Positional accuracy or spatial accuracy refers to the accuracy of a sample feature when compared to a control feature. The positional accuracy of a point feature is usually provided by the Euclidean distance between the sample point and a control point. The error can be reported as errors in x, errors in y, and errors in z. For lines and polygons determination of positional accuracy is more complex since these are composed of multiple sample points.
The following is a brief summary of the most common methods for accessing the spatial accuracy of lines taken from (Atkinson and Ariza, 2002).
- Epsilon Bands (Skidmore and Turner, 1992). These methods depend on measuring the offset of the test from control along a cross-section perpendicular to the control. There are various ways of generating the control line.
- Hausdorff Distance (Abbas, Grussenmeyer, and Hottier, 1995). This is simply the smallest and largest offset of test measured perpendicular to the control.
- Maximum Proportion Standard/Maximum Distortion Standard (Veregin, 2000). Examines the areas created by a test line repeatedly crossing a control. Then by creating different ratios of the areas on one or the other side of the control measures of proportion/distortion are derived.
- Buffer Overlay Statistics (Tveite and Langaas, 1999). Overlays buffers created around both test and control and then examines the ratio of the different resulting areas. In our case, the most interesting of these is the ratio of the areas that intersect both control and test compared to the size of the control buffer. As this approaches one the control buffer more and more accurately measures the positional accuracy.
- Buffer Overlay (Goodchild and Hunter, 1997). Creates a buffer around a control line and measures the amount of test within the control buffer. The buffer is increased until a buffer size is found that includes all of test. The size of the last buffer is said to be the accuracy of test.
Of all these methods Buffer Overlay is the most commonly applied since it can readily be calculated using off the shelf GIS functions, the results are simple to interpret, and it is statistically based. In my next post we will use buffer overlay to determine the spatial accuracy of polygons.