The paper proposes an original algorithm for the automatic detection of the edges within a cloud of
terrestrial laser points: it is based on the analysis of the numerical values of the local mean curvature, this last estimated by applying a non parametric interpolation model. The discontinuity
contours are thus characterized by points having mean curvature greater than a threshold, so
automatically detected by means of a procedure implementing such a criterion. Several numerical
experiments on simulated and real noisy laser data have shown the proper applicability of the
proposed method for the edge detection: if compared with the results obtained by applying TIN decimation software, those accomplished by the proposed algorithm seem very promising.
