Answers are grounded in this paper and cite their sources.
Hybrid ensemble learning improves crowd behavior recognition by integrating optical flow and deep motion features, pointing to enhanced surveillance outcomes.
Crowd Behavior analysis has become a crucial component of modern video surveillance systems, enabling automatic detection of abnormal events such as panic, congestion, and violence. Traditional approaches often fail to generalize under complex environmental conditions, while deep learning methods alone require large datasets and extensive computation. This paper proposes a hybrid ensemble learning framework that integrates optical flow–based motion features with deep motion representations extracted from convolutional neural networks (CNNs) to achieve real-time and robust crowd behavior recognition. The ensemble model combines Random Forests (RFs), Gradient Boosting (GB), and a lightweight CNN classifier via weighted voting. Experiments conducted on benchmark datasets, such as the UCSD Anomaly Detection Dataset and Violent Flows (VF), demonstrate that the proposed framework outperforms individual classifiers and state-of-the-art deep models in terms of accuracy, F1-score, and processing speed. The results confirm that ensemble learning effectively bridges the gap between handcrafted motion cues and deep spatio-temporal representations for practical surveillance applications.
No takes yet. Share an insight, caveat, or question.
Mabrouk et al. (2025) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: