Million Man March
(adapted from Boston University Press Release dated 27 October 1997)
Dr. Farouk EL-Baz, director of the Center for Remote Sensing at Boston University, today announced the results of the Center’s final analysis of the crowd size present during the “Million Man March” held Monday, October 16, in Washington, DC – 837,000, with a margin of error of plus or minus 20 percent.
This was the Center’s second effort to analyze National Park Service photos taken during the march. The Park Service’s original estimate of 400,000 marchers, which was based on an analysis of images taken from a videotape, brought criticism from the event ‘s organizers, the Nation of Islam, which felt the crowd reached a number between 1.5 million to 2 million.
The new estimate represents an improvement over last week’s effort because the Center’s team was able to work from original negatives. A further reduction in the size of the margin of error could be achieved only through an improvement in the data-gathering methodology employed by the National Park Service.
Last week, Dr. Michael Guillen, science editor of ABC’s “Good Morning America” program, asked Dr. El-Baz if the Center could conduct an analysis of some 35 mm photographs obtained by the television network from the Park Service. Dr. El-Baz gathered a team of 10 research associates and graduate students of the Center in order to develop a methodology for applying remote sensing techniques to the problem of crowd size estimation. After working overnight, the team on Thursday morning, October 19, produced an estimate of 870,000 people in attendance at about 3:30 p.m., with a margin of error of about 25 percent, which meant the actual size of the crowd at that time could have been as low as 650,000 or as high as 1.1 million.
Dr. El-Baz offered to meet with representatives of the National Park Service in order to explain how his team arrived at their estimate and to discuss how the Park Service might obtain appropriate photographs and apply advanced methodologies in the future. At the request of the Park Service, representatives of the Nation of Islam also attended the meeting, which was held Tuesday, October 24, in one of the Center’s laboratories in Boston.
Leading the delegation from the National Park Service was Robert G. Stanton, field director for the National Capitol Area. Representing the National of Islam was Chief of Staff Leonard F. Muhammad. Dr. El-Baz and his colleagues explained to the two groups how they analyzed the photos and suggested specific measures that should be adopted by the Park Service to improve future crowd estimates.
After the six-hour long meeting, the parties agreed to issue the following statement: “As a result of its meeting with Dr. Farouk El-Baz at Boston University this morning, the National Park Service has concluded that the 400,000 number can no longer be considered final. It is proceeding, in cooperation with Boston University, to re-analyze available photographic evidence to generate an improved count. Boston University’s high-tech computerized procedures will take several days to arrive at a better estimate.”
For the second analysis at Boston University, the National Park Service provided the Center with the original negatives of the 35mm photos it had used to produce the Park Service’s first estimate. By digitizing the negatives, instead of color prints, the Center’s team was able to use higher quality images than were previously available, although the resolution in
35 mm film offers, at best, marginal quality.
Other limitations still faced the Center’s team because of the Park Service’s data-gathering methodology. The Park Service shot the videotape and still-photos from a helicopter that flew alongside the area of the march. A helicopter is an unstable platf orm for aerial photography due to vibration, and variations in both altitude and the tilt of the aircraft. This results in a different slant range in each photo and, because the helicopter flew alongside the march, rather than directly overhead, a great obliquity of the line-of-sight.
The final number, based on a count using negatives provided by the NPS, was 837,214 people, with a 20 percent margin of error.
Technical Methodology Description
EL-Baz and his colleagues adapted a methodology originally developed to count dunes in a desert and trees in a forest in order to determine the number of participants at the October 16, 1995 Million Man March.
On October 18, ABC News in Washington scanned (at 300 dots per inch) a series of color prints (from 35 mm negatives) taken of the march by the National Park Service. The scanned images were uploaded through the internet to the Center for Remote Sensing.
The photos were taken from a helicopter with a hand-held camera and showed views of the crowd along the Mall. They were taken from several different heights and at various oblique angles. Scientists at the Center downloaded the images to their computers and began literally to count heads in places where the crowd was dispersed.
The process began with two graduate students breaking the photos down into small sections and enlarging those sections on their computer screens. By concentrating on open areas where individuals were clearly visible, they painstakingly “clicked” on the shadows corresponding to each marcher.
In areas where the crowd was tightly packed, Dr. El-Baz and his colleagues estimated the maximum density per unit area – that is, how many individuals could stand in a single square meter. He simply measured a square meter on the lab floor and saw how many of his students would fit comfortably. He found that six was a high-density count and concluded that six people per square meter corresponded to the most densely packed areas of the march, such as the base of the Capitol and the area around the half-dozen closed-circuit television screens on the Mall.
The Center’s scientists then traced boundaries around the most densely packed areas of the computer images – those with six people per square meter – and around areas of other densities: four per square meter, two per square meter, or as low as just one per 10 square meters where there were comparatively few marchers.
The National Park Service photos were taken from different oblique angles (not parallel to the ground), so the individuals in the foreground looked larger than those in the background, which complicated the process of determining crowd density. Dr. El-Baz and his team therefore had to “straighten” or “flatten” out the pictures so areas of different density could be compared. By taking out the obliquity and simulating a vertical view, the common landmarks on the different photos could also be compared. For example, the corner of two streets and the base of the Washington Monument would have the same relationship to each other in the different images. THis enabled the team to analyze photographic coverage of the entire Mall area.
Finally, the count began. By comparing the flattened photos to a map of the Mall, the team measured the total surface areas of each density and multiplied the number of square meters by the estimated number of people per square meter. Then the subtotals of different density areas were added.
The initial count, made on October 18 and 19, was handicapped by the poor quality of the scanned, downloaded images. The final count, made on October 25 and 26, used the original negatives.
Dr. El-Baz first used this method – tallying shadows and determining densities – to count the number of dunes in the deserts of Egypt and Kuwait, and the Center for Remote Sensing last year completed a six-year study for the U.S. Forest Service to count the number of trees in the forests of California.