Improving and evaluating boundary algebra filling for identifying polygon intersections

Polygon intersection is important for data processing in geographic information systems. For large datasets, spatial indexing methods such as R-tree allow the identification of polygon intersections, but often retrieve inaccurate results. An improved boundary algebra filling (iBAF) method was preliminarily proposed as an alternative to R-tree. However, its applicability, performance, and accuracy require optimization, and its application conditions remain to be unveiled. This study develops version iBAF 2.0 for a more efficient identification and evaluates performance for different computational intensities and applications. Both intersecting polygons and raster zones within intersections can be rapidly grouped in the rasterized cells of input polygons. The resulting polygons can then be generated by configuring the polygon groups or converting the zones into vectors. We use complexity ratio CR, which is defined as the sum of the number of polygons in each actually intersecting group divided by the total number of polygons, to represent the computational intensity. Two land-use datasets containing 4295 and 741,562 polygons are considered, and we establish test cases containing the same polygons with varying CR. Experimental results show that iBAF 2.0 outperforms R-tree when applied to topology verification; however, its performance is conditional for polygon overlay and area calculation between two layers. Specifically, iBAF 2.0 exhibits higher-efficiency grouping of polygons and raster zones when CR exceeds specific thresholds. In addition, better scalability is achieved compared to R-tree when polygons with complex shapes and additional layers are considered.