Occlusion detection based on fractal texture analysis in surveillance videos using tree-based classifiers

Abstract

Occlusion detection in video has been an active research for decades. This interest is motivated by numerous applications, such as visual surveillance, human-computer interaction, and sports event analysis. In this paper, an occlusion detection approach based on fractal texture analysis is proposed. Texture features are extracted from the segmented images using Segmentation-based Fractal Texture Analysis (SFTA) algorithm. The experiments are carried out using a PNNLParking- Lot dataset and the various tree-based classifiers such as random forest, random tree, decision tree (J48), and REP tree are used for classification. In the experiment results, random forest classifier showed the best performance with an overall accuracy rate of 98.3% for SET- 1, 98.2% for SET-2, and 83.7% for SET-3, which outperforms other algorithms.