Enhanced Approaches to Detect Track and Classify Objects for Video Surveillance

Abstract

Major objectives To propose object background subtraction techniques that develop hybrid models to obtain the benefits of existing algorithms along with a model that takes advantage of newlinetransformation technique to effectively identify objects of interest that is significantly different to the background in a video sequence newlineTo propose object enhancement techniques that perk up the quality of detected objects in three manners Noise removal using enhanced morphological filter Illumination and lighting variation correction using enhanced reflectance model and Shadow removal algorithm using cluster based cast shadow and estimatorbased newlineself shadow algorithms To design an enhanced voting based object tracking techniques by associating target objects in consecutive video frames over a time period To propose enhanced object classification techniques based on enhanced silhouettebased newlinemethod and enhanced Support Vector Machine classifier to automatically recognize the tracked moving objects and group them into four categories namely human human group vehicles and animals