Published on July 1, 2020.

Crowd Source

Someday soon, crowds will once again be a defining feature of the human experience. When they are, facility managers and safety officials can utilize a new tool developed by NFPA to help them evaluate and manage potentially dangerous crowd dynamics.


BY VICTORIA HUTCHISON

  
Large crowds can present some of the most complex challenges faced by safety officials, crowd managers, and facility owners. Crowd management has been a long-standing life safety concern for both fire-related and non-fire emergencies in an array of venues, including sports facilities, concert halls, clubs, malls, and fairgrounds. Crowd dynamics can lead to trampling incidents, crowd crushes, and violence, and when these events combine with insufficient means of egress and ineffective crowd management, injuries and deaths can occur, sometimes in staggering numbers.

But what if crowd managers and authorities having jurisdiction (AHJs) could evaluate crowd dynamics with a real-time crowd monitoring system? What if they could use detailed, data-informed situational awareness to identify rapid changes in crowd density, movement, and other behaviors and neutralize potentially dangerous situations?

Those are the aims of an ongoing Fire Protection Research Foundation (FPRF) project on data-informed crowd management. Launched last year in collaboration with the NFPA Data and Analytics Group, the goal of the project is to develop a low-cost, open-source, proof-of-concept tool that can be used by safety officials to improve their real-time situational awareness of crowds. Funded by the National Institute of Standards and Technology (NIST), the crowd density tool will be capable of collecting, analyzing, visualizing, and reporting crowd data extracted from still images and video feeds. It is not intended to replace safety officials; instead, the tool can be used as part of the event planning process, and during live events to support crowd managers as they make timely, informed decisions.

As a research project manager at the FPRF, part of my role is to listen and respond to the needs of our stakeholders. As one of the lead researchers on this project, it is exciting to participate in a cutting-edge effort that is a direct response to requests from safety officials. This research combines the latest advancements in data science with commercial off-the-shelf hardware and open-source software to provide reasonable and accurate solutions for the crowd-safety community. The tool has the potential to deliver a high-quality solution that is affordable and easily implemented.

RELATED: Read Hutchison's first-person account of a scary situation in a large crowd 
 
The project is expected to conclude this fall, and we have developed an initial proof-of-concept version of the tool. Our next steps will focus on testing the tool on an array of venues, validating the estimated crowd counts, and evaluating the feasibility of full-scale implementation. In October, the FPRF expects to release a final report and the programming code associated with the tool, along with guidance on how to implement it. The coronavirus pandemic has reduced or eliminated crowds around the world for the past several months, but they’ll return soon enough—and when they do, safety officials and facility managers will have an effective new tool to help them meet the challenge.

The hazards of crowds

The problem of crowds does not lie with the crowd itself, but rather in our incomplete understanding of its anticipated behavior and limitations in our ability to quickly adapt to that behavior.

Many variables go into that behavior, starting with the venue itself—each one has its own vulnerabilities and challenges. Similarly, every crowd has its own character, influenced by internal and external factors such as purpose, organization (or lack thereof), and emotional engagement. Crowd dynamics are influenced by time, space, the information available to people, and the group’s collective energy. The inherent complexity and volatility of these events underscore the need to provide crowd managers and other safety officials with accurate and relevant information in real time.
 

DEADLY CROWDS Top: Fans are pulled to safety during a crowd crush incident that killed 96 people during a soccer match in England in 1989. Center: A crush incident killed 21 at the Love Parade festival in Germany in 2010. Bottom: At least 2,000 people died in a stampede and crush incident near Mecca, Saudi Arabia, during the annual Hajj in 2015. Photographs: Getty Images
 
History is filled with examples of volatility boiling over into chaos. In 2015, during the annual Hajj, the Islamic pilgrimage to Mecca, Saudi Arabia, a stampede and crowd crush incident in the nearby city of Mina resulted in the deaths of at least 2,000 people—some estimates put the toll at more than 2,400, with at least 900 people injured. In 1991, nine spectators died and 28 were injured in a crowd crush at a college gymnasium in New York City. The event, a charity basketball game featuring celebrity rap stars, drew thousands and turned deadly when fans surged toward a single entryway at the bottom of a stairwell. In 1989, a crowd-crush incident inside Hillsborough Stadium in Sheffield, England, resulted in 96 deaths and more than 760 people injured. Inquests found that negligence among the police and ambulance services, as well as design features of the stadium, were among the factors responsible for the disaster.

