Countries that acknowledge their fundamental responsibility to understand the economic and cultural status of their people have long valued a national census. Without knowing where the people are, who they are, how many there are, and, by question and analysis, what they are doing — within meaningful timeframes — a nation cannot adequately serve its citizens. Having such data provides competent governance; valuable tax base and infrastructure assessment; equitable funding for improvements, economic growth and enhanced living conditions; and effective management of social issues. Censuses have nobly served these purposes for decades, if not centuries. But over the last several years, there has been a global decline in support for national censuses, based largely on their increasing costs, which are directly caused by the increasing challenges of detecting and maintaining accurate spatiotemporal changes to the database features identifying populations.
With the advent of the digital age and the growth in GIS technology, it was reasonable to expect an increase in the accuracy of the census at a lower cost — by an extraordinary measure over the days of pencil and paper! Yet the opposite has occurred. Concerns over cost, and the quality and usefulness of census data, are now leading many nations to devalue and defund the national census.
GIS – Boon or bust?
GIS technology includes "simple” vector features — building structure points and road centerlines — which are beneficial for accurately locating the population with specific housing units and addresses. Creating an initial database with these features was an expensive proposition, but national stakeholders believed that the payoff of having accurate temporal population data would yield significant return on investment, making it a vital endeavor. Unfortunately, the effort required to maintain a complete, correct and temporal database has proven to be extensive and costly, and history shows that maintaining any database, regardless of its initial value, through increasing costs and decreasing accuracy, eventually leads to its demise.
The problem
Fully automated change detection promised speed and low cost updates, but it failed to deliver on accuracy and completeness. Meanwhile, the cognitive benefits of labor intensive analysis met the fundamental needs for accuracy and completeness, but were too costly and slow to keep pace with the rate of change. Semi-automated processes combining these two extremes held some promise, but opposing forces worked against each other until collapsing far short of acceptable low-cost and high-accuracy requirements.
Meanwhile, experts have long recognized that direct vector-to-orthoimage change detection processes are critical — indeed, the only way — to accurately baseline an existing database and continuously maintain temporal updates. This significant fact was supported by independent subject matter experts contracted by the U.S. Bureau of the Census, Geography Division, in their 2011 Report, "Task 3 - Identifying the Current State and Anticipated Future Direction of Potentially Useful Developing Technologies," and again in the "Change Detection Technology Evaluation”, FY2012 Report. These reports reveal the ongoing need for direct processes, and provide insight on incomplete and costly alternatives that are contributing to the globally decreasing support for national censuses.
Solution realized
FinitEdge™ technologies, developed by World of Change, USA, has solved this direct vector-to-orthoimage change detection need for updating building structure and road centerline databases. FinitEdge™ combines fully automated image segmentation and vector matching processes with a proprietary global analyst crowd workforce for iterative classification processes, resulting in rapid feature extraction and near real-time change detection for authoritative updates. FinitEdge™ detects new map features and concisely identifies the existing point/polyline/polygon features that need to be deleted or geospatially updated to true positions — providing specific update detail rather than inefficient “change versus no-change” indicators of potential change.
For example, FinitEdge™ analyzes rooftop areas detected in new imagery to classify a present dataset of Pintroids™, which are “rooftop” building structure points, as “Existing” (on a roof), “Missing” (not on a roof), and “Unknown” (no roof detected in obscure conditions such as clouds, shadows and trees). A “New” Pintroid is extracted within the roof area of building structures having no current building structure points or Pintroid within the structure’s detected roof edges. The following samples show classified results to an existing dataset:
Existing (Green) The Pintroid falls on a building structure detected within the orthoimagery.
Unknown (Red) The Pintroid falls in obscure conditions, e.g. shadows or vegetation, with no evidence of a building structure detected.
New (Blue) Potential new Pintroids are extracted on building structures that are detected within the orthoimagery but do not contain an existing Pintroid.
Missing (Red) The Pintroid falls where no building structure is detected within the orthoimagery.
Security concerns eliminated
FinitEdge™ processing requires no feature attributes, eliminating all personal privacy concerns. To meet rigid GEOINT security concerns, FinitEdge™ uses un-georeferenced input and applies multiple levels of random encryption to highly segmented data, which is then indiscriminately distributed to a proprietary and totally disparate global crowd workforce.
High accuracy, affordable and fast
FinitEdge™ analyzes orthoimagery and then segments it into pixel groups for edge detection with respect to entropic influences like image perspective, sun angle, shadows, no shadows, clouds, vegetation, vehicles and more. Supervised classification is rapidly applied throughout the process, analyzing each flagged image segment by two to three independent crowd workforce analysts until accuracies exceeding 98% are assured – authoritative change detection and feature extraction.
FinitEdge™ change detection requires no building polygons and has been collectively proven more than 99.5% accurate in Guam, South Africa, Canada and hundreds of counties across the United States, including proficiency in detecting some classes of hidden and hard to capture housing units.
FinitEdge™ creates and maintains complete, correct and temporal feature points on all building structures and, on average per housing unit, costs less than half the price of a postage stamp.
Starting up and scaling from scratch, FinitEdge™ would take approximately 18 months to baseline the building structure points database of a country the size of the U.S.; recognizing that even interim deliveries would typically exceed an agency's ability to ingest the baseline changes and geocode address updates in that timeframe. Subsequent processing for ongoing maintenance of temporal addresses would be processed in near real-time, through an SaaS interface, as often as the country desired and imagery was available. A similar scenario exists for road centerlines which can be run concurrently.
Today
Master Address File feature-specific change detection of this magnitude forever changes the temporal consistency and the statistical relevancy of data in the years leading up to a census, as well as in the intervening years between decennial censuses. This now simple change detection solution will also drive the cost of a national census well below the days of pen and paper, as rightfully anticipated. For example, the expensive $13 billion U.S. 2010 Census cost 56 percent more than the $8.1 billion 2000 Census. The opposite should have been expected, and is now possible.
Technology is now available — for individual countries or United Nations initiatives — to provide current, correct and affordable change detection to the fundamental geospatial data comprising the heart of every census worldwide. FinitEdge™ provides rapid analysis that goes beyond a decennial census to provide for timely updates on the highly transient and displaced populations of this day and age. This breakthrough innovation in change detection technology, with unprecedented accuracy and low cost, will likely bring about a census revival.