Robust moving object detection based on fusing Atanassov's Intuitionistic 3D Fuzzy Histon Roughness Index and texture features

Abstract

Background modeling is a crucial step in various computer vision applications such as video surveillance, object tracking, and moving object detection. Classifying image pixels as foreground or background is yet a challenging task particularly in complicated situations such as illumination variations, rippling water, camera jitter, and the presence of fast and slow moving objects. Therefore, for a better detection of the moving objects, the multi-modal nature of a scene in those intricate situations should be modeled by multiple models for each image pixel. To this end, in this article, we improve our previous work by fusing color features and texture features using Choquet fuzzy integral. Thereby, our proposed spatial color features that are described by Atanassov's Intuitionistic 3D Fuzzy Histon Roughness Index are fused by the texture features extracted using a covariance matrix. As handling multi-modal background updating is an arduous task, we also propose a new model updating for tackling various challenges such as model initializing with moving objects, existence of fast and slow moving objects in a scene, and existence of the moving objects that stop for a while. We intensively evaluate our proposed approach on diverse benchmark datasets. Experimental results demonstrate the robustness and supremacy of our proposed approach compared to its previous version and the state-of-the-art algorithms in the field.