Crowd Estimation of the Black Nazarene Procession in Manila, Philippines

Abstract

Developments in crowd counting and estimation are currently in demand brought about by the increasing population and corresponding improvements in public safety planning. Various crowd counting methodologies are successful in addressing challenges and limitations, such as occlusions and wide density variations. However, none have tackled the risky dynamic scenario of the annual Black Nazarene Procession in Manila City, Philippines. Extreme densities are reached as participants follow a moving subject. Yearly reported crowd estimates vary greatly as estimation methods used for this event remain undisclosed, undefined, or unpublished. Considering the strengths of both detection-based and regression-based crowd counting methods, a novel pedestrian estimation method is proposed to appropriately provide an accurate pedestrian estimate. Using video graphics, a static grid analysis is performed to systematically capture and evaluate actual participant density. From the recorded pedestrian densities, functions were developed to estimate densities at varying distances ahead and behind the carriage. A pedestrian joining density function was established to account for devotees who merge with the crowd way ahead of the procession. The 2019 event involved a moving carriage faithfully followed by thousands of devotees as it traveled along a 6.94-km route for 21.35 hours. Employing a 95% confidence interval for the density function intercepts and coefficients, 176,086 to 484,215 active pedestrian devotees during the procession were estimated using the developed systematic pedestrian estimation method. The social value of a more accurate crowd estimation method lies in providing policymakers with reliable crowd estimates that will enable the authorities to deploy the appropriate number of security personnel for crowd management, as well as medical staff for emergency situations during mass gatherings of similar nature to the Black Nazarene Procession.