Novel Approach by Fuzzy Logic to Deal with Dynamic Analysis of Shadow Elimination and Occlusion Detection in Video Sequences of High-Density Scenes

Abstract

Monitoring of high-density images from video sequences provides an important potential for crowd detection and classification. In fact, shadow presence from video sequences causes detection failures of results or mistakes in interpretation. In this contribution, we present an automatic system to deal with shadow elimination based on extracting the vector size of the movement and detection of occlusion management with the Fourier series approach because of the position and orientation of the camera, speed magnitude and visual tracking of crowd scenes, and mathematical morphology of discrete data in a non-linear approach. The model consists of distinctive real objects from fused data and crowded scene caused by shades, which often has consequence such as the failure counting and grading of vehicle and people, while dealing with traditional methods. As we reveal, our technique is principally appropriate for UMN and PETS data to eliminate shadow nuisance and detection occlusion with quite good performance. For classification, we use two classes in each image, for each category of events detected by fuzzy logic. Although this comes down to easy system modeling, as it relates to the use of fuzzy rules. The provided results advocate in favor of our method in terms of effectiveness and precision. Indeed, the proposed approach provides the ability to segment more specific objects, such as people and vehicles in real-time space. The results were compared to other methods, namely Covariance Matrices for Crowd Behavior, Social Force Method, and Ground Truth Technique.