Sparse Point Registration

Abstract

This work introduces a Sparse Point Registration (SPR) method for performing robust registration given the geometric model of the object and few sparse point-measurements (<20) of the object’s surface. Such a method is of critical importance in applications such as probing-based surgical registration, manipulation, etc. Our approach for SPR is iterative and in each iteration, the current best pose estimate is perturbed to generate several poses. Among the generated poses, the best pose as evaluated by an inexpensive cost function is used to estimate the locally optimum registration. This process is repeated, until the pose converges within a tolerance bound. Two variants of the SPR are developed: deterministic (dSPR) and probabilistic (pSPR). Compared to the pSPR, the dSPR is faster in converging to the local optimum, and requires fewer parameters to be tuned. On the other hand, the pSPR provides uncertainty information in addition to the registration estimate. Both the approaches were evaluated using various standard data sets and the compared to results obtained using state-of-the-art methods. Upon comparison with other methods, both dSPR and pSPR were found to be robust to initial pose errors as well as noise in measurements. The effectiveness of the approach is also demonstrated with an application of robot-probing based registration.