Vissarion Fysikopoulos

Recently, there is a growing interest in geospatial trajectory computing. We call trajectories the sequences of time-stamped locations. As the technology for tracking moving objects becomes cheaper and more accurate, massive amounts of spatial trajectories are generated nowadays by smartphones, infrastructure, computer games, natural phenomena, and many other sources.

In this talk we will present the set of tools available in Boost Geometry to work with trajectories highlighting latest as well as older library developments. Starting with more basic operations like length, distance and closest points computations between trajectories we move forward to more advanced operations like compression or simplification as well as the conceptually opposite operation of densify by interpolating or generating random points on a given trajectory. We conclude with the important topic of similarity measurements between trajectories.

All implemented algorithms are parameterized by using the Boost Geometry's strategy mechanism that control the accuracy-efficiency trade-off and work for 3 different coordinate systems (namely, cartesian, spherical and ellipsoidal) each of which comes with its own advantages and limitations.

How to compute the two closest points between two geometries and how this differs from distance computation? What happens when some points are on opposite/antipodal sides of the globe? How can one create equidistant points along a line composed of one or more line segments?

We discuss solutions to those questions highlighting some of the latest developments in Boost Geometry, the library that is currently being used to provide GIS support to MySQL. The implemented algorithms are parameterized by strategies that control the accuracy-efficiency trade-off. The proposed solutions work for 3 different coordinate systems (namely, cartesian, spherical and ellipsoidal) each of which comes with its own advantages and limitations.

We sum up by briefing the plan of future development and a short discussion.

What is the shortest distance between two administrative units in a city? How similar are two hurricane trajectories? In the heart of both questions there is distance computation. In this talk we will discuss distance computation in Boost.Geometry, the library that is currently being used to provide GIS support to MySQL.

We study and implement several families of algorithms for distance such as iterative, series approximation, elliptic arc length, flat earth approximation, spherical. We show particular examples using those algorithms to compute distance between points or polygons. Finally we compare them w.r.t. performance and accuracy. Our ultimate goal is a distance function that given a user defined accuracy utilize the most efficient algorithm.

We sum up by briefing next steps of development in Boost.Geometry and ideas for GSoC'18.

What is a great circle, a loxodrome or a geodesic? What are the differences between them and which one is more suitable for each GIS application? This talk addresses this kind of questions and how geography is implemented in Boost.Geometry. The library that is currently being used to provide GIS support to MySQL.

Following up the introductory talk on Boost.Geometry we discuss the algorithmic, the implementation as well as the user perspective of the development of geography in Boost.Geometry. We define basic geometric objects such as geodesics, and the modeling of the Earth as a sphere or ellipsoid. We try to understand the effect of different Earth models to the accuracy and speed of fundamental geometric algorithms (such as length, area, intersection etc.) by showing particular examples. Finally, we are having a look towards the future of geography in Boost.Geometry.

Comment: This is an entry-level talk on subjects of GIS, geometric algorithms and Boost.Geometry. Attending the introductory talk on Boost.Geometry will help the audience to have a more detailed understanding of the Boost.Geometry development but it is not considered as a prerequisite.