CT Image Segmentation Method of Composite Material Based on Improved Watershed Algorithm and U-Net Neural Network Model

doi:10.1007/s12204-021-2385-2

Abstract

Abstract: In the study of the composite materials performance, X-ray computed tomography (XCT) scanning has always been one of the important measures to detect the internal structures. CT image segmentation technology will effectively improve the accuracy of the subsequent material feature extraction process, which is of great significance to the study of material performance. This study focuses on the low accuracy problem of image segmentation caused by fiber cross-section adhesion in composite CT images. In the core layer area, area validity is evaluated by morphological indicator and an iterative segmentation strategy is proposed based on the watershed algorithm. In the transition layer area, a U-net neural network model trained by using artificial labels is applied to the prediction of segmentation result. Furthermore, a CT image segmentation method for fiber composite materials based on the improved watershed algorithm and the U-net model is proposed. It is verified by experiments that the method has good adaptability and effectiveness to the CT image segmentation problem of composite materials, and the accuracy of segmentation is significantly improved in comparison with the original method, which ensures the accuracy and robustness of the subsequent fiber feature extraction process

Key words: image segmentation, composite material, segmentation of adhered objects, watershed algorithm, U-net neural network

摘要： 在复合材料性能的研究中，XCT扫描一直是检测其内部结构特点的重要手段之一，利用CT图像分割技术将显著提高后续材料特征提取精度，对材料性能研究有重要意义。本研究针对复合材料CT图像中纤维截面粘连导致的图像分割准确率低的问题，在芯层区域通过形态学指标评估区域有效性并基于分水岭算法提出迭代分割策略；在过渡层区域利用人工标记训练U-net神经网络模型并应用于分割结果预测，进而提出一种基于改进分水岭算法结合U-net模型的纤维复合材料CT图像分割方法。经实验验证，本方法对复合材料CT图像分割问题具有良好的适应性和有效性，相较于未改进方法分割准确率得到显著提升，保证了后续纤维特征提取过程的准确率和鲁棒性。

关键词: 图像分割，复合材料，粘连个体分割，分水岭算法，U-net神经网络模型

CLC Number:

XUE Yongboa (薛永波)，LIU Zhaob (刘钊)， LI Zeyanga (李泽阳)，ZHU Pinga* (朱平). CT Image Segmentation Method of Composite Material Based on Improved Watershed Algorithm and U-Net Neural Network Model[J]. J Shanghai Jiaotong Univ Sci, 2023, 28(6): 783-792.

References 13

[10]	LIU X B, ZHANG Y W. Ore image segmentation method of conveyor belt based on U-Net and Res UNet models [J]. Journal of Northeastern University (Natural Science), 2019, 40(11): 1623-1629 (in Chinese).
[1]	Editorial Department of China Journal of Highway and Transport. Review on China’s bridge engineering research: 2014 [J]. China Journal of Highway and Transport, 2014, 27(5): 1-96 (in Chinese).
[11]	VINCENT L, SOILLE P. Watersheds in digital spaces: An efficient algorithm based on immersion simulations [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1991, 13(6): 583-598.
[2]	XING L Y, BAO J W, LI S M, et al. Development status and facing challenge of advanced polymer matrix composites [J]. Acta Materiae Compositae Sinica, 2016, 33(7): 1327-1338 (in Chinese).
[12]	XU L L, LU H X, ZHANG M. Automatic segmentation of clustered quantum dots based on improved watershed transformation [J]. Digital Signal Processing, 2014, 34: 108-115.
[3]	TE?MANN M, MOHR S, GAYETSKYY S, et al. Automatic determination of fiber-length distribution in composite material using 3D CT data [J]. EURASIP Journal on Advances in Signal Processing, 2010, 2010: 545030.
[13]	RONNEBERGER O, FISCHER P, BROX T. U-net: Convolutional networks for biomedical image segmentation [M]//Medical image computing and computerassisted intervention: MICCAI 2015. Cham: Springer, 2015.
[4]	ELBERFELD T, DE BEENHOUWER J, DEN DEKKER A J, et al. Parametric reconstruction of glass fiber-reinforced polymer composites from X-ray projection data: A simulation study [J]. Journal of Nondestructive Evaluation, 2018, 37(3): 62.
[5]	LIU Z, QUINN K P, SPERONI L, et al. Rapid threedimensional quantification of voxel-wise collagen fiber orientation [J]. Biomedical Optics Express, 2015, 6(7): 2294-2310.
[6]	GIUSTI R, ZANINI F, LUCCHETTA G. Automatic glass fiber length measurement for discontinuous fiberreinforced composites [J]. Composites Part A: Applied Science and Manufacturing, 2018, 112: 263-270.
[7]	WEI X, CAO Y, FU G, et al. A counting method for complex overlapping erythrocytes-based microscopic imaging [J]. Journal of Innovative Optical Health Sciences, 2015, 8(6): 1550033.
[8]	LIU X Y, WU X, SUN W, et al. Image segmentation of pellet particles based on morphological reconstruction and GMM [J]. Chinese Journal of Scientific Instrument, 2019, 40(3): 230-238 (in Chinese).
[9]	GUAN T. Segmentation and classification of optical microscopic cervical cell images [D]. Changsha: National University of Defense Technology, 2015 (in Chinese).
[10]	LIU X B, ZHANG Y W. Ore image segmentation method of conveyor belt based on U-Net and Res UNet models [J]. Journal of Northeastern University (Natural Science), 2019, 40(11): 1623-1629 (in Chinese).
[11]	VINCENT L, SOILLE P. Watersheds in digital spaces: An efficient algorithm based on immersion simulations [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1991, 13(6): 583-598.
[12]	XU L L, LU H X, ZHANG M. Automatic segmentation of clustered quantum dots based on improved watershed transformation [J]. Digital Signal Processing, 2014, 34: 108-115.
[13]	RONNEBERGER O, FISCHER P, BROX T. U-net: Convolutional networks for biomedical image segmentation [M]//Medical image computing and computerassisted intervention: MICCAI 2015. Cham: Springer, 2015.