Video generates a rich set of image information and often the useful information is only a very limited set of the available information. Another well-known fact is that visually reviewing of long video recordings is a time demanding task. In combination with the continuously increasing number of video surveillance systems, this leads to an increasing need for automated analysis of long image sequences. The goal for this work is to develop and evaluate a method for automatic detection and tracking of events recorded onto a surveillance video, such as appearance of persons or vehicles in a surveyed area, to evaluate the usefulness for forensic applications and real time applications. One core problem is the fact that both the background and the objects move, where only the physical motion of moving objects are of interest and needs to be separated from the camera motion. Another core problem in many of the video processing algorithms is parameter estimation despite invariance for accurate modeling of the desired features. Varying scale, color, lightning conditions and occlusion of the object of interest can for example cause invariance. The technical approaches for this work is to separate object invariance, all pixels are considered independently and no feature parameters are needed. If the basic optical flow constraint is satisfied, the motion is classified as global motion. If not, the motion is considered caused by local motion, noise or other phenomena. An object that undergoes local motion can then be detected and tracked as is forms a trace in the temporal domain, while the noise appears on an intermittent basis and will be disregarded. The results from applying this method on several image sequences were compared and the robustness and ability to deal with invariance has been evaluated. The result clearly shows that in realistic situations, where visual reviewing can be quite a difficult task, computer based methods for automatic detection are useful to detected moving objects in long video recordings.
展开▼
