Polyline Spatial Join Evaluation Using Raster Approximation

Abstract: The main subject of spatial joins is polygons and polylines. Typical polygons and polylines can occupy several Kbytes. Since approximations use much less space, the processing of spatial joins can be greatly improved by the use of filters that reduce the need for examining the exact geometry of spatial objects in order to find the intersecting ones. Candidate pairs of approximations of spatial objects are evaluated using such filters. As a result, three possible sets of answers are identified: the positive one, composed of intersecting pairs; the negative one, composed of non-intersecting pairs; and the inconclusive one, composed of the remaining pairs of candidates. There are many approximations designed for polygons, but few are suitable for approximating polylines. This paper presents a spatial join processor based on the MSQP architecture [24]. We have developed a new polyline approximation, named Five Color Direction Raster Signature (5CDRS)[18]. It is used as the filter of MSQP ́s second step. The performance of the join processor and the approximation was evaluated with real world data sets. The results showed that our approach, when compared to others presented in the literature, reduced the inconclusive answers by 44% in the average. Consequently, the need for retrieving the representation of polylines and carrying out exact geometry tests is reduced by the same factor. As the exact geometry test is the most time consuming step, we have noticed that the overall time is also reduced by 38% in the average. The disk accesses are reduced by 41% in the average.