Publications

In this paper, a new methodology is presented for contour line generalization. The methodology is based on the combination of artificial neural network, decision tree and classification & regression tree algorithms into an auction schema where the “best” parameter for contour line generalization is selected. The contour lines are generalized using a tension parameter locally adapted to a selection set of line characteristics. The proposed methodology determines the tension to be applied to the curve in function of the contour line characteristics (fractal dimension, the length and others). A test is performed over a 1:25k scale map. The resulting 1:50k scale contour lines were compared with contour lines generalized interactively with the same algorithm and a global precision of 81% was achieved.