Size measurement of pellets in industry is usually performed by manual sampling and sieving techniques. Automatic on-line analysis of pellet size based on image analysis techniques would allow non-invasive, frequent and consistent measurement. We evaluate the statistical significance of the ability of commonly used size and shape measurement methods to discriminate among different sieve-size classes using multivariate techniques. Literature review indicates that earlier works did not perform this analysis and selected a sizing method without evaluating its statistical significance. Backward elimination and forward selection of features are used to select two feature sets that are statistically significant for discriminating among different sieve-size classes of pellets. The diameter of a circle of equivalent area is shown to be the most effective feature based on the forward selection strategy, but an unexpected five-feature classifier is the result using the backward elimination strategy. The discrepancy between the two selected feature sets can be explained by how the selection procedures calculate a feature's significance and that the property of the 3D data provides an orientational bias that favours combination of Feret-box measurements. Size estimates of the surface of a pellet pile using the two feature sets show that the estimated sieve-size distribution follows the known sieve-size distribution.

Size measurement of rocks is usually performed by manual sampling and sieving techniques. Automatic on-line analysis of rock size based on image analysis techniques would allow non-invasive, frequent and consistent measurement. In practical measurement systems based on image analysis techniques, the surface of rock piles will be sampled and therefore contain overlapping rock fragments. It is critical to identify partially visible rock fragments for accurate size measurements. In this research, statistical classification methods are used to discriminate rocks on the surface of a pile between entirely visible and partially visible rocks. The feature visibility ratio is combined with commonly used 2D shape features to evaluate whether 2D shape features can improve classification accuracies to minimize overlapped particle error.

The size distribution as a function of weight of particles is an important measure of product quality in the mining and aggregates industries. When using manual sampling and sieving, the weight of particles is readily available. However, when using a machine vision system, the particle size distributions are determined as a function of the number of particles. In this paper we first show that there can be a significant weight-transformation error when transforming from one type of size distribution to another. We also show how the problem can be overcome by training a classifier and scaling the results according to calibrated average weights of rocks. The performance of the algorithm is demonstrated with results of measurements of limestone particles on conveyor belts.

Evaluation of Spherical Fitting as a technique for sizing iron ore pellets is performed. Size measurement of pellet in industry is usually performed by manual sampling and sieving techniques. Automatic on-line analysis of pellet size would allow non-invasive, frequent and consistent measurement. Previous work has used an assumption that pellets are spherical to estimate pellet sizes. In this research we use a 3D laser camera system in a laboratory environment to capture 3D surface data of pellets and steel balls. Validation of the 3D data against a spherical model has been performed and demonstrates that pellets are not spherical and have physical structures that a spherical model cannot capture.

Size measurement of pellets in industry is usually performed by manual sampling and sieving techniques. Automatic on-line analysis of pellet size based on image analysis techniques would allow non-invasive, frequent and consistent measurement. We make a distinction between entirely visible and partially visible pellets. This is a significant distinction as the size of partially visible pellets cannot be correctly estimated with existing size measures and would bias any size estimate. Literature review indicates that other image analysis techniques fail to make this distinction. Statistical classification methods are used to discriminate pellets on the surface of a pile between entirely visible and partially visible pellets. Size estimates of the surface of a pellet pile show that the overlapped particle error is overcome by only estimating the surface size distribution with entirely visible pellets.

We present a new approach to approximate continuous-domain mathematical morphology operators. The approach is applicable to irregularly sampled signals. We define a dilation under this new approach, where samples are duplicated and shifted according to the flat, continuous structuring element. We define the erosion by adjunction, and the opening and closing by composition. These new operators will significantly increase precision in image measurements. Experiments show that these operators indeed approximate continuous-domain operators better than the standard operators on sampled one-dimensional signals, and that they may be applied to signals using structuring elements smaller than the distance between samples. We also show that we can apply the operators to scan lines of a two-dimensional image to filter horizontal and vertical linear structures.

This paper proposes a way of better approximating continuous, two-dimensional morphologyin the discrete domain, by allowing for irregularly sampled input and output signals. We generalizeprevious work to allow for a greater variety of structuring elements, both flat and non-flat. Experimentallywe show improved results over regular, discrete morphology with respect to the approximation ofcontinuous morphology. It is also worth noting that the number of output samples can often be reducedwithout sacrificing the quality of the approximation, since the morphological operators usually generateoutput signals with many plateaus, which, intuitively do not need a large number of samples to be correctlyrepresented. Finally, the paper presents some results showing adaptive morphology on irregularlysampled signals.

Mathematical morphology (MM) on grayscale images is commonly performed in the discretedomain on regularly sampled data. However, if the intention is to characterize or quantify continuousdomainobjects, then the discrete-domain morphology is affected by discretization errors that may bealleviated by considering the underlying continuous signal, given a correctly sampled bandlimited image.Additionally, there are a number of applications where MM would be useful and the data is irregularlysampled. A common way to deal with this is to resample the data onto a regular grid. Often this createsproblems where data is interpolated in areas with too few samples. In this paper, an alternative way ofthinking about the morphological operators is presented. This leads to a new type of discrete operatorsthat work on irregularly sampled data. These operators are shown to be morphological operators thatare consistent with the regular, morphological operators under the same conditions, and yield accurateresults under certain conditions where traditional morphology performs poorly

This paper introduces a new operator that can be used to ap-proximate continuous-domain mathematical morphology on irregularly sampled surfaces. We define a new way of approximating the continuous domain dilation by duplicating and shifting samples according to a flat continuous structuring element. We show that the proposed algorithm can better approximate continuous dilation, and that dilations may be sampled irregularly to achieve a smaller sampling without greatly com-promising the accuracy of the result.

The particle size distribution of fragmented rock in mines significantly affects operational performance of loading equipment, materials handling and crushing systems. A number of methods to measure rock fragmentation exist at present, however these systems have a number of shortcomings in an underground environment. This paper outlines the first implementation of high resolution 3D laser scanning for fragmentation measurement in an underground mine. The system is now used routinely for fragmentation measurement at the Ernest Henry sublevel-cave mine following extensive testing and calibration. The system is being used to study the effects of blasting parameters on rock fragmentation to optimise blast design. Results from 125 three dimensional scans measured the average P₅₀ and P₈₀ to be 230mm and 400mm respectively. The equipment, methodology and analysis techniques are described in detail to enable application of the measurement system at other mines.

We present an up-to-date survey on the topic of adaptive mathematical morphology. A broad review of research performed within the field is provided, as well as an in-depth summary of the theoretical advances within the field. Adaptivity can come in many different ways, based on different attributes, measures, and parameters. Similarities and differences between a few selected methods for adaptive structuring elements are considered, providing perspective on the consequences of different types of adaptivity. We also provide a brief analysis of perspectives and trends within the field, discussing possible directions for future studies.

A priority-based method for pixel reconstruction and incrementalhole filling in incomplete images and 3D surface data is presented.The method is primarily intended for reconstruction of occluded areasin 3D surfaces and makes use of a novel prioritizing scheme, based on apixelwise defined confidence measure, that determines the order in whichpixels are iteratively reconstructed. The actual reconstruction of individualpixels is performed by interpolation using normalized convolution.The presented approach has been applied to the problem of reconstructing3D surface data of a rock pile as well as randomly sampled imagedata. It is concluded that the method is not optimal in the latter case,but the results show an improvement to ordinary normalized convolutionwhen applied to the rock data and are in this case comparable to thoseobtained from normalized convolution using adaptive neighborhood sizes.

Mathematical Morphology is a common strategy for non-linear filtering of image data. In its traditional form the filters used, known as structuring elements, have constant shape once set. Such rigid structuring elements are excellent for detecting patterns of a specific shape, but risk destroying valuable information in the data as they do not adapt in any way to its structure.We present a novel method for adaptive morphological filtering where the local structure tensor, a well-known method for estimation of structure within image data, is used to construct adaptive elliptical structuring elements which vary from pixel to pixel depending on the local image structure. More specifically, their shape varies from lines in regions of strong single-directional characteristics to disks at locations where the data has no prevalent direction.

Continuous casting is a highly efficient process used to produce most of the world steel production tonnage, but can cause cracks in the semi-finished steel product output. These cracks may cause problems further down the production chain, and detecting them early in the process would avoid unnecessary and costly processing of the defective goods. In order for a crack detection system to be accepted in industry, however, false detection of cracks in non-defective goods must be avoided. This is further complicated by the presence of scales; a brittle, often cracked, top layer originating from the casting process. We present an approach for an automated on-line crack detection system, based on 3D profile data of steel slab surfaces, utilizing morphological image processing and statistical classification by logistic regression.The initial segmentation successfully extracts 80\% of the crack length present in the data, while discarding most potential pseudo-defects (non-defect surface features similar to defects). The subsequent statistical classification individually has a crack detection accuracy of over 80\% (with respect to total segmented crack length), while discarding all remaining manually identified pseudo-defects. Taking more ambiguous regions into account gives a worst-case false classification of 131~mm within the 30~600~mm long sequence of 150~mm wide regions used as validation data. The combined system successfully identifies over 70\% of the manually identified (unambiguous) crack length, while missing only a few crack regions containing short crack segments.The results provide proof-of-concept for a fully automated crack detection system based on the presented method.

We demonstrate how known convolution techniques for uncertain data can be used to set the shapes of structuring elements in adaptive mathematical morphology, enabling robust morphological processing of partially occluded or otherwise incomplete data. Results are presented for filtering of both gray-scale images containing missing data and 3D profile data where information is missing due to occlusion effects. The latter demonstrates the intended use of the method: enhancement of crack signatures in a surface inspection system for casted steel.The presented method is able to disregard unreliable data in a systematic and robust way, enabling adaptive morphological processing of the available information while avoiding any false edges or other unwanted features introduced by the values of faulty pixels.

A novel method for automated inspection of small corner cracks in casted steel is presented, using a photometric stereo setup consisting of two light sources of different colors in conjunction with a line-scan camera. The resulting image is separated into two different reflection patterns which are used to cancel shadow effects and estimate the surface gradient. Statistical methods are used to first segment the image and then provide an estimated crack probability for each segmented region. Results show that true cracks are successfully assigned a high crack probability, while only a minor proportion of other regions cause similar probability values. About 80% of the cracks present in the segmented regions are given a crack probability higher than 70%, while the corresponding number for other non-crack regions is only 5%. The segmented regions contain over 70% of the manually identified crack pixels. We thereby provide proof-of-concept for the presented method.

Knowledge about pellet microstructure such as porosity and oxidation degree is essential in improving the pellet macro-behavior such as structural integrity and reduction properties. Manual optical microscopy is commonly used to find such information but is both highly time consuming and operator dependent. This paper presents research to automate image capture and analysis of entire cross-sections of baked iron ore pellets to characterize proportions of magnetite, hematite, and other components.The presented results cover: semi-automated image acquisition of entire pellets, separation of pellet and epoxy and calculation of 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 labeled as magnetite, hematite and pores and additives. Furthermore, spatial distributions of magnetite, hematite and pores and 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.

In general terms, sintering describes the bonding of particles into a more coherent structure, where joins form between packed particles, usually as a result of heating. Characterization of sintering is an important topic in the fields of metallurgy, steel, iron ore pellets, ceramics, and snow for understanding material properties and material strength. Characterization using image analysis has been applied in a number of these fields but is either semi-automatic, requiring human interaction in the analysis, or based on statistical sampling and stereology to characterize the sample. This paper presents a novel fully automatic image analysis algorithm to analyze and determine the degree of sintering based on analysis of the particle joins and structure. Quantitative image analysis of the sintering degree is demonstrated for samples of iron ore pellets but could be readily applied to other packed particle materials. Microscope images of polished cross-sections of iron ore pellets have been imaged in their entirety and automated analysis of hundreds of images has been performed. Joins between particles have been identified based on morphological image processing and features have been calculated based on the geometric properties and curvature of these joins. The features have been analyzed and determined to hold discriminative power by displaying properties consistent with sintering theory and results from traditional pellet diameter measurements on the heated samples, and a statistical evaluation using the Welch t-test.

Image analysis as a technique for fragmentation measurement of rock piles has been the subject of research since the 1980’s and to date, run of mine (ROM) fragmentation optimisation studies have primarily relied on particle size measurement using photographic based 2D imaging systems. Disadvantages of 2D imaging systems include particle delineation errors due to variable lighting and material colour and texture variation; no direct measure of scale & perspective distortion; and inability to distinguish overlapped particles, non-overlapped particles and areas-of-fines. With the development of 3D imaging technologies, there is an opportunity to develop techniques that could improve data collection and overcome the limitations of existing 2D image based systems. This paper describes the first attempt to use 3D high resolution laser scanning techniques to quantify “whole of muckpile” fragmentation from full scale production blasting. During two monitoring campaigns in 2013, high resolution laser scanning data was collected from production blasts at Esperanza Mine (Antofagasta Minerals Group). Fully automated analysis of the 3D data was possible in all cases where the data was of sufficiently high resolution. Manual pre-processing was required when the data was of low resolution to specify the region of fines. Overall results indicated that run of mine fragmentation requirements were meeting specified targets despite the marked differences in powder factors. This was particularly the case for those blasts conducted in similar geological domains. This work has demonstrated that high resolution laser scanning can be used as an alternative technique to measure “whole of muckpile” fragmentation in production blasting.

This paper investigates the possibility to automatically match and recognize individual Scots pine (Pinus sylvestris L.) boards using a fusion of two feature detection methods. The first method denoted Block matching method, detects corners and matches square regions around these corners using a normalized Sum of Squared Differences (SSD) measure. The second method denoted the SURF (Speeded-Up Robust Features) matching method, matches SURF features between images (Bay et al., 2008). The fusion of the two feature detection methods improved the recognition rate of wooden floorboards substantially compared to the individual methods. Perfect matching accuracy was obtained for board pieces with more than 20 knots using high quality images. More than 90% matching accuracy was achieved for board pieces with more than 10 knots, using both high- and low quality images.

In the wood industry, there is a wish to recognize and track wood products through production chains. Traceability would facilitate improved process control and extraction of quality measures of various production steps. In this paper, a novel wood surface recognition system that uses scale and rotationally invariant feature descriptors called K-plets is described and evaluated. The idea behind these descriptors is to use information of how knots are positioned in relation to each other. The performance and robustness of the proposed system were tested on 212 wood panel images with varying levels of normally distributed errors applied to the knot positions. The results showed that the proposed method is able to successfully identify 99–100 % of all panel images with knot positional error levels that can be expected in practical applications

Today, a large number of people are manually grading and detecting defects in wooden lamellae in the parquet flooring industry. This paper investigates the possibility of using the ensemble methods random forests and boosting to automatically detect cracks using ultrasound-excited thermography and a variety of predictor variables. When friction occurs in thin cracks, they become warm and thus visible to a thermographic camera. Several image processing techniques have been used to suppress the noise and enhance probable cracks in the images. The most successful predictor variables captured the upper part of the heat distribution, such as the maximum temperature, kurtosis and percentile values 92–100 of the edge pixels. The texture in the images was captured by Completed Local Binary Pattern histograms and cracks were also segmented by background suppression and thresholding.

The classification accuracy was significantly improved from previous research through added image processing, introduction of more predictors, and by using automated machine learning. The best ensemble methods reach an average classification accuracy of 0.8, which is very close to the authors’ own manual attempt at separating the images (0.83).

Measurement and image analysis of 3D surface profiledata of blasted rock piles in an open-pit mine are presented.A proof-of-concept/demonstration project into determining thesize distribution of the visible rocks on the pile was performed.The results demonstrate the capacity to collect high resolution3D surface profile data using a high-end two-axis scanning laserrange-finder. Furthermore, automated image analysis was appliedto this data to identify and size the rocks on the pile. Areas ofvery fine particles, too small to individually detect, are able to bedetected and classified as areas-of-fines. Detection of these areasof-fines is extremely important as the amount of fine material isa key factor in evaluating blasting outcomes. The algorithms toperform this segmentation and classification analysis are outlinedand results are shown in the form of images and sizing graphs.

Fully automated online measurement of the size distribution of limestone particles on conveyor belt is presented based on 3D range data collected every minute during 13 hours of production. The research establishes the necessary measurement technology to facilitate automatic control of rock crushing or particle agglomeration processes to improve both energy efficiency and product quality. 3D data from laser triangulation is used to provide high resolution data of the surface of the stream of rocks. The 3D data is unaffected by color variation in the material and is not susceptible to scale or perspective distortion common in 2D imaging. Techniques are presented covering; sizing of particles, determination of non-overlapped and overlapped particles, and mapping of sizing results to distributions comparable to sieving. Detailed variations in the product sieve-size are shown with an abrupt change when the size range of the limestone particles was changed.

Fully automated online measurement of the size distribution of limestone fragments on conveyor belt is presented based on 3D range data collected every minute during 13 hours of production. The research establishes the necessary measurement technology to facilitate automatic control of particle breaking or aggregating processes to improve both energy efficiency and product quality. Techniques are presented covering; sizing of fragments, determination of non-overlapped and overlapped fragments, and mapping of sizing results to distributions comparable to sieving. Detailed variations in the product sieve size are shown with an abrupt change when the size range of the limestone fragments was changed.

Advances in developing vision systems for fully automated, non-invasive, rapid particle sizing in the mining and aggregates industries are presented using the example of fragmentation measurement of ore in an underground LHD unit bucket. 3D surface data (Figure 1) 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. Results are presented covering; 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 presented research allows the possibility of feedback to blasting, and automatic control of mills and crushers when applied to conveyor belt applications.

