Fast and scalable point cloud indexing

I’m pleased to an­nounce that our pa­per with the title “A sys­tem for fast and scal­able point cloud in­dex­ing us­ing task par­al­lel­ism” has been ac­cep­ted for present­a­tion at the in­ter­na­tional con­fer­ence on Smart Tools and Ap­plic­a­tions in Graph­ics (STAG). The con­fer­ence will take place from 12 to 13 Novem­ber 2020. It will be held fully on­line. Ad­mis­sion is com­pli­ment­ary, so feel free to re­gister and listen to our talk!

In our pa­per, we in­tro­duce a sys­tem for fast, scal­able in­dex­ing of ar­bit­rar­ily sized point clouds based on a task-par­al­lel com­pu­ta­tion model. Points are sor­ted us­ing Mor­ton in­dices in or­der to ef­fi­ciently dis­trib­ute sets of re­lated points onto mul­tiple con­cur­rent in­dex­ing tasks. To achieve a high de­gree of par­al­lel­ism, a hy­brid top-down, bot­tom-up pro­cessing strategy is used. Our sys­tem achieves a 2.3x to 9x spee­dup over ex­ist­ing point cloud in­dex­ing sys­tems while re­tain­ing com­par­able visual qual­ity of the res­ult­ing ac­cel­er­a­tion struc­tures. It is also fully com­pat­ible with widely used data formats in the con­text of web-based point cloud visu­al­iz­a­tion. We demon­strate the ef­fect­ive­ness of our sys­tem in two ex­per­i­ments, eval­u­at­ing scalab­il­ity and gen­eral per­form­ance while pro­cessing data­sets of up to 52.5 bil­lion points.

We’re look­ing for­ward to meet­ing you at the con­fer­ence!

Reference

Bor­mann, P., & Krämer, M. (2020). A sys­tem for fast and scal­able point cloud in­dex­ing us­ing task par­al­lel­ism. In S. Biasotti, R. Pin­tus, & S. Ber­retti (Eds.), Smart Tools and Apps for Graph­ics - Euro­graph­ics Italian Chapter Con­fer­ence. ht­tps://​doi.org/​10.2312/​stag.20201250

Download

The Euro­graph­ics li­cense agree­ment al­lows us to share the fi­nal ver­sion of the pa­per here.


Posted by Michel Krämer
on November, 5th 2020.