Sample sizing in the iron ore industry requires significant manual effort to collect, transport, and sieve the samples as well as record, process and report the data. Efforts have been made to automate this process, which essentially emulate the manual process, except that a person is replaced by a robot. Even in these cells, the sieving and sizing operation consumes a significant proportion of robot time._x000D_
The proposed doubling of production from Australia's iron ore mines over the next five years will place increasing pressure on sample sizing and interest is increasing in alternative automated methods._x000D_
One such method is the adaptation of computer vision technology which has been successfully introduced into the mining industry in a range of applications including feed size distribution to SAG mills and oversize detection from screening stations. The limitation of this technology is that only the top surface is visible and this is used as an indicator of total size distribution._x000D_
Obviously in a sample plant, where 100 per cent of the sample must be measured for acceptable quality assurance, this approach is unacceptable._x000D_
In order to capture 100 per cent of the sample using computer vision technology a monolayer must be formed either by spreading the sample across a moving surface, such as a conveyor or vibrating platform, and passing this under a camera. This will become a surging' load which presents specific problems for some types of computer vision systems, namely those relying on a 2D approach._x000D_
In this paper a 3D computer vision technology will be described which overcomes the problems associated with measuring particle size in a monolayer and experimental results will be presented which demonstrate its effectiveness for lump and fine iron ore sample size measurement.