THE USE OF RANDOM-FIELD THEORY FOR FAULT-DETECTION IN AUTOMATED VISUAL INSPECTION OF CARPETS

A unified method for detecting all types of textural fault on a plain carpet by using machine vision is presented. The surface to be examine d is divided into discrete areas, and the characteristics of each area are compared against the characteristics of carpet of acceptable (goo d) quality. The Gaussian Markov random-field (GMRF) model is used for the modelling of the carpet-surface texture. An experimental device in volving the use of a line-scan camera and an IBM personal computer sim ulates on-line inspection of woven carpets to detect various types of fault arising in the production process, Measures for detecting faults are derived from the GMRF model based on sufficient statistics, This measure is very effective in detecting textural differences. The detec tion of unlevel, linear, and other types of fault is discussed. In com bination with a previous linear-faults-detection method, it is possibl e to detect all types of textural fault on a plain carpet in ape effic ient way. With some additional techniques, this method can also be use d for the detection of faults in coloured patterned carpets.