Intelligent unmanned aerial vehicles surveillance for detecting abnormalities in crowds under occlusion and multi target conditions

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Analysis demonstrates 98.7% F1-score in crowd surveillance with hyperparameter adjustment, highlighting real-time performance.

Anomaly detection achieves 98.7% F1-score with a 0.03 MAE using the VisDrone dataset, showcasing effective crowd monitoring.
Noise reduction is maximized between 83% and 91% in occluded environments, ensuring reliable detection amid challenges.
The approach integrates YOLO and Generational adversarial Occlusion Networks for robust performance under tight constraints.
Future work emphasizes parallel processing to enhance algorithm adaptability and efficiency in large-scale monitoring.

Ye et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark