Measuring similarity for query in geometric data

Bahram Sadeghi Bigham,Raheleh Abyar Langroudi

doi:10.1007/s42452-019-1184-y

Bahram Sadeghi Bigham, Raheleh Abyar Langroudi

Open Access

https://doi.org/10.1007/s42452-019-1184-y

Copy DOI

Abstract

One of the most important steps which is used in every data mining projects is searching an object or some similar objects in a data set. For geometric data, there are some methods that measure the difference between two objects. In recent years, researchers have focused on these types of metrics and used them in different applications (e.g., shape matching, machine vision, map generation, etc.). The query problem in these kinds of applications is more complicated when we have big data. In this paper, a new metric is presented which works efficiently when the geometric objects are in discrete form (e.g., polygon or chain). The presented method is important from a theoretical point of view, and its differences with other similar metrics are discussed in this paper.

Highlights

Nowadays, with the advent and development of devices such as PDA smartphones and car navigation systems equipped with positioning systems such as GPS, the possibility of collecting position data generated by moving objects has increased dramatically
Query processing is done on track data, which is difficult in terms of space and time complexity
One of the methods of measuring the similarity is to approximate each curve with a set of points and use the Hausdorff distance defined as: H(A ⋅ B) = max max aA

Summary

Introduction

With the advent and development of devices such as PDA smartphones and car navigation systems equipped with positioning systems such as GPS, the possibility of collecting position data generated by moving objects has increased dramatically. Sensor tracing techniques, such as satellites and radars, collect and process large volumes of motion data. These tools generate raw data of motion with an identifier of the object and its position at a moment’s time. Query processing is done on track data, which is difficult in terms of space and time complexity. To address these issues, a wide range of approaches and ideas has been proposed, and we classify them according to the main method of data mining.

Some common metrics

Determining similarity in queries

Compliance with ethical standards

Conclusion and future works