Perspective-independent point cloud processing: towards streamlining 3D computer vision workflows and enhancing 3D indoor scene perception
Bibliographic record
Abstract
The advent of commercially accessible depth sensors, augmented reality headsets, and smartphones equipped with depth sensing technologies has revolutionized the acquisition of 3D data, enabling comprehensive spatial understanding in real-world environments. This advancement has led to the widespread adoption of 3D data, offering significant benefits across a range of applications, from augmented reality to autonomous navigation. However, the complexity of indoor scenes poses significant challenges for 3D computer vision systems, particularly in cluttered environments where background surfaces hinder the detection and analysis of relevant foreground objects. This PhD thesis presents a comprehensive study of perspective-independent point cloud processing techniques tailored to address the challenges posed by cluttered and complex indoor environments. The primary objectives focus on streamlining 3D computer vision workflows by contextually segmenting and subtracting 3D background surfaces while enhancing 3D scene perception through accurate identification of these surfaces and spatial relationships within indoor scenes. Alongside the primary objectives, this thesis also addresses two sub-objectives: size reduction of indoor point clouds and labeling of various elements within complex indoor scenes. To achieve these objectives, four research endeavors are presented. Initially, two techniques were implemented for bounding surface segmentation and removal: Iterative Region-based RANdom SAmple Consensus (IR-RANSAC) and orientation-based M-estimator SAmple Consensus (MSAC). They considerably reduce the size of 3D datasets and the search space of various 3D computer vision applications, resulting in enhanced performance and faster processing times. IR-RANSAC demonstrates robust performance with a mean F₁ score above 94%, while Orientation-based MSAC achieves a mean F1 score exceeding 98%, showcasing its superior performance and notable computational efficiency. In the subsequent work, PiGPDS, a perspective-independent ground plane detection and segmentation method, was introduced as a method for detecting and segmenting ground planes in 3D complex indoor environments, where the position and orientation of the sensor are unrestricted and unknown. PiGPDS demonstrated exceptional performance, achieving an average F1 score of 96.01%, accurately segmenting ground surfaces of complex 3D indoor scenes acquired from diverse locations with varying pitches and yaws. Finally, in the concluding endeavor, PiPCS, a Perspective-Independent Point Cloud Simplifier, stands as a significant advancement, building upon the foundational research laid out in earlier studies. PiPCS redefines conventional 3D background subtraction techniques by contextually segmenting and eliminating 3D background components, yielding precisely segmented 3D foreground objects without relying on colour or historical data. PiPCS demonstrates outstanding performance, achieving an average F1 score of 91.27% and substantial size reductions averaging 74.11% across all dataset's point clouds. PiPCS optimizes 3D computer vision systems by streamlining their workflows, enhancing indoor scene perception, reducing point cloud size, and enabling precise labeling within complex indoor environments.
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 | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.002 |
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".