These are the types of painful lessons that form the backbone of the regulatory requirements for managing large crowds in NFPA 101®, Life Safety Code®. According to the code, crowd managers are responsible for understanding crowd dynamics, crowd management techniques, and the venue’s emergency response plan, among other characteristics. Crowd management is rooted in the need to safely evacuate occupants during an unwanted fire emergency, but today this has evolved into an all-hazards focus for fire-related emergencies, non-fire emergencies, and non-emergency situations. To prepare for large assembly events, crowd managers must evaluate specific actions necessary for various situations, anticipated occupancy levels, the adequacy of means of ingress and egress, and anticipated human behavior.

Management challenge

As if this weren’t enough, crowd managers must also be ready for unanticipated events—moments where circumstances can change rapidly, influencing real-time crowd dynamics and requiring immediate action on the part of crowd managers.

While poor management of crowded events routinely leads to civilian injuries and deaths, modern tools and technologies can be used to enhance crowd management strategies. Existing technologies and simulations designed to assess crowd densities and dynamics can be useful, but they are often expensive and challenging to implement, with only questionable accuracy that provides little assistance to crowd managers.

The NFPA web-based crowd monitoring tool, by contrast, is focused on overcoming these challenges, and utilizes modern technologies and algorithms to evaluate crowd movement over time in high-risk spaces. The tool can be accessed with a laptop, tablet, or mobile phone. Users would simply upload a still image or a short video clip to the site, or connect to a streaming video feed. For still images, the tool can run an evaluation for crowd count in a given area. For video options, the tool can capture individual frames at a user-specified interval, ranging from one to 30 seconds, or manually as desired. The frame capture is then evaluated by the model, which predicts the count and provides a crowd density map, which is rendered for the user in about five seconds. The counts are then graphed at the timestamp of each capture, based on the user-specified interval, and displayed to identify trends in crowd density. Early trials have been very promising; the tool has been accurate when subjects were either in the foreground or in the background—an important improvement over previous models, which tended to search for images of complete individuals and had difficulty identifying people who were obscured
.
 

WATCHFUL EYE The crowd management tool under development by NFPA can analyze a photograph or video clip of a crowd and generate a heat map illustrating the areas of highest density. This real-time analysis can help safety officials make potentially live-saving decisions. Photographs: NFPA

Our tool is based on a concept known as computer vision, a subset of computer science that enables us to teach computers to recognize objects in a given scene. Within the computer vision and deep learning communities, this process is defined as congested scene recognition, or CSR, which has two primary objectives: counting the number of individuals in a crowd, and identifying their spatial distribution. The second part is especially important to facility managers, since two different scenes containing the same number of people can have vastly different distributions.

We are testing the feasibility and functionality of our initial prototype, including the accuracy of crowd estimations, timeliness of reporting, value of information, ease of implementation and use, economic feasibility, and computational demand. Our hope is that this tool puts data-informed crowd management within reach of safety officials who have found previous versions to be expensive, inaccurate, or otherwise problematic. That’s in part why we intentionally built the tool using open-source platforms and general-purpose programming languages, where the source code is freely available.

The risks associated with large crowds in a variety of venues will never go away, and crowd managers will always face immense responsibilities that cannot be replaced by technology alone. But it is our hope that new tools and technologies can give crowd managers the relevant, real-time information they need to make split-second decisions and keep people safe. 


VICTORIA HUTCHISON
is a research project manager with the Fire Protection Research Foundation. FREDERICK MACDONALD, a data scientist at NFPA, contributed to this article. Top photograph: Getty Images