Authors:Â S.P. Suganya Shri, Dr. L. Latha
Abstract:Â
Foreground extraction in a dynamic environment is a challenging task during video surveillance. In computer vision, evaluation of movement is a major task for which only the changes in the scene are required. While there are a number of techniques available to analyze the video sequence in real time using parametric model, here we use non-parametric model to extract foreground using coarse-to-fine detection theory, after which the performance is measured using various formulae. Later, SILTP algorithm is used to detect the object and track them. Foreground objects are extracted using binary descriptor based background modeling method. Foreground and background instances are represented using galaxy descriptor. In region and pixel levels, the detection theory is applied to obtain rough and detailed shapes of foreground objects. Illumination changes and dynamic background problems of background modeling are dealt in this method. In this paper, we propose a technique to focus on foreground tracking and event analysis in background detection. Performance is measured using recall, precision and f-measure.
Full Issue
| View or download the full issue |