In the recent years, number of surveillance cameras deployed has increased significantly. However it is important that these cameras are functioning as intended and capturing meaningful data. Offenders resort to techniques such as blocking the camera view, using a foreign object to cover the lens, spray painting or de-focusing the camera lens to prevent capture of their images and recording of their actions. These cameras are often unattended or security guard might have lost his focus and if discovery of such an attempt is not immediate, even though there is no failure in the system and the images are still being recorded, these recordings are not useful. In this paper, we propose methods for real-time discovery of visibility loss and covered camera lens using background subtraction in wavelet domain. We also propose some methods to increase robustness of the system in real-life scenarios.
展开▼