Ongoing research into an industrial prototype 3D imaging and analysis system is presented for non-contact measurement of the size of iron ore green pellets on the conveyor belt. The imaging hardware has been installed at a pellet production plant and captures 3D surface data of pellets on the conveyor belt. Segmentation methods based on mathematical morphology are applied to the 3D surface data to identify individual pellets. Pellet sizing based on the segmented 3D surface data has been performed and demonstrates a correlation with pellet sieve size.

This project contributed advances in developing vision systems for fully automated, non-invasive, rapid particle sizing for fragmentation measurement of ore in an underground LHD unit bucket. 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 delivered a prototype fully automated online measurement system for size distribution measurement of limestone aggregate on conveyor for Nordkalk. The system is based on 3D profile data from laser triangulation and measures an approximately 1.5m long section of limestone every minute, producing an estimate of the sieve size distribution. A specific goal of the project was to demonstrate that the image processing results would trend in the right direction tracking changes in the material size. This has been achieved with the sizing result producing larger results when large rocks are on the conveyor and smaller when small rocks are on the conveyor. Additionally, the installation has proven that the measurement hardware works, that the system has the ability to distinguish between overlapped and non‐overlapped rocks, and produce sizing results based on the non‐overlapped rocks. A stage two project is currently in application to perform the necessary field trials and experimentation to adapt the sieve size estimation to the broad range of material sizes used at the mine site. The underlying research has many applications including; quality control of bulk material before transportation to the customer, process control on the input to ovens or crushers, quantifying the output from these same processes, and to validate the size of material received from the supplier. These applications contribute to the knowledge and process flow in concepts such as "mine‐to‐mill" for overall mine efficiency. The project was a collaboration between Nordkalk, MinBaS, MBV-Systems, ProcessIT Innovations and Luleå University of Technology and was partly supported by the INTERREG IVA program of the European Union.

The size distribution of piles of blasted and crushed rock in mining is so closely related to the economics of the mine that it must be measured quickly and accurately. This thesis develops a mathematical and 3D imaging solution for determining the size distribution of laboratory rock piles based on surface measurements of the pile.

An industrial prototype 3D imaging and analysis system has been developed that measures the pellet sieve size distribution into 9 sieve size classes between 5mm and 16+mm. The system is installed and operational at a pellet production plant capturing and analysing 3D surface data of piled pellets on the conveyor belt. 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 mathematical morphology are applied to the 3D surface data to identify individual pellets. Determination of the entirely visible pellets is made using a new two feature classification, the advantage being that this system eliminates the resulting bias due to sizing partially visible (overlapped) particles based on their limited visible profile. Literature review highlights that in the area of size measurement of pellets and rocks, no other researchers make this distinction between entirely and partially visible particles. Sizing is performed based on best-fit-rectangle, classified into size classes based on one quarter of the measured sieving samples, and then compared against the remaining sieve samples.

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

Optimisation and control of blasting, comminution and agglomeration processes is a complex task with large potential for gains in energy efficiency and productivity in the mining and aggregates indus-tries. In order to realise these benefits, fully automated, non-contact, on-line particle size measurement tech-nology is required to provide the necessary feedback. Results from two installations are presented for measurement of crushed rock on conveyor belt from a variety of sieved products between 0 and 100mm and material from a primary crusher up to 250 mm. Detailed sieve-size distribution results are presented with results calculated directly from the 3D surface profile data of the conveyor. No calibration against sieve samples has been performed clearly demonstrating the capability of the system to operate in a plug-and-play low maintenance setup. The technology measures 3D surface profile data which is used to delineate individual particles, identify non-overlapped particles, identify areas-of-fines, and produce a sieve-size distribution.

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. However, the generic morphological erosion and dilation operation in the CUDA NPP library is relatively naive, and performance scales expensively with increasing structuring element size. The objective of this work is to produce a freely available GPU capability for morphological operations so that fast GPU processing can be readily available to those in the morphological image processing community. Open-source extensions to CUDA (hereafter referred to as LTU-CUDA) have been produced for erosion and dilation using a number of structuring elements for both 8 bit and 32 bit images. Support for 32 bit image data is a specific objective of the work in order to facilitate fast processing of image data from 3D range sensors with high depth precision. Furthermore, the implementation specifically allows scalability of image size and structuring element size for processing of large image sets. Images up to 4096 by 4096 pixels with 32 bit precision were tested. This scalability has been achieved by forgoing the use of shared memory in CUDA multiprocessors. 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 NPP. However, 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. This performance limitation is mitigated for larger structuring elements using an optimised transpose function, which is not default in NPP, and applying the horizontal structuring element. LTU-CUDA is an ongoing project and the code is freely available at https://github.com/VictorD/LTU-CUDA.

Once segmentation of 3D surface data of a rock pile has been performed, the next task is to determine the visibility of the surface rocks. A region boundary-following algorithm that accommodates irregularly spaced 3D coordinate data is presented for determining this visibility. We examine 3D surface segmentations of laboratory rock piles and determine which regions in the segmentation correspond to entirely visible rocks, and which correspond to overlapped or partially visible rocks. This is a significant distinction as it allows accurate size determination of entirely visible rocks, separate handling of partially visible rocks, and prevents erroneous bias resulting from mishandling partially visible rocks as smaller entirely visible rocks. Literature review indicates that other rock pile sizing techniques fail to make this distinction. The rock visibility results are quantified by comparison to manual surface classifications of the laboratory piles and the size results are quantified by comparison to the sieve size.

Image segmentations have been performed to identify the surface fragmentation of rock piles using 3D surface data, and quantified. The advantages for fragmentation measurement using image analysis are significant and include: quantifying image segmentation performance in isolation of the downstream processes of fragment classification and size distribution calculation, utilization of 3D data to overcome various limitations of photographic-based image analysis, and the capacity to use 3D fragment data to eliminate the misclassification of partially visible fragments as smaller entirely visible fragments. The segmentation results have been quantified by comparison with the 3D surface data of each individual rock fragment. Mathematical morphology and image segmentation algorithms have been extended from greyscale image-based definitions and applied to irregularly spaced 3D coordinate surface data. 3D coordinate surface data can now be morphologically processed directly in 3D, segmented, visualized, and directly compared to the actual surface fragmentation in order to quantify the results.

To assess the present-day functionality of large-scale sublevel caving (SLC) at LKAB Kiruna a comprehensive measurement program was undertaken involving blast function, fragmentation and gravity flow. As part of this assessment, a fragmentation measurement trial was performed based on 3D imaging ofthe draw-point and corresponding bucket load of the underground load-haul-dump (LHD) excavator. 3D image data from stereo photogrammetry was collected and an automated image analysis strategy developed. A number of data sets were collected for each of the draw-point and LHD bucket, along withsieving results for four of the LHD bucket loads (totally about 70 tonnes). Two of the sieving results were used to inform the automated image analysis strategy, and two were held back as a comparison. Large variations in the visible particles are apparent when comparing corresponding draw-points and LHD bucketshighlighting the impact of sampling location and the need to measure large quantities of data in order to avoid bias from small samples. The results show that 3D imaging and analysis can produce fully automated measurement and analysis of the visible particle size distribution. Although this is not the same as the sievesize distribution it provides useful estimation of both the larger size classes and a bulk estimate of fine material below approximately 60mm. The 3D stereo photogrammetry measurement system used produced very high 3D point density but this was achieved using a custom up-sampling technique which significantly smoothed the data, removing small particles, smoothing edges, and this negatively affected the particle delineation algorithms.

Fragmentation in sublevel caving (SLC) is vitally important. Both gravity flow and any downstreamprocesses are affected. Fairly coarse fragmentation may lead to larger draw bodies (isolatedextraction zones) and hence potentially higher primary ore recovery and depressed/delayedwaste rock inflow from above. Fewer flow disturbances are expected by mitigating oversize. Inaddition, it requires less boulder handling and reduces wear and possible hang-up problems inorepasses. To assess the present-day functionality of large-scale SLC, a multiyear, comprehensivemeasurement program was initiated. It covers the main elements for SLC, namely blast function,fragmentation and gravity flow. The present paper focuses on fragmentation measurements. Animage acquisition system was used to document the drawpoint and the load-haul-dump (LHD)bucket for each mucking cycle. At the beginning, four buckets (about 70 t total) were sieved tovalidate the results of both 2D and 3D image analysis techniques. The fundamental and specificproblems are discussed herein. At the moment, fragmentation of the SLC rings is evaluated usinga quick rating system and a 2D fragment delineation software. The results enable a descriptionof fragmentation and its variation during mucking but also – combined with the gravity flowmeasurements – conclusions on the fragmentation for different parts of the SLC ring. Possibleinfluences on flow disturbances and ore recovery/dilution are investigated. The recent findingsallow a better understanding of breakage and flow and support future process improvements.