Automatic Fracture Detection in CT Scan Images of Rocks Using Modified Faster R-CNN Deep-Learning Algorithm with Rotated Bounding Box
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
In this study, we propose a new approach for automatic fracture detection in CT scan images of rock specimens. This approach is built on top of two-stage object detection deep learning algorithm called Faster R-CNN with a major modification of using rotated bounding box. The use of rotated bounding box plays a key role in the future work to overcome several inherent difficulties of fracture segmentation relating to the heterogeneity of uninterested background (i.e., minerals) and the variation in size and shape of fracture. Comparing to the commonly used bounding box (i.e., axis-align bounding box), rotated bounding box shows a greater adaptability to fit with the elongated shape of fracture, such that minimizing the ratio of background within the bounding box. Besides, an additional benefit of rotated bounding box is that it can provide relative information on the orientation and length of fracture without the further segmentation and measurement step. To validate the applicability of the proposed approach, we train and test our approach with a number of CT image sets of fractured granite specimens with highly heterogeneous background and other rocks such as sandstone and shale. The result demonstrates that our approach can lead to the encouraging results on fracture detection with the mean average precision (mAP) up to 0.89 and also outperform the conventional approach in terms of background-to-object ratio within the bounding box.
Abstract
In this study, we propose a new approach for automatic fracture detection in CT scan images of rock specimens. This approach is built on top of two-stage object detection deep learning algorithm called Faster R-CNN with a major modification of using rotated bounding box. The use of rotated bounding box plays a key role in the future work to overcome several inherent difficulties of fracture segmentation relating to the heterogeneity of uninterested background (i.e., minerals) and the variation in size and shape of fracture. Comparing to the commonly used bounding box (i.e., axis-align bounding box), rotated bounding box shows a greater adaptability to fit with the elongated shape of fracture, such that minimizing the ratio of background within the bounding box. Besides, an additional benefit of rotated bounding box is that it can provide relative information on the orientation and length of fracture without the further segmentation and measurement step. To validate the applicability of the proposed approach, we train and test our approach with a number of CT image sets of fractured granite specimens with highly heterogeneous background and other rocks such as sandstone and shale. The result demonstrates that our approach can lead to the encouraging results on fracture detection with the mean average precision (mAP) up to 0.89 and also outperform the conventional approach in terms of background-to-object ratio within the bounding box.
참고문헌 (21)
- "TensorFlow: Large-scale machine learning on heterogeneous systems" (2015)[URL]
- Andrä, H., "Digital rock physics benchmarks-Part I: Imaging and segmentation", Comput. Geosci, 50, pp.25~32 (2013)
- Crandall D, "CT scanning and flow measurements of shale fractures after multiple shearing events", Int J Rock Mech Min Sci, 100, pp.177~187 (2017)
- Girshick, R, "Fast R-CNN", ICCV (2015)
- He, K., "Mask R-CNN", CoRR abs, pp.arxiv.org, abs[DOI]
- He, K., "Delving deep into rectifiers: Surpassing human-level performance on ImageNet classification", Proceedings of the IEEE international conference on computer vision, IEEE, pp.1026~1034 (2015)
- Ketcham, R. A., "Acquisition, optimization and interpretation of X-ray computed tomographic imagery: applications to the geosciences", Comput. Geosci. Geosc., 27, pp.381~400 (2001)
- Kling, T., "Simulating stress-dependent fluid flow in a fracturedcore sample using real-time X-ray CT data", Solid Earth, 7, pp.1109~1124 (2016)
- Kyle, J. R., "Application of high resolution X-ray computed tomography to mineral deposit origin, evaluation, and processing", Ore Geol. Rev., 65, pp.821~839 (2015)
- Lin, T., "Feature pyramid networks for object detection", CVPR (2017)
- Lin, T., "“Microsoft COCO: common objects in context”, in Computer Vision - ECCV 2014 - 13th European Conference, Zurich,Switzerland, September 6-12, 2014, Proceedings, Part V, ser", LectureNotes in Computer Science, 8693, pp.740~755 (2014)
- Liu, W., "SSD: Single Shot MultiBox Detector", ArXiv151202325 Cs. 2016; 9905:21-37, ~ (2016)
- Nobuyuki Otsu, "A Threshold Selection Method from Gray-Level Histograms", IEEE Transactions on Systems, Man, and Cybernetics, 9(1), pp.62~66 (1979), Institute of Electrical and Electronics Engineers (IEEE)[DOI]
- Redmon, J., "You only look once: Unified, real-time object detection", IEEE Computer Society Conference on Computer Vision and Pattern Recognition (2016)
- Ren, S., "Faster R-CNN: Towards real-time object detection with region proposal networks", NIPS (2015)
- Ronneberger, P. Fischer, "U-Net: Convolutional Networks for Biomedical Image Segmentation", LNCS, 9351, pp.234~241 (2015), Springer
- Schmitt, M., "Classification and quantification of pore shapes in sandstone reservoir rocks with 3-D X-ray microcomputed tomography", Solid Earth, 7, pp.285~300 (2016)
- Voorn, M., "Multiscale Hessian fracture filtering for the enhancement and segmentation of narrow fractures in 3D image data", Comput. Geosci., 57, pp.44~53 (2013)
- Vu, T. X., "Cascade RPN: Delving into high-quality region proposal network with adaptive convolution", Proc. NIPS, pp.1~11 (2019)
- Wennberg, O. P., "Computed tomography scan imaging of natural open fractures in a porous rock; geometry and fluid flow", Geophys. Prospect., 57, pp.239~249 (2009)
- Yen, J. Y., "An algorithm for finding shortest routes from all source nodes to a given destination in general networks", Quarterly of Applied Mathematics, 27(4), pp.526~530 (1970)[DOI]