Industrial image analysis

The industrial image analysis group is a group within signal processing with a focus on both applied and theoretical image analysis. The group has a strong focus on industrial measurement projects particularly within the mining, aggregates and metallurgy industries, and theoretical advances in morphological image processing.

The research theme encompasses, automated online industrial applications in image analysis, morphological image processing, statistical classification and geometric algorithms, with a specific focus on the analysis and interpretation of 3D surface data.

Projects are peformed within the ProcessIT Innovations centre for industrial collaboration and the group has strong collaborations with providers of measurement hardware and system integration from both local SMEs (machine vision companies) and the Optical Measurement Laboratory at Lapland University of Applied Sciences in Kemi.

Contact Matthew Thurley

LTUCUDAbanner.png

Fast Morphological Image Processing Extensions to CUDA

GPU architectures offer a significant opportunity for faster morphological image processing, and the NVIDIA CUDA architecture offers a relatively inexpensive and powerful framework for performing these operations. This work is developing a freely available GPU extensions (refered to as LTU-CUDA) for NVIDIA CUDA for constant time morphological operations with respect to structuring element size. Structuring elements for both 8 bit and 32 bit images are supported to facilitate fast processing of image data from 3D range sensors with high depth precision. The vHGW algorithm for erosion and dilation independent of structuring element size has been implemented for horizontal, vertical, and 45 degree line structuring elements with significant performance improvements over the generic CUDA library NPP. Memory handling limitations hinder performance in the vertical line case providing results not independent of structuring element size and posing an interesting challenge for further optimisation.   LTU-CUDA is an ongoing project and the code is freely available at https://github.com/VictorD/LTU-CUDA.

SteelSlabsBanner.PNG

Automated Crack Detection in Steel Slabs using Machine Vision : 2009 to 2014

Continuous casting is an efficient industrial technique for casting molten steel into large steel slabs and is a global high production technique. Depending on steel composition, casting parameters and cooling rates, cracks can form in these slabs. Small cracks in slabs become large holes when slab steel is rolled into thin sheet which may not be detected before the thin sheet is used in manufacturing. Early detection of cracks, before shipping to the rolling mill, will prevent significant costs in transportation, slab reheating, rolling costs, and customer dissatisfaction.

This research project is detecting thin sub-millimetre wide cracks in large steel slabs up to 12 m long and 2 m wide. The surface of these slabs is highly textured with oscillation marks, color varation, peeling layers of oxidised steel known as scales and numerous other features. Algorithms are being developed to identify these features and find cracks. Preliminary trials accurately identify slabs with cracks and produce no false detections.

PelletStructureBanner.png
Some of these images are courtesy of LKAB

Automated image analysis for quantitative characterization of iron ore pellet structures : 2011 to 2014

In order to improve the mechanical strength in magnetite iron ore pellets it is importance to characterize and quantitatively measure the degree of sintering and features that impact the process of sintering. This research will develop fully automated image analysis techniques for quantitative characterization of sintering related features. Furthermore to provide these techniques as automated analysis tools for LKAB to apply in internal research to develop a better understanding and control of sintering related phenomena and improve pellet strength and quality

This work is a continuation of the following research trainee Masters Thesis.

Automated Image Analysis of Pellet Structure Using Optical Microscopy : Research Trainee Thesis 2010

Knowledge about pellet microstructure such as porosity and oxidation degree is essential in improving the pellet macro behaviour such as structural integrity and reduction properties. Manual optical microscopy is commonly used to find such information but is both highly time consuming operator dependent.

This thesis presents research to fully automate image capture and analysis of entire cross-sections of baked iron ore pellets to characterize proportions of magnetite, hematite, and other components. Results are presented coving; automation of image acquisition, separation of pellet and epoxy, and calculating total percentages of magnetite, hematite and pores. Using the Leica Qwin microscope software and a segmentation method based on Otsu thresholding these three objectives have been achieved with the phases labelled as magnetite, hematite and pores & additives. Furthermore, spatial distributions of magnetite, hematite and pores & additives are produced for each pellet, graphing proportions in relation to the distance to the pellet surface. The results are not directly comparable to a chemical analysis but comparisons with manual segmentation of images validates the method. Different types of pellets have been tested and the system has produced robust results for varying cases.

limestone_collage.JPG

Automated Online Size Measurement of Rock Fragmentation / Particle Size Distribution

A series of projects have delivered two prototype fully automated online measurement system for size distribution measurement of rocks/particles on conveyor. The system is based on 3D profile data from laser triangulation and measures an approximately 1.5m long section of rocks every 20 seconds, producing an estimate of the sieve size distribution.

