Spatial join query is one of the fundamental operations in spatial database systems and geographic information systems. Existing estimating methods often make uniformity assumption, which seldom holds in realistic situation. Moreover, little attention has been paid to distance join, instead of intersection join. This paper first develops a MBR-buffer based method, which can be used to get satisfying estimation results with only several statistics. Then, to improve relative errors, an enhanced method using line segment distribution law for distance join among different kinds of datasets is proposed. Without scanning the entire dataset, this method can provide accurate estimating results. Experiments show that our techniques are more efficient for estimating the selectivity of distance join, and more applicable to realistic datasets.
展开▼