Limits...
Iterative most-likely point registration (IMLP): a robust algorithm for computing optimal shape alignment.

Billings SD, Boctor EM, Taylor RH - PLoS ONE (2015)

Bottom Line: Our GTLS approach has improved accuracy, efficiency, and stability compared to prior methods presented for this problem and offers a straightforward implementation using standard least squares.We evaluate the performance of IMLP relative to a large number of prior algorithms including ICP, a robust variant on ICP, Generalized ICP (GICP), and Coherent Point Drift (CPD), as well as drawing close comparison with the prior anisotropic registration methods of GTLS-ICP and A-ICP.The performance of IMLP is shown to be superior with respect to these algorithms over a wide range of noise conditions, outliers, and misalignments using both mesh and point-cloud representations of various shapes.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Science, Johns Hopkins University, Baltimore, MD, United States of America.

ABSTRACT
We present a probabilistic registration algorithm that robustly solves the problem of rigid-body alignment between two shapes with high accuracy, by aptly modeling measurement noise in each shape, whether isotropic or anisotropic. For point-cloud shapes, the probabilistic framework additionally enables modeling locally-linear surface regions in the vicinity of each point to further improve registration accuracy. The proposed Iterative Most-Likely Point (IMLP) algorithm is formed as a variant of the popular Iterative Closest Point (ICP) algorithm, which iterates between point-correspondence and point-registration steps. IMLP's probabilistic framework is used to incorporate a generalized noise model into both the correspondence and the registration phases of the algorithm, hence its name as a most-likely point method rather than a closest-point method. To efficiently compute the most-likely correspondences, we devise a novel search strategy based on a principal direction (PD)-tree search. We also propose a new approach to solve the generalized total-least-squares (GTLS) sub-problem of the registration phase, wherein the point correspondences are registered under a generalized noise model. Our GTLS approach has improved accuracy, efficiency, and stability compared to prior methods presented for this problem and offers a straightforward implementation using standard least squares. We evaluate the performance of IMLP relative to a large number of prior algorithms including ICP, a robust variant on ICP, Generalized ICP (GICP), and Coherent Point Drift (CPD), as well as drawing close comparison with the prior anisotropic registration methods of GTLS-ICP and A-ICP. The performance of IMLP is shown to be superior with respect to these algorithms over a wide range of noise conditions, outliers, and misalignments using both mesh and point-cloud representations of various shapes.

No MeSH data available.


Related in: MedlinePlus

Registration errors for registering a point-cloud target shape. (Experiment 4).Source shapes were randomly generated from a mesh model of a human hip (Fig. 1A), misaligned by (A): [15, 30] mm / degrees and (B): [30, 60] mm / degrees, and registered back to a point-cloud representation of the mesh. The test cases represent different noise models used to generate noise on the source shape (Table 4). For each test case, 300 randomized trials were conducted, with successful registrations being used to compute an average target registration error (TRE). The error bars provide approximate standard deviations of the reported average TRE values. The proposed IMLP algorithm was evaluated relative to standard ICP [1], GICP [11], and CPD [20], as well as relative to near-comparisons of GTLS-ICP [10] and A-ICP [12] using the two variants IMLP-CP and IMLP-MD, which modify IMLP’s most-likely match criteria to that of closest-point and Mahalanobis-distance matching, respectively.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4352012&req=5

pone.0117688.g005: Registration errors for registering a point-cloud target shape. (Experiment 4).Source shapes were randomly generated from a mesh model of a human hip (Fig. 1A), misaligned by (A): [15, 30] mm / degrees and (B): [30, 60] mm / degrees, and registered back to a point-cloud representation of the mesh. The test cases represent different noise models used to generate noise on the source shape (Table 4). For each test case, 300 randomized trials were conducted, with successful registrations being used to compute an average target registration error (TRE). The error bars provide approximate standard deviations of the reported average TRE values. The proposed IMLP algorithm was evaluated relative to standard ICP [1], GICP [11], and CPD [20], as well as relative to near-comparisons of GTLS-ICP [10] and A-ICP [12] using the two variants IMLP-CP and IMLP-MD, which modify IMLP’s most-likely match criteria to that of closest-point and Mahalanobis-distance matching, respectively.

Mentions: The registration accuracies achieved by each algorithm for this experiment are presented in Fig. 5. Similar results were obtained for both ranges of initial misalignment. As seen in the figure, IMLP achieves significantly better registration accuracy than any other algorithm across all test cases for both ranges of misalignment, with exception of CPD for which IMLP achieves comparatively better accuracy in more than half of the test cases considered. Note that unlike Experiment 2A, in this experiment IMLP strongly outperforms ICP even for the initial test cases involving isotropic measurement noise. The reason for this stems from the surface-model covariances used to model unmeasured surface regions surrounding each sample point.


