Information from on-line imaging sensors has great potential for the monitoring and control of
spatially distributed systems. The major difficulty lies in the efficient extraction of information
from the images in real-time, information such as the frequencies of occurrence of specific features
and their locations in the process or product space. This paper uses multivariate image analysis
(MIA) methods based on multiway principal component analysis to decompose the highly
correlated data present in multispectral images. The frequencies of occurrence of certain features
in the image, regardless of their spatial locations, can be easily monitored in the space of the
principal components (PC). The spatial locations of these features in the original image space
can then be obtained by transposing highlighted pixels from the PC space into the original image
space. In this manner it is possible to easily detect and locate (even very subtle) features from
real-time imaging sensors for the purpose of performing statistical process control or feedback
control of spatial processes. Due to the current lack of availability of such multispectral sensors
in industrial processes, the concepts and potential of this approach are illustrated using a
sequence of multispectral images obtained from a LANDSAT satellite, as it passes over a certain
geographical region of the earth’s surface.