الفهرس | Only 14 pages are availabe for public view |
Abstract This research is going to introduce an algorithm to detect moving ob- jects from video sequence captured by a static camera. The detection and extraction of moving objects from frames of a recorded video sequence is widely used in many fields such as: computer vision, vid - eo surveillance and traffic flow monitoring. Robust tracking and clas- sification of moving objects in the scene could be achieved by having reliable and effective detection unit. There are many challenges in developing good detection algorithm. Firstly the algorithm must be robust against illumination changes. Secondly the algorithm should avoid detection of non-stationary backgrounds (swaying trees, leaves, rain and snow). Finally the algorithm should be quick in adapting to stop and start of moving objects in the scene. Therefore high preci- sion and computational complexity issues are very important while trying to choose an algorithm for a particular environment. Three dif- ferent motion detection techniques are implemented and compared. These are namely, traditional frame difference, background subtrac- tion (traditional and selective running average model) and hybrid technique. The objective of the proposed research is to design and implement a robust moving object detection algorithm based on combination of adaptive frame difference and background subtraction techniques to overcome the weakness existing in both traditional frame difference and background subtraction techniques. The pro- posed hybrid algorithm is tested qualitatively by using different five video sequences and quantitatively by three different video se- quences. The proposed algorithm outperforms other commonly used object detection algorithms. |