Accelerating the computation of multi-scale visual curvature for simplifying a large set of polylines with Hadoop

Over the years, the vast majority of curvature-based simplification algorithms for vector data have employed pseudo-curvatures, rather than real curvatures. This is because the vector data in the field of Geographical Information Science (GIScience) is usually represented in the form of polylines, but polylines do not meet the require-ments of traditional curvature calculation. This situation has been improved since the multi-scale visual curvature (MVC) was proposed. However, due to the high complex-ity and huge computation needed for the algorithm, it is difficult to make effective use of MVC in GIScience. In this paper, the MVC algorithm is used to simplify big vector data in a Hadoop-based cloud. The main challenge is that both the data and computa-tion are simultaneously intensive. An accelerated MVC algorithm for simplification is proposed in this paper. This algorithm is performed by adopting a two-level accelera-tion approach: (1) a simplified calculation method of MVC for the vector data in GIScience, and (2) a parallelization strategy for the MVC algorithm in the Hadoop-based cloud. When simplifying big vector data in gigabyte (GB) size, the execution time is reduced to less than 2.2 % of the original time. The proposed accelerated MVC algorithm has great potential in many GIScience applications, including map general-ization, DEM simplification, and spatial-temporal data compression