Combining error models to reduce the imprecision of geometric length measurement in vector databases

Abstract

Length measurements calculated from the geometry of vector geographic objects, called geometric measurements, are inherently imprecise. The imprecision of the measurements is due to the accumulation of causes of various origins, related to the production processes, and the rules of data representation. In order to reduce the overall imprecision of geometric length measurements, this article proposes to identify the causes of measurement error in the data, to model their respective impact, and finally to combine these different impacts. To do so, five causes of geometric measurement error have been modeled: map projection, terrain disregard, polygonal approximation of curves, digitizing error, and cartographic generalization. To estimate the overall measurement imprecision, three combination methods are proposed: selection of the maximum error, sum of the errors, and quadratic aggregation of the errors. An experiment conducted on a sample of roads represented at a medium scale demonstrates that quadratic error aggregation is the most effective combination method for reducing the imprecision of geometric length measurements.