A Wide Area Multiview Static Crowd Estimation System Using UAV and 3D Training Simulator

Crowd size estimation is a challenging problem, especially when the crowd spread over significant geographical area. It has applications in monitoring of rallies and demonstrations calculating assistance requirements humanitarian disasters. Therefore, accomplishing surveillance system for large crowds constitutes issue. UAV-based techniques are an appealing choice region, but they present variety interesting challenges, such as integrating per-frame estimates through video without counting individuals twice. Large quantities annotated training data required to design, train, test system. In this paper, we have first reviewed several techniques, existing simulators sets available analysis. Later, described simulation provide data, avoiding need tedious error-prone manual annotation. Then, evaluated synthetic from simulator using various single-frame techniques. Our findings show that simulated can be used train estimation, thereby providing suitable platform develop We also propose automated 3D approximately static or slow-moving crowds, public events, political rallies, natural man-made evaluate results by applying our new framework scenarios with varying sizes. The proposed gives promising widely accepted metrics including MAE, RMSE, Precision, Recall, F1 score validate results.