A partitioned collapse method for long and narrow patches

Collapse is a basic generalization operation in land-use data multi-scale representation. It is a compute-intensive operation and involves complex topological transformations. A feasible way to improve the computation effi-ciency is to adopt a parallel strategy, which means that the collapse of a large spatial range of long and narrow (LN) patches via a partitioning-based approach. However, many topological inconsistencies occur around the partition-cell boundaries. Hence, we propose a partitioned collapse method for LN patches and focus on the reconstruction problem of topological inconsistencies. First, topological inconsistencies were classified into four patterns based on their morphological characteristics. Second, for each of these patterns, a corresponding al-gorithm was formulated to reconcile the topological inconsistencies. Finally, our approach was validated using the national geographical condition data of Chishui City, Guizhou Province. The experiment results show that our method generates smooth split lines from land-use patches by effectively eliminating topology inconsistency patterns near the partition-cell boundary lines and is 8.7 times more efficient than the traditional unpartitioned method within a city region.