International Society for Industrial Process Tomography

Investigating X-Ray Images For Studying Gravitational Flow In Silos Using Crowdsourcing Annotations And Analysis

A. Romanowski¹, K. Grudzień¹

¹Institute of Applied Computer Science, Lodz University of Technology, Lodz, Poland

^*Email: androm@iis.p.lodz.pl

ABSTRACT

In this article we present a novel approach to investigate process tomography imaging, namely analysis and annotation of the collected pictures with use of human crowd sources. The exemplary domain for human computation, specifically for crowdsourcing, presented here is the study of a gravitational flow in silo. Experimental images collected during experiments with trace particles immersed into the bulk solid in model silo are visualized with X-ray industrial tomography setup. Due to nonhomogeneous damping and absorption of radiation, the resulting images are difficult to be processed efficiently with automatic algorithms. Employment of human crowdsourcing paradigm enables to prepare the meaningful, compacted visual data ready for analysis and interpretation. Additionally, crowdsourcing uses a number of workers to cope with mundane work of annotating images and therefore saves time of domain experts that they could devote to interpret pre-processed images. Obtained results show it was possible to analyse the volumetric material structure movement based on the 2D radiography data conveying information about location and movement of tiny metal trace particles. Later, details on the location and movement were coupled with the velocity of material flow obtained directly from the experimental settings. As a result, quantitative description of the flow was derived for different parts of silo model volume.

Keywords crowdsourcing, particle tracking, process tomography, X-ray imaging, silo flow

Industrial ApplicationA4 Human-machine interaction in IPT systems.