Scientific publication

Airborne based High Performance Crowd Monitoring for Security Applications

Publication from Digital

Roland Perko and Thomas Schnabel, Gerald Fritz and Alexander Almer, Lucas Paletta

Scandinavian Conference on Image Analysis (SCIA) , 1/2013


Crowd monitoring in mass events is a highly important technology to support the security of attending persons. Proposed methods based on terrestrial or airborne image/video data often fail in achieving sufficiently accurate results to guarantee a robust service. We present a novel framework for estimating human density and motion from video data based on custom tailored object detection techniques, a regression based density estimate and a total variation based optical flow extraction. From the gathered features we present a detailed accuracy analysis versus ground truth information. In addition, all information is projected into world coordinates to enable a direct integration with
 existing geoinformation systems. The resulting human counts demonstrate a mean error of 4% to 9% and thus represent a most efficient measure that can be robustly applied in security critical services.