2019/06/06 by Antoine Rimboux, Rimboux, Antoine, Rob Dupre +9
Computer Science · Social Sciences · #Anomaly Detection Techniques and Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Human Mobility and Location-Based Analysis #Video Surveillance and Tracking Methods
paper · pdf · doi:10.48550/arxiv.1906.03994
openalex publication_date 2019/06/06 · openalex created_date 2022/07/23 · openalex updated_date 2026/07/28
Smart video sensors for applications related to surveillance and security are\nIOT-based as they use Internet for various purposes. Such applications include\ncrowd behaviour monitoring and advanced decision support systems operating and\ntransmitting information over internet. The analysis of crowd and pedestrian\nbehaviour is an important task for smart IoT cameras and in particular video\nprocessing. In order to provide related behavioural models, simulation and\ntracking approaches have been considered in the literature. In both cases\nground truth is essential to train deep models and provide a meaningful\nquantitative evaluation. We propose a framework for crowd simulation and\nautomatic data generation and annotation that supports multiple cameras and\nmultiple targets. The proposed approach is based on synthetically generated\nhuman agents, augmented frames and compositing techniques combined with path\nfinding and planning methods. A number of popular crowd and pedestrian data\nsets were used to validate the model, and scenarios related to annotation and\nsimulation were considered.\n