Prediction Intervals for Surface Growing Range Segmentation

The surface growing framework presented by Besl and Jain [2] has served as the basis for many range segmentation techniques. It has been augmented with alternative fitting techniques [17], model selection criteria [11, 15], and solid modelling components [6]. All of these surface growing approaches, however, require global thresholds. Range scenes typically cannot satisfy the global threshold assumption since it requires data noise characteristics to be constant throughout the scene. Furthermore, these approaches can only be applied to range scenes where large seed regions can be isolated. As scene complexity increases, the number of surfaces, discontinuities, and outliers increase, hindering the identification of large seed regions. We present statistical criteria based on multivariate regression to replace the traditional decision criteria used in surface growing. We use local estimates and their uncertainties to construct criteria which capture the uncertainty associated with extrapolating estimated fits. Our criteria allow small robust seed regions to grow to large surface patches without the use of global thresholds. To make the best use of these criteria, we restrict the surface expansion process to very localized extrapolations. This increases the sensitivity to discontinuities and allows regions to refine their estimates and uncertainties as they expand. Our approach has a small number of parameters which are either statistical thresholds or cardinality measures, i.e. we do not use thresholds defined by specific range distances or orientation angles.