2019年
Contact based hierarchical segmentation for granular materials
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
- ,
- ,
- ,
- ,
- 巻
- 11564 LNCS
- 号
- 開始ページ
- 428
- 終了ページ
- 440
- DOI
- 10.1007/978-3-030-20867-7_33
© Springer Nature Switzerland AG 2019. Segmentation of tomography images of granular materials has mostly been done without any semantic notion of the materials these images represent. Poor segmentation; undersegmentation or oversegmentation can be mitigated through the introduction of a prior that defines what a valid boundary between two labels should be. In this paper, we present an approach to multiscale tomography segmentation of granular materials, based on hierarchical segmentation using minima. A 3D tomography image of granular materials is segmented with the introduction of a contact prior combined with other extinction for the purpose of formulating the hierarchy. The contact prior based hierarchy is biased towards a segmentation with a realistic mechanically valid contact network. The proposed method is observed to significantly increase the segmentation accuracy of these grains.
- リンク情報
- ID情報
-
- DOI : 10.1007/978-3-030-20867-7_33
- ISSN : 0302-9743
- eISSN : 1611-3349
- SCOPUS ID : 85068227501