Bayesian Classification of Moving Objects Using Adaptive Threshold Approach

Abstract

Object identification and tracking has become essential in areas like aerial image analysis and monitoring in crowded public places. An Active Contour and Color Recognizing approach were developed for detecting and tracking the object in sequential images. However, the presence of occluded and dynamic shadows of objects was a challenging issue in foreground segmentation. Therefore, the proposed method introduces object segmentation based on threshold values chosen using multiple texture features of objects. Next, the research work introduces spatial classification of objects using multi-classifier Bayes algorithm. This integrated approach improves the segmentation and classification accuracy with reduced time. Experimental results reveals the effectiveness of the above said methods with respect to evaluation metrics like segmentation time, peak signal-to-noise ratio, classification accuracy, false positive rate, classification time and object detection rate.