Detection Of Defective Videosurveillance Camera In Train Stations

In the past decade, imaging sensors have known noteworthy improvements and a growing interest from both academics and industries. In particular, videosurveillance cameras have widely spread across the world, and many systems depends on their inputs. Either to provide visual information to human operators or to give input data to Artificial Intelligence algorithms, it is important to be able to assess the proper operation of sensors and the quality of images and videos they record. However, existing methods are either very specific (detect only limited kinds of default) or perform poorly in unpredictable environment. Artificial Intelligence algorithms have proven to be useful to detect anomalies and could be used to highlight a wide range of defaults, including noise, obstruction and intentional tampering such as knocking the camera out of alignment. In this work, the authors present a comparative study for defective camera detection, in the context of videosurveillance.