Registration and Integration of Multiple Range Images by Matching Signed Distance Fields for Object Shape Modeling

Abstract

Modeling object shapes from multiple range images requires three processes: correction of measurement errors, registration of data shapes, and integrating them as a unified shape representation. We propose a method by which these tasks can be solved simultaneously. Discrete samples of the signed distance field (SDF) of the object surface are used as the shape representation. If the data shapes are registered correctly, the SDFs should match in the common coordinate system. The data shapes are first integrated by averaging the data SDFs assuming that they are roughly preregistered. Then, each data shape is registered to the integrated shape by estimating the optimal transformation. Integration and registration are alternately iterated until the input shapes are properly registered to the integrated shape. Weighting values are controlled to reject outliers derived from measurement errors and wrong correspondences. The proposed method does not suffer from cumulative registration errors because all data shapes are registered to the integrated shape. From the SDF shape representation, a polygon surface model is directly generated. The method was tested on synthetic and real range images.