An adaptive video surveillance approach for dynamic environment

This paper presents an edge extraction based automatic algorithm that is suitable for both indoor and outdoor video surveillance, where the background scene as well as its illumination changes frequently. The proposed method focuses especially on achieving robustness through adaptation to the changes in background. We extract pieces of edge information from images and represent these segments as efficiently designed edge classes The reference edge segments are updated with the presence of new objects or the missing of existing objects in the input images for long period. The representation of edge information as segments and the update in reference edges incorporate video surveillance system with acclimation to the dynamic background, high sensitivity to the presence of new objects and robustness to illumination changes. We report on results of our experiment using the framework for detecting motion events for outdoor parking lot.