A polygon aggregation method with global feature preservation using superpixel segmentation

Abstract

As the map scale decreases, conflicts can appear among polygonal features such as water areas and buildings. Aggregation is usually employed to clearly represent polygonal features on small-scale maps. Over the past several decades, a number of polygon aggregation algorithms based on vector data have been proposed by various scholars. In contrast, few existing aggregation methods are based on raster data, and it is difficult to simultaneously consider polygonal features with different shape characteristics such as water areas and buildings. However, with the continuous development and progress of computer vision technology, advanced theories and methods, such as superpixel segmentation, have provided brand new opportunities and challenges for polygon aggregation. Both superpixel segmentation and area object aggregation employ spatial clustering to increase the representation level at a coarser resolution. Therefore, this paper proposes a new algorithm called superpixel polygon aggregation (SUPA) for the aggregation of general polygons and buildings based on raster data. In this method, general polygons are first segmented using superpixel algorithms. Then, general polygons are globally aggregated by superpixel selection. In this process, the different semantic characteristics of an object, such as a building or natural water area, control the aggregation decisions, such as the handling of boundaries. Finally, the aggregate boundaries of general polygons (buildings) are locally adjusted by Fourier descriptors (superpixel filling and removal). To test the proposed SUPA method, both water areas and buildings are used to perform aggregation. Compared with the existing traditional method in ArcGIS software, the results show that the proposed SUPA method can preserve the global features of general polygons and the orthogonal features of buildings while maintaining reliable aggregation results.