New Algorithms for 3D Registration Using Raw Point Techniques
Bibliographic record
Abstract
Registration of two 3D point clouds is a problem encountered in many domains in 3D computer vision. A correct correspondence between the overlapping portions of the point clouds can be used to generate a transform that registers the two point clouds. The search for correspondences can be done using techniques that attempt to find similarity between local or global surface geometry. However, such methods have their limitations. Global methods fail when there is significant surface occlusion, while local ones degrade in performance in the face of noise and outliers. In cases of significant occlusion, noise, and outliers, it is best to rely on a large number of correspondences between very small subsets of points [3]. The higher the number of correspondences, the more likely it is to find the correspondence that represents the correct transformation to achieve registration. In this thesis, a generalization to these subsets that allows us to control the degree of their ambiguity is introduced. Our generalization provides a way to optimize the number of correspondences as to achieve the maximum speed up without sacrificing robustness. We show that for the problem of offline registration we can achieve a speed up factor of up to 4.4x using our generalized version of the algorithm. We also use our generalization with and improved version of 4PCS [52], and show that we can achieve further efficiency improvements over the state of the art in raw point registration iIn addition to the generalization of the 4-Point Congruent sets method, we present a novel RANSAC framework for 3D registration. Unlike the standard RANSAC ap- proach, our approach requires sampling only 2 points and thus reducing the worst time complexity of the algorithm. We present two flavours of this approach and evalu- ate it by comparing it to 4PCS and Super 4PCS. We achieve a speed up improvement of up to 57x over 4PCS and are on par with Super 4PCS in most cases.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.004 |
| Meta-epidemiology (narrow) | 0.002 | 0.002 |
| Meta-epidemiology (broad) | 0.002 | 0.003 |
| Bibliometrics | 0.004 | 0.004 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.003 | 0.005 |
| Open science | 0.005 | 0.005 |
| Research integrity | 0.003 | 0.005 |
| Insufficient payload (model declined to judge) | 0.011 | 0.011 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".