyzed and computed for real-time dynamic image of surveillance. During image captured process, images can be stored in a database by surveillance system at anytime. The fixed cameras are often applied to conduct safety surveillance, but they are occasionally accidentally or maliciously destroyed by external forces, e.g. defocus, fog, spray paint, displacement, obstruction. This paper focus on two security issues to ensure the security of video surveillance systems for the camera tamper detection. The proposed model, Camera Tamper Detection Model (CTDM), is used as a testing method for the camera tamper detection. This model includes three stages, initialization, training and real-time detection. In the initialization stage, CTDM uses Background Codebook Model based on pixels as the background and foreground detection model for video images. According to the color model, both chroma and brightness of pixels are adopted as feature values and they are able to tackle the problem of light and shadow movement. In the training stage, CTDM uses computer memory to establish a short-term memory area storing the sequence of real-time video image samples to segregate the foreground, and then build a good adaptive background model. The model can be applied to the tamper detection of the fixed surveillance cameras. Finally, in the real-time detection stage, the research proposes the warning mechanism of camera tamper through the changes of feature data of foreground pixel brightness, and then classifies the changes to tell whether they fall into Displacement or Obstruction types, in order to ensure the security of video surveillance system. The experimental results show that the accuracy rate of camera tamper detection can reach 100%, either in the indoor or outdoor experimental scenes and the accuracy rate of a tamper type classification can also reach 90%, either in the Displacement or Obstruction types. Thus, this dissertation proves that our proposed methods can be efficiently applied to the real-time detection of the camera tamper. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|