ETRI Journal, vol. 36, no. 5, Oct. 2014
This figure shows the three steps of the MCT AdaBoost detector for the robust sign recognition system at subway stations. First, to train the classifiers, it is necessary to obtain a large set of sign and non-sign images; second, we used the modified census transform for feature extraction; and third, we trained the classifiers of AdaBoost. See page 696.