Multivariate Image Analysis in Mineral Processing - Advanced Control and Supervision of Mineral Processing Plants - page 125

3

318

303

2

328

313

293

1

333

269

338

323

0

343

283

273

-1

348

-2

-3

SS1
SS2
SS3
SS4
SS5
SS6

-4

-5

-6

-3 -2 -1 0 1 2 3 4

t 1

293 318

273

338 348

323

Figure 3.21 Process transition from steady-state SS2 to SS3 in the t 1 t 2 score space of the PCA

model. Selected images are also shown along the transition. From Liu et al. [37]

−

3.5.3 Estimation of Run-of-Mine Ore Composition on Conveyor
Belts

In this last case study, the problem of estimating ROM ore composition ( i.e. , litho-
logical type, not elemental composition) on conveyor belts is investigated using
multivariate imaging techniques. The very challenging nickel mineral system of
Raglan's mine (Xstrata Nickel Canada, Inc.) is used to illustrate the methods. Each
mineral is very heterogeneous, some have similar visual characteristics ( i.e. , colors
and surface texture), and the rock fragments can be dry or wet due to rain or snow.
The machine vision approach therefore needs to take these issues into account in