2022 Fiscal Year Final Research Report
Development and Application of an Automatic Complementation System for Multiple Object Point Cloud Data from 3D LIDAR
| Project/Area Number |
20K04399
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 20020:Robotics and intelligent system-related
|
|---|
| Research Institution | Wakayama University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Keywords | 3次元点群データ補完 / 深層学習 |
|---|
| Outline of Final Research Achievements |
In this study, we developed a deep learning network that performs completion processing for large-scale 3D point cloud data obtained by measuring the surrounding environment while minimizing spatial computational complexity, and a new evaluation index for completion of 3D point cloud data that indicates whether the gaps in the input point clouds are appropriately filled (completed). We also experimentally showed that the new evaluation index can be used as part of the loss function to train the deep learning network so that it can generate point clouds appropriately.
To demonstrate the effectiveness of the proposed method, validation experiments were conducted using the well-known SemanticKITTI and SemanticPOSS datasets, which are real-world 3D point cloud datasets for self-driving cars.
|
| Free Research Field |
知能ロボット
|
| Academic Significance and Societal Importance of the Research Achievements |
最近の自動運転車などに多く用いられているLiDARセンサで取得された3次元点群データは，疎かつ偏在しているため，本来の構造を正確に捉えることができないことがある．この問題を解決するために，本研究では，疎に分布している，または，一部が欠損したような3次元点群データを，密かつ一様に分布する3次元点群のデータに変換する手法を開発した．本研究課題で扱ったような大規模な3次元点群データを対象とした手法は，我々の知る限りこれまでに存在しない．また，開発した手法により，本来の構造を正確に捉えた3次元点群データを生成することが可能になり，その結果として自動運転車の安全性を向上することに寄与することができる．
|