Iterative most-likely point registration (IMLP): a robust algorithm for computing optimal shape alignment.

Billings SD, Boctor EM, Taylor RH - PLoS ONE (2015)

Registration errors for registering a point-cloud target shape. (Experiment 4).Source shapes were randomly generated from a mesh model of a human hip (Fig. 1A), misaligned by (A): [15, 30] mm / degrees and (B): [30, 60] mm / degrees, and registered back to a point-cloud representation of the mesh. The test cases represent different noise models used to generate noise on the source shape (Table 4). For each test case, 300 randomized trials were conducted, with successful registrations being used to compute an average target registration error (TRE). The error bars provide approximate standard deviations of the reported average TRE values. The proposed IMLP algorithm was evaluated relative to standard ICP [1], GICP [11], and CPD [20], as well as relative to near-comparisons of GTLS-ICP [10] and A-ICP [12] using the two variants IMLP-CP and IMLP-MD, which modify IMLP’s most-likely match criteria to that of closest-point and Mahalanobis-distance matching, respectively.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4352012&req=5

pone.0117688.g005: Registration errors for registering a point-cloud target shape. (Experiment 4).Source shapes were randomly generated from a mesh model of a human hip (Fig. 1A), misaligned by (A): [15, 30] mm / degrees and (B): [30, 60] mm / degrees, and registered back to a point-cloud representation of the mesh. The test cases represent different noise models used to generate noise on the source shape (Table 4). For each test case, 300 randomized trials were conducted, with successful registrations being used to compute an average target registration error (TRE). The error bars provide approximate standard deviations of the reported average TRE values. The proposed IMLP algorithm was evaluated relative to standard ICP [1], GICP [11], and CPD [20], as well as relative to near-comparisons of GTLS-ICP [10] and A-ICP [12] using the two variants IMLP-CP and IMLP-MD, which modify IMLP’s most-likely match criteria to that of closest-point and Mahalanobis-distance matching, respectively.
Mentions: The registration accuracies achieved by each algorithm for this experiment are presented in Fig. 5. Similar results were obtained for both ranges of initial misalignment. As seen in the figure, IMLP achieves significantly better registration accuracy than any other algorithm across all test cases for both ranges of misalignment, with exception of CPD for which IMLP achieves comparatively better accuracy in more than half of the test cases considered. Note that unlike Experiment 2A, in this experiment IMLP strongly outperforms ICP even for the initial test cases involving isotropic measurement noise. The reason for this stems from the surface-model covariances used to model unmeasured surface regions surrounding each sample point.

Bottom Line: Our GTLS approach has improved accuracy, efficiency, and stability compared to prior methods presented for this problem and offers a straightforward implementation using standard least squares.We evaluate the performance of IMLP relative to a large number of prior algorithms including ICP, a robust variant on ICP, Generalized ICP (GICP), and Coherent Point Drift (CPD), as well as drawing close comparison with the prior anisotropic registration methods of GTLS-ICP and A-ICP.The performance of IMLP is shown to be superior with respect to these algorithms over a wide range of noise conditions, outliers, and misalignments using both mesh and point-cloud representations of various shapes.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Science, Johns Hopkins University, Baltimore, MD, United States of America.

ABSTRACT
We present a probabilistic registration algorithm that robustly solves the problem of rigid-body alignment between two shapes with high accuracy, by aptly modeling measurement noise in each shape, whether isotropic or anisotropic. For point-cloud shapes, the probabilistic framework additionally enables modeling locally-linear surface regions in the vicinity of each point to further improve registration accuracy. The proposed Iterative Most-Likely Point (IMLP) algorithm is formed as a variant of the popular Iterative Closest Point (ICP) algorithm, which iterates between point-correspondence and point-registration steps. IMLP's probabilistic framework is used to incorporate a generalized noise model into both the correspondence and the registration phases of the algorithm, hence its name as a most-likely point method rather than a closest-point method. To efficiently compute the most-likely correspondences, we devise a novel search strategy based on a principal direction (PD)-tree search. We also propose a new approach to solve the generalized total-least-squares (GTLS) sub-problem of the registration phase, wherein the point correspondences are registered under a generalized noise model. Our GTLS approach has improved accuracy, efficiency, and stability compared to prior methods presented for this problem and offers a straightforward implementation using standard least squares. We evaluate the performance of IMLP relative to a large number of prior algorithms including ICP, a robust variant on ICP, Generalized ICP (GICP), and Coherent Point Drift (CPD), as well as drawing close comparison with the prior anisotropic registration methods of GTLS-ICP and A-ICP. The performance of IMLP is shown to be superior with respect to these algorithms over a wide range of noise conditions, outliers, and misalignments using both mesh and point-cloud representations of various shapes.

No MeSH data available.


Related in: MedlinePlus