Researchers develop fast image processing for industryPublished On: Sat, Oct 20th, 2012 | Computer Science | By BioNews
The mining and steel industries, as well as the aggregates industry that produces rock and stone for construction and industry needs, were the focus end-users for new fast image analysis capabilities for measurement feedback in industry processes. Researchers at Luleå University of Technology (LTU) have developed LTU-CUDA, a software for fast image processing based on high performance graphics cards commonly found in today’s personal computers.
Image processing on a graphics card with LTU-CUDA would mean fast feedback in seconds of particle size, computing the size fractions of material on a conveyor belt. This in turn means that the mills and crushers in an industrial process could be automatically regulated according to the input and output material sizes, says Matthew Thurley researcher at Luleå University of Technology.
It is in this need for fast analysis where LTU-CUDA comes into the picture. There is already an automated online machine vision system for measurement of particle size fractions on conveyor developed at LTU that works well and measures the production per minute. However, it is not fast enough if you want to measure the size fractions of the rock material to automatically control the crushers and mills. With LTU-CUDA, split-second image analysis of the size fractions could be made using morphological image processing. This field of image processing is a mathematical technique for analyzing and processing geometrical structures, i.e. structures of different large and small rock fragments on a conveyor. Morphological image processing can be used for many others types of imaging and machine vision solutions, and LTU-CUDA has been made available free to download on the internet.
This research and development is a small part of a larger portfolio of projects in progress at the Luleå University of Technology with a focus in the process industry. The development of LTU-CUDA was financed from the EU INTERREG IVA North in close collaboration with the Centre of Excellence Process IT Innovations at LTU. Matthew Thurley has developed LTU CUDA along with one of his masters students Victor Danell and recently received a scientific paper accepted by IEEE Journal of Selected Topics in Signal Processing.
We managed to get a “paper” accepted in a scientific journal and how many students can say they have an IEEE publication, it’s really cool, says Matthew Thurley who leads a group of researchers at the university in the field of image analysis with an industrial focus.
In a global market where competition for the process industry is getting tougher, efficiency is increasingly important. LTU-CUDA can play an important role in providing fast measurement feedback enabling automatic control of industry processes in order to save energy, increase production capacity, and reduce maintenance and downtime.
In the mining industry, numerous processes involve material handling of rock, from blasting to crushing, milling and mineral processing. For crushers, automatic control would be beneficial to account for the uneven size of the input rock material so that production performance is improved and the output is best suited to the downstream processes. For the rotary mills, which is a giant rolling barrel that grinds rock down to very small particles, control of the input rock material size is also important for production performance. Furthermore in the steel industry, it is useful to be able to quickly detect cracks in the steel surface.
Crushing is also important in the aggregates industry that manufactures rock and natural stone into specific size fractions to satisfy customers in a broad array of industries; including production of cement and concrete, road and railway construction and raw materials needed in the steel and paper industries.
The mining and aggregates industries globally would benefit from automatically controlled crushers that can adjust based on the size fraction of rock material that comes into it, so that it can produce just the size that you want out, says LTU researcher Matthew Thurley.
Fast Morphological Image Processing Open-Source Extensions for GPU Processing With CUDA, , IEEE Journal of Selected Topics in Signal Processing.2012, Volume: 6 , Issue: 7 , Page(s): 849 – 855