This system is now commercialised and available with global distirbution. More information here.  Furthermore the same algorithms have been extended to apply to fragmentation analysis for blasting and caving operations in mining.

Results trend in the right direction tracking changes in the material size largely because the system has the ability to distinguish between overlapped and non‐overlapped rocks, which prevents mis-sizing overlapped rocks as if they were small rocks.

Furthermore, the system can detect areas of visible fines, preventing their mis-classification and sizing as large rocks.

This capacity to identify overlaped rocks, and areas of visible fines mitigates two sources conflicting sources of error common to traditional photographic systems.

Volumetric based size estimation allows direct calculation of a relative weight and sieve-size for each non-overlapped rock coupled with volumetric estimation of fines to provide a cumulative sieve-size distribution estimated directly from the measured data.

The underlying research has many applications including; quality control of particulate material, and process control on inputs or outputs to ovens, crushers or grinding mills. These applications contribute to the knowledge and process flow in concepts such as "mine‐to‐mill" for overall mine efficiency.

The project has been partly supported by the INTERREG IVA program of the European Union, and is commercialised by MBV-Systems AB. Commercial systems are installed at the Boliden Tara Mine in Ireland, and the Sachtleben pigment manufacturer in Finland.

Crusher Output
Particle size distribution measurements after a primary crusher : over 4 hours

The following graph shows product size over more than 6 hours of production loading during which the 40-70mm and 60-90mm products was being loaded.

Daily Log, P20, P50, P80 percent passing values lhd_collage.JPG

Automated Measurement of Rocks in Underground Excavator Buckets

This project contributed advances in developing vision systems for fully automated, non-invasive, rapid particle sizing for fragmentation measurement of ore in an underground excavator bucket (LHD unit). 3D surface data of the bucket contents was collected during operation and fully automated offline processing of the data was performed on 424 data sets, determining the individual fragments in the bucket and estimating their sieve size.

The results demonstrate fully automatic fragment identification, determination of non-overlapped and overlapped fragments to eliminate misclassification of overlapped fragments as smaller fragments, automatic identification of areas of fine material below the resolution of the 3D sensor, and sizing based on the measured 3D fragment profile that takes fragment overlap into consideration.

The project demonstrated the techniques that could be used to provide rapid feedback to blasting, and automatic control of crushers when applied to conveyor belt applications.

This project was a collaboration between LKAB, Softcenter, ProcessIT Innovations and Luleå University of Technology.

mining_and_steel_collage_695px.JPG

Automated Online Measurement of the Size of Soft (Green) Iron Ore Pellets on Conveyor

This project developed an industrial prototype 3D imaging and analysis system for measuring the size distribution of iron ore green pellets. The system has been operational at a pellet production plant since 2007 capturing and analysing 3D surface data of piled pellets on the conveyor belt.

This system is now commercialised and available with global distirbution. More information here.  Furthermore the same algorithms have been extended to apply to fragmentation analysis for blasting and caving operations in mining.

It provides fast, frequent, non-contact, consistent measurement of the pellet sieve size distribution and opens the door to autonomous closed loop control of the pellet balling disk or drum in the future.

Segmentation methods based on morphological image processing are applied to the 3D surface data to identify individual pellets. Determination of entirely visible pellets (non-overlapped) is made using a two feature classification strategy, the advantage being that this system eliminates the bias the results from sizing overlapped particles based on their limited visible profile. The system achieves what a number of commercial 2D fragmentation measurement systems could not satisfactorily achieve for the pellet producer, that is, accurate sizing of the green pellets.

This project was a collaboration between LKAB, Boliden, SSAB, MBV-Systems AB, ProcessIT Innovations and Luleå University of Technology and was funded by the VINNOVA Mining Research Program.