書誌事項
- タイトル別名
-
- Research for Detecting Road Features from Point Cloud Data Based on Geometric Shape and Topology
この論文をさがす
説明
国土交通省では,i-Constructionの下,道路管理の効率化に向けて地方整備局の道路管理用車両にセンシング装置を搭載し,地方管理道路を含めた一般道の点群データの収集を開始した.しかし,点群データは地物の情報を保持していないため,対象地物は膨大な点の中から人手で探索する必要がある.この作業を省力化するため,機械学習を用いて点群データから道路地物を識別する手法を一部提案した.しかし,それは,対象地物が少なく,実運用を想定した検証ができていない.そこで,本研究では,維持管理業務の主要な地物の仕様や基準を整理するとともに,識別の判断指標となる特徴を見い出し,それらに基づいた地物識別技術を提案する.そして,実証実験より,機械学習の性能限界を明らかにし,今後の地物識別の実用化に向けた展開を論じる.
In order to improve the efficiency of road management made i-Construction, the Ministry of Land, Infrastructure, Transport and Tourism has installed sensing device on the road management vehicle of the Regional Development Bureau and has started collecting point cloud data for road. However, since the point cloud data does not hold information on features, it is necessary to manually extract feature points from the enormous number of points. In order to save labor in this work, a method that uses machine learning to detect road features from point cloud data has been proposed in a part of previous research. However, in that research, there were few target features, and it was not possible to perform verification assuming actual operation. Thus, in this research, we newly add major features of maintenance work and clarify the definition of features by organizing specifications and standards on geometric shape and topology. And, we propose method for detecting road features based on indicators for detection. We clarify the performance limits of machine learning by demonstration experiment assuming actual operation and discuss future developments toward the practical application of detecting road features.
収録刊行物
-
- 情報処理学会論文誌
-
情報処理学会論文誌 62 (5), 1218-1233, 2021-05-15
dummy publisher
- Tweet
詳細情報 詳細情報について
-
- CRID
- 1390853649853891712
-
- NII論文ID
- 170000184883
-
- NII書誌ID
- AN00116647
-
- ISSN
- 18827764
-
- Web Site
- http://id.nii.ac.jp/1001/00211089/
-
- 本文言語コード
- ja
-
- 資料種別
- journal article
-
- データソース種別
-
- JaLC
- IRDB
- CiNii Articles