||2012 marks 15 years since the start of Matthew's PhD research to determine whether it was possible to estimate the size distribution of a pile of particles based on measurement of the surface. 2012 also marks the completion of the main research and development questions and the production of a system that achieves this goal for piled, overlapping particle piles also containing fine material below the resolution of the sensor.
Matthew will present the nature of the problem and a number of the algorithms in the solution. 3D profile data of the surface of the pile is collected and segmented using morphological image processing and watershed segmentation. Topological variation in the 3D surface profile data provides clues to the type of material identified in each segmented region and allows the determination of whether particles are overlapped, non-overlapped, or are actually areas of fine material below the resolution of the sensor. Finally volumetric size estimation of individual particles provides an approximation of particle weight, in order to produce sieve-size-distributions by weight as required by industry. This allows the system to produce realistic results directly from the measured data without any requirement for "calibration" against sieve samples, resulting in a system that can be easily installed in industrial measurement locations.
Optimisation and quality control of blasting, comminution and agglomeration processes
is a complex task with large potential for gains in energy efficiency and productivity in the numerous industries, particularly in mining. In order to realise these benefits fully automated, noncontact,on-line particle size measurement technology has been developed to provide feedback for control and optimisation. Results from two industrial installations are presented showing results for crushed rock on conveyor belt with a variety of sieved products between 0 and 100mm and material from a primary crusher up to 250 mm. Detailed size distribution results are presented with results calculated directly from the 3D surface profile data of the conveyor.