The Filter/Refine Strategy: A Study On The Land-Use Resource Dataset In Norway

The use of a spatial index followed by an exact geometry check (a filter/refine strategy) is commonly applied in spatial database management systems. Its efficiency in selecting a small subset of geographic features from a large feature collection is undebatable. But there is a problem with the strategy: For a pure overlap test in the filter phase, the filter/refine strategy gets less optimal as the query window grows larger. Loosely speaken, the reason for this is that the candidates falling completely within the area of the query window are not included in the result although a decision can be made from the filter phase. We argue that an overlap test followed by an inside test will reduce the query costs from quadratic to linear as the query window doubles in northand east dimension. A study on the Land-use resource data set in Norway (DMK) gives an indication of this.