A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: April - June 2004 (open access)

A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: April - June 2004

This twelfth quarterly technical report discusses work on classification algorithms and an improved lighting system. Measurements on core in the Stillwater Mine core room showed that false positive sulfide classifications occurred at breaks in the core and there were also problems with camera saturation due to glints from crystal facets within the core. To reduce false positives due to noisy data, a wavelet transform smoothing program was explored. Results indicate classifications improve only marginally with smoothed data for spectral angle mapping techniques. A new lighting technique was developed to decrease saturation due to glints from facets within the core samples. This appears to have decreased the glints and it has the added benefit of greatly decreasing false positives near the breaks in the core.
Date: July 29, 2004
Creator: Swanson, Rand
System: The UNT Digital Library
A Real-Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: April - June 2003 (open access)

A Real-Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: April - June 2003

This eighth quarterly technical report discusses the progress made on a machine vision technique for determining coal content and preparations for Year-3 system deployment. Classification maps for coal have been generated and shown to two coal-mining executives. An application for licensing high-speed hyperspectral data analysis software from the Naval Research Laboratory (NRL) has been made. Both Western Energy and Stillwater Mining Company have offered platforms for Year-3 deployment. Barretts Minerals has expressed renewed interest in using Resonon's machine vision system for identifying dolomite in their talc ore and have agreed to provide samples to the Montana Tech team.
Date: July 21, 2003
Creator: Swanson, Rand
System: The UNT Digital Library
A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: January - March 2004 (open access)

A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: January - March 2004

This eleventh quarterly technical report discusses the installation of a spectral machine vision system in the Stillwater mine's core room. In brief, the system has been fabricated, installed, and preliminary measurements have been made. A first round of refinements has been made, included replacing a bad bearing and applying filters to the lighting. A high-speed Spectral Angle Mapper (SAM) program was written to classify the cores in real time. This program identifies sulfides in the core sample quite well, but also produces false positives at boundaries and breaks in the core. Additionally, bright reflections from facets within the ore occasionally saturate the camera. Overall, the project is on schedule, but additional refinement in the algorithm and lighting is required to obtain more accurate results.
Date: April 27, 2003
Creator: Swanson, Rand
System: The UNT Digital Library
A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: January - March 2003 (open access)

A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: January - March 2003

This seventh quarterly technical report discusses the progress made on a machine vision technique for determining coal content and ore grades. Considerable progress has been made on coal analysis. Naval Research Laboratory (NRL) target recognition software has been tested and incorporated into the system. This software decreases analysis time considerably and is more intuitive to use. Work with board-level computers has proceeded well; ultimately this will make the technology more compact and fieldable. Work with talc will be delayed because the graduate student working on this project is leaving the program. Ongoing work is devoted to more detailed coal analysis, improving the software interface, and developing procedures and a users manual.
Date: April 28, 2003
Creator: Swanson, Rand
System: The UNT Digital Library
A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: October - December 2003 (open access)

A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: October - December 2003

This tenth quarterly technical report discusses the progress made on a machine vision technique for ore grading based on hyperspectral imaging. A graduate student at Montana Tech has successfully defended her thesis related to this project. Arrangements with Stillwater Mining Company to deploy a machine vision system in their core room have been completed. Designs for they system that will be installed next quarter have been drawn and parts are being machined. Presentations on the spectral imaging system developed during this effort have been made to Stillwater Mining Company and at a remote sensing symposium at Montana State University.
Date: January 23, 2004
Creator: Swanson, Rand
System: The UNT Digital Library
A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: July - August 2004 (open access)

A Real Time Coal Content/Ore Grade (C2OG) Sensor, Technical Report: July - August 2004

This thirteenth quarterly technical report describes data collection at the Stillwater Mine and an additional improvement to the lighting system. The data collection system was returned to the Stillwater Mine during this reporting period and a large amount of data was collected. The data will be analyzed and correlated with fire assays in the next reporting period. The majority of work done this quarter has been devoted to collecting data from cores scanned in the Stillwater Mining Company core room. This work is somewhat tedious and tiresome, but essential to: (1) obtain enough data to reliably determine the correlation between assay results and spectral imaging results; (2) find bugs and glitches in the system that arise only periodically or after long periods of use; and (3) obtain data on the natural (and man-made) variations in the Stillwater ore that may confuse the machine vision algorithms.
Date: October 22, 2004
Creator: Swanson, Rand
System: The UNT Digital Library