There’s an interesting feature in the Chicago Tribune about remote sensing techniques used to count crowds for last weekend’s Air and Water Show. The feature points to the difficulty of counting crowds, and the work done by the Boston University Center for Remote Sensing to capture the crowd count using aerial imagery. The feature indicates aerial surveying is more scientific than other methods, but that there’s no exact science.
Crowd counting can be a politically charged issue. There’s an interesting San Francisco Chronicle feature from 2003 that discusses the techniques and accuracy of crowd counting for a rally of war protesters in 2003. Again, aerial surveying was used to determine a count of 65,000 participants, when organizers asserted more than 200,000 were in attendance.
People counters for retailers have long been in use. The company Traf-Sys sells sensor technology and software to help count people. Their thermal sensors, surround sensors and wireless beams are capable of counting people as they cross a threshold, and aren’t affected by lighting and other sensor disturbances. A company called Prodco offers similar technology for retail threshold counting, but neither are aimed at counting large crowds of people.
The problem of counting crowds is that thresholds would be street intersections or much larger entryways such as parks, making it really difficult to deploy sensors for this work. The dynamic nature of crowd movement for events such as marches and parades add another level of uncertainty. This is typically addressed in imagery scenarios by applying a grid and looking at density within the grid to estimate the totals, taking the photos at peak times and counting only the peak.
Police use a formula based on the maximum number of people in a given areas, and estimate the percentage of open space to come up with an estimate based on ratios. Both news articles referenced above indicate that police counts are typically conservative and much lower than actual numbers.
Is crowd counting a viable business where geospatial technologies could be better applied? I’m not sure there’s a good business case for developing more accurate technology, but certainly the geospatial toolbox could be applied to better affect here. The use of automated imagery analysis tools would be practical. Thermal sensors would be a much better tool than simply aerial imagery. And perhaps oblique angles would prove more useful than simple overhead views, to factor in overhangs and other obstructions in a dense urban core. Video sensors could also prove useful to collect data, but an automated tool would be needed to comb through that large data stream.
From the brief research that I’ve conducted, technologies applied to counting crowds haven’t advanced much since the 1960s. What technologies and methodologies would you deploy to accurately count a crowd?