搜索

x

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

超声图像中复合材料褶皱形态的Mask-RCNN识别方法

张海燕 徐心语 马雪芬 朱琦 彭丽

引用本文:
Citation:

超声图像中复合材料褶皱形态的Mask-RCNN识别方法

张海燕, 徐心语, 马雪芬, 朱琦, 彭丽

Mask-RCNN recognition method of composite fold shape in ultrasound images

Zhang Hai-Yan, Xu Xin-Yu, Ma Xue-Fen, Zhu Qi, Peng Li
PDF
HTML
导出引用
  • 复合材料在制造和使用过程中不可避免地会产生褶皱缺陷, 因其形态变化多样, 形变程度较小, 人工辨认存在一定障碍, 容易出现错漏情况. 为提高检测效率, 提出利用Mask-RCNN(Mask region-based convolutional neural networks)目标检测算法对复合材料超声图像中不同形态的褶皱缺陷进行检测并分类. 制备含有不同形态褶皱缺陷的碳纤维复合材料层合板, 利用超声相控阵采集全矩阵数据; 通过波数成像算法得到复合材料层合板纵切面图像, 根据地质层中褶皱的几何学特征, 将复合材料层合板中存在的不同褶皱分为三类, 进而建立褶皱形态与材料损伤程度之间的关系; 提出Mask-RCNN算法用于褶皱缺陷的自动检测并分类, 该算法中语义分割的引入可显示褶皱缺陷的位置和形状. 实验结果表明: Mask-RCNN对不同形态褶皱识别的准确率分别达到92.1%, 90.9%和93.3%, 褶皱分类识别准确、有效. 为实现复合材料层合板数据采集-成像-缺陷判别一体化、自动化提供了理论支撑.
    Wrinkle defects will be inevitably produced during composite manufacturing and the in-service life of composite structures. Because of their diverse morphological changes and small deformation, it is difficult to manually identify the wrinkle with important errors. In order to improve the inspection efficiency, a Mask-RCNN algorithm is proposed to detect and classify different forms of wrinkle defects in composites based on phased array images. Carbon fiber composite laminates are prepared first in different forms of wrinkle defects. Secondly, the ultrasonic phased array is used to collect full matrix data. The longitudinal scanning image of the composite laminate is then obtained through the wavenumber imaging algorithm. According to the geometric characteristics of the folds in the geological layer, the wrinkles in the composite laminate are divided into three categories, and the relationship between the wrinkle shape and the material damage degree is established. The Mask-RCNN algorithm is finally proposed for automaticaly detecting and classifying the wrinkle defects. The introduction of semantic segmentation in this algorithm can help to reveal the positions and shapes of wrinkle defects. The experimental results show that the accuracies of Mask-RCNN in the recognition of different forms of wrinkles reach 92.1%, 90.9%, and 93.3%, respectively, and the classification and recognition of wrinkles are accurate and effective. It provides theoretical support for the integration and automation of data acquisition-imaging-defect recognition in composite industries.
      通信作者: 张海燕, hyzh@shu.edu.cn
    • 基金项目: 国家自然科学基金(批准号: 12174245, 11874255, 11904223)资助的课题.
      Corresponding author: Zhang Hai-Yan, hyzh@shu.edu.cn
    • Funds: Project supported by the National Natural Science Foundation of China (Grant Nos. 12174245, 11874255, 11904223).
    [1]

    赵伟, 刘立武, 孙健, 冷劲松, 刘彦菊 2021 宇航材料工艺 51 73Google Scholar

    Zhao W, Liu L W, Sun J, Leng J S 2021 Aerosp. Mater. Technol. 51 73Google Scholar

    [2]

    黎盛寓 2020 塑料科技 48 81Google Scholar

    Li S Y 2020 Plast. Sci. Technol. 48 81Google Scholar

    [3]

    陈珍, 李锐 2021 农机使用与维修 06 33Google Scholar

    Chen Z, Li R 2021 Agric. Machinery Using Maintenance 06 33Google Scholar

    [4]

    张梦杰, 周春鹏, 陈华辉, 张刚利 2019 中国临床神经外科杂志 24 356Google Scholar

    Zhang M G, Zhou C P, Chen H H, Zhang G L 2019 Chin. J. Clin. Neurosurg. 24 356Google Scholar

    [5]

    于晓东, 胡海晓, 贾欲明, 王敏, 曹东风 2020 复合材料学报 37 1932Google Scholar

    Yu X D, Hu H X, Jia Y M, Wang M, Cao D F 2020 Acta Mater. Compos. Sin. 37 1932Google Scholar

    [6]

    Pandey R K, Sun C T 1999 Compos. Sci. Technol. 59 405Google Scholar

    [7]

    Hsiao H M, Daniel I M 1996 Compos. Sci. Technol. 56 581Google Scholar

    [8]

    Hsiao H M, Daniel I M 1996 Composites Part A 27 931Google Scholar

    [9]

    Alexander C, Kulkarni S 2008 Int. Pipeline Conf. 7 61

    [10]

    李韦清, 杨涛, 杨冠侠, 刘思南, 杜宇, 刘畅 2018 固体火箭技术 41 621Google Scholar

    Li W Q, Yang T, Yang G X, Liu S N, Du Y, Liu C 2018 J. Solid Rocket Technol. 41 621Google Scholar

    [11]

    张婷, 黄爱华, 李向前 2021 航空制造技术 64 78Google Scholar

    Zhang T, Huang A H, Li X Q 2021 Aeronaut. Manuf. Technol. 64 78Google Scholar

    [12]

    Sandhu A, Reinarz A, Dodwell T J 2018 Composites Part B 205 58Google Scholar

    [13]

    Koichi M, Yoshihiro M, Akira T, Yoshiro S 2016 Composites Part B 86 84Google Scholar

    [14]

    张海燕, 宋佳昕, 任燕, 朱琦, 马雪芬 2021 物理学报 70 060501Google Scholar

    Zhang H Y, Song J X, Ren Y, Zhu Q, Ma X F 2021 Acta Phys. Sin. 70 060501Google Scholar

    [15]

    周正干, 李洋, 周文彬 2016 机械工程学报 52 1Google Scholar

    Zhou Z G, Li Y, Zhou W B 2016 Chin. J. Mech. Eng. 52 1Google Scholar

    [16]

    Zhang H Y, Song J X, Ren Y, Zhu Q, Ma X F 2021 J. Compos. Mater. 56 1Google Scholar

    [17]

    Callow H J, Hayes M P, Gough P T 2022 Electron. Lett. 38 336Google Scholar

    [18]

    于爱暄 2020 硕士学位论文 (大庆: 东北石油大学)

    Yu A X 2020 M. S. Thesis (Daqing: Northeast Petroleum University) (in Chinese)

    [19]

    Xie Z K, He F X, Fu S P, Sato I, Tao D C, Sugiyama M 2021 Neural Comput. 33 2163Google Scholar

    [20]

    Zhang W P, Chen Y G, Yang W M, Wang G J, Xue J H 2021 IEEE Trans. Neural. Networks. Learn. Syst. 32 4742Google Scholar

    [21]

    Abdulhamit S 2013 Comput. Biol. Med. 43 576Google Scholar

  • 图 1  褶皱缺陷自动分类识别流程图

    Fig. 1.  Flow chart of automatic classification and recognition of wrinkle defects.

    图 2  褶皱形态分类 (a) form-Ⅰ类别褶皱形态; (b) form-Ⅱ类别褶皱形态; (c) form-Ⅲ类别褶皱形态

    Fig. 2.  Classification of wrinkle morphology: (a) Form-Ⅰ category wrinkle morphology; (b) form-Ⅱ category wrinkle morphology; (c) form-Ⅲ category wrinkle morphology

    图 3  Mask-RCNN网络结构

    Fig. 3.  The network structure of Mask-RCNN.

    图 4  ResNet50网络构成

    Fig. 4.  ResNet50 network composition.

    图 5  实验装置图

    Fig. 5.  Experimental device.

    图 6  检测试样及其波数成像 (a) 显微镜图; (b) 波数成像法图; (c) 对图(b)加噪声; (d) 对图(b)调整亮度并沿x轴翻转

    Fig. 6.  est sample and its wavenumber imaging: (a) Microscope image; (b) wavenumber algorithm; (c) add noise to Figure (b); (d) adjust the brightness of Figure (b) and flip along the x-axis.

    图 7  测试图像及其分类结果 (a) 含有一个褶皱的原始图片; (b) 图(a)的测试结果图; (c)含有同一种类多个褶皱的原始图片; (d) 图(c)的测试结果图; (e) 含有多种类多个褶皱的原始图片; (f) 图(e)的测试结果

    Fig. 7.  Test image and its results: (a) The original picture containing one wrinkle; (b) the test result picture of Figure (a); (c) the original picture containing multiple wrinkles of the same type; (d) the test result picture of Figure (c); (e) the original picture containing multiple types of wrinkles; (f) the test result picture of Figure (e).

    表 1  超声相控阵参数设置

    Table 1.  Parameter settings of ultrasonic phased array.

    参数阵元数阵元宽度 /mm阵元
    中心距/mm
    中心频率/MHz采样频率/MHz
    320.91.0550
    下载: 导出CSV

    表 2  分类结果的混淆矩阵

    Table 2.  Classification results confusion matrix.

    真实标签预测标签
    类别A类别B
    类别ATPFN
    类别BFPTN
    下载: 导出CSV

    表 3  form-Ⅰ混淆矩阵数据

    Table 3.  The form-Ⅰ confusion matrix data.

    真实标签预测标签
    form-Ⅰ其他
    form-Ⅰ354
    其他326
    下载: 导出CSV

    表 4  form-Ⅱ混淆矩阵数据

    Table 4.  The form-Ⅱ confusion matrix data.

    真实标签预测标签
    form-Ⅱ其他
    form-Ⅱ100
    其他123
    下载: 导出CSV

    表 5  form-Ⅲ混淆矩阵数据

    Table 5.  The form-Ⅲ confusion matrix data.

    真实标签预测标签
    form-Ⅲ其他
    form-Ⅲ283
    其他231
    下载: 导出CSV
  • [1]

    赵伟, 刘立武, 孙健, 冷劲松, 刘彦菊 2021 宇航材料工艺 51 73Google Scholar

    Zhao W, Liu L W, Sun J, Leng J S 2021 Aerosp. Mater. Technol. 51 73Google Scholar

    [2]

    黎盛寓 2020 塑料科技 48 81Google Scholar

    Li S Y 2020 Plast. Sci. Technol. 48 81Google Scholar

    [3]

    陈珍, 李锐 2021 农机使用与维修 06 33Google Scholar

    Chen Z, Li R 2021 Agric. Machinery Using Maintenance 06 33Google Scholar

    [4]

    张梦杰, 周春鹏, 陈华辉, 张刚利 2019 中国临床神经外科杂志 24 356Google Scholar

    Zhang M G, Zhou C P, Chen H H, Zhang G L 2019 Chin. J. Clin. Neurosurg. 24 356Google Scholar

    [5]

    于晓东, 胡海晓, 贾欲明, 王敏, 曹东风 2020 复合材料学报 37 1932Google Scholar

    Yu X D, Hu H X, Jia Y M, Wang M, Cao D F 2020 Acta Mater. Compos. Sin. 37 1932Google Scholar

    [6]

    Pandey R K, Sun C T 1999 Compos. Sci. Technol. 59 405Google Scholar

    [7]

    Hsiao H M, Daniel I M 1996 Compos. Sci. Technol. 56 581Google Scholar

    [8]

    Hsiao H M, Daniel I M 1996 Composites Part A 27 931Google Scholar

    [9]

    Alexander C, Kulkarni S 2008 Int. Pipeline Conf. 7 61

    [10]

    李韦清, 杨涛, 杨冠侠, 刘思南, 杜宇, 刘畅 2018 固体火箭技术 41 621Google Scholar

    Li W Q, Yang T, Yang G X, Liu S N, Du Y, Liu C 2018 J. Solid Rocket Technol. 41 621Google Scholar

    [11]

    张婷, 黄爱华, 李向前 2021 航空制造技术 64 78Google Scholar

    Zhang T, Huang A H, Li X Q 2021 Aeronaut. Manuf. Technol. 64 78Google Scholar

    [12]

    Sandhu A, Reinarz A, Dodwell T J 2018 Composites Part B 205 58Google Scholar

    [13]

    Koichi M, Yoshihiro M, Akira T, Yoshiro S 2016 Composites Part B 86 84Google Scholar

    [14]

    张海燕, 宋佳昕, 任燕, 朱琦, 马雪芬 2021 物理学报 70 060501Google Scholar

    Zhang H Y, Song J X, Ren Y, Zhu Q, Ma X F 2021 Acta Phys. Sin. 70 060501Google Scholar

    [15]

    周正干, 李洋, 周文彬 2016 机械工程学报 52 1Google Scholar

    Zhou Z G, Li Y, Zhou W B 2016 Chin. J. Mech. Eng. 52 1Google Scholar

    [16]

    Zhang H Y, Song J X, Ren Y, Zhu Q, Ma X F 2021 J. Compos. Mater. 56 1Google Scholar

    [17]

    Callow H J, Hayes M P, Gough P T 2022 Electron. Lett. 38 336Google Scholar

    [18]

    于爱暄 2020 硕士学位论文 (大庆: 东北石油大学)

    Yu A X 2020 M. S. Thesis (Daqing: Northeast Petroleum University) (in Chinese)

    [19]

    Xie Z K, He F X, Fu S P, Sato I, Tao D C, Sugiyama M 2021 Neural Comput. 33 2163Google Scholar

    [20]

    Zhang W P, Chen Y G, Yang W M, Wang G J, Xue J H 2021 IEEE Trans. Neural. Networks. Learn. Syst. 32 4742Google Scholar

    [21]

    Abdulhamit S 2013 Comput. Biol. Med. 43 576Google Scholar

  • [1] 刘俊杰, 左慧玲, 谭鑫, 董健生. 褶皱状单层GeSe各向异性的能带漏斗效应. 物理学报, 2024, 73(23): . doi: 10.7498/aps.20241155
    [2] 刘俊杰, 左慧玲, 谭鑫, 董健生. 褶皱状单层GeSe各向异性的能带漏斗效应. 物理学报, 2024, 73(23): 236801. doi: 10.7498/aps.73.20241155
    [3] 刘淑倩, 张海燕, 张辉, 朱文发, 陈祎婷, 刘雅洁. 融合环形统计矢量的复合材料褶皱缺陷超声相位迁移成像. 物理学报, 2024, 73(17): 174301. doi: 10.7498/aps.73.20240714
    [4] 孟雨欣, 赵漪凡, 李绍春. 褶皱状蜂窝结构的单层二维材料研究进展. 物理学报, 2021, 70(14): 148101. doi: 10.7498/aps.70.20210638
    [5] 张海燕, 宋佳昕, 任燕, 朱琦, 马雪芬. 碳纤维增强复合材料褶皱缺陷的超声成像. 物理学报, 2021, 70(11): 114301. doi: 10.7498/aps.70.20210032
    [6] 许宏, 苑争一, 黄彤飞, 王啸, 陈正先, 韦进, 张翔, 黄元. 层状材料褶皱对几种地质活动机理研究的启示. 物理学报, 2020, 69(2): 026101. doi: 10.7498/aps.69.20190122
    [7] 徐艳, 王培光, 杨青, 董江涛. 时空相关多通道聚类的运动目标检测. 物理学报, 2019, 68(16): 164203. doi: 10.7498/aps.68.20190161
    [8] 王彬, 冯雅辉, 王秋实, 张伟, 张丽娜, 马晋文, 张浩然, 于广辉, 王桂强. 化学气相沉积法制备的石墨烯晶畴的氢气刻蚀. 物理学报, 2016, 65(9): 098101. doi: 10.7498/aps.65.098101
    [9] 覃业宏, 唐超, 张春小, 孟利军, 钟建新. 硅晶体表面石墨烯褶皱形貌的分子动力学模拟研究. 物理学报, 2015, 64(1): 016804. doi: 10.7498/aps.64.016804
    [10] 余森江. 硅油基底上受限金属薄膜自组装褶皱的原子力显微镜研究. 物理学报, 2014, 63(11): 116801. doi: 10.7498/aps.63.116801
    [11] 金左轮, 韩静, 张毅, 柏连发. 基于纹理显著性的微光图像目标检测. 物理学报, 2014, 63(6): 069501. doi: 10.7498/aps.63.069501
    [12] 吴瑕, 陈建文, 鲍拯, 郭德阳. 火箭尾焰对高频回波的影响及其频域能量凝聚检测方法. 物理学报, 2014, 63(11): 119401. doi: 10.7498/aps.63.119401
    [13] 高文, 汤洋, 朱明. 复杂背景下目标检测的级联分类器算法研究. 物理学报, 2014, 63(9): 094204. doi: 10.7498/aps.63.094204
    [14] 侯旺, 于起峰, 雷志辉, 刘晓春. 基于分块速度域改进迭代运动目标检测算法的红外弱小目标检测. 物理学报, 2014, 63(7): 074208. doi: 10.7498/aps.63.074208
    [15] 行鸿彦, 龚平, 徐伟. 海杂波背景下小目标检测的分形方法. 物理学报, 2012, 61(16): 160504. doi: 10.7498/aps.61.160504
    [16] 许元男, 赵远, 刘丽萍, 张宇, 孙秀冬. 基于伪Wigner-Ville分布和Rényi熵的显著图目标检测. 物理学报, 2010, 59(2): 980-988. doi: 10.7498/aps.59.980
    [17] 何四华, 杨绍清, 石爱国, 李天伟. 基于图像区域Lyapunov指数的海面舰船目标检测. 物理学报, 2009, 58(2): 794-801. doi: 10.7498/aps.58.794
    [18] 姜 斌, 王宏强, 黎 湘, 郭桂蓉. 海杂波背景下的目标检测新方法. 物理学报, 2006, 55(8): 3985-3991. doi: 10.7498/aps.55.3985
    [19] 郏东耀, 丁天