Adaptive geofields similarity measure based on binary similarity measures generalization

Abstract

The authors discuss the task of geofield similarity measurement. The local and global approaches to it are reviewed. The rationale of inapplicability of measures, originally developed for images, for general case of geofields is given. The adaptive function family based on generalization of binary similarity measures is proposed. It assumes adaptation to specific scales, tasks and subject domain. The software implementation of this functions family is discussed. Its applicability for detection of similarities and difference of several geofield special cases is considered. The computational performance of the proposed mechanism different use is studied. It is shown that MT-IoU one (Multi-threshold Intersection-over-Union) is flexible and performant framework for building specialized geofield similarity measure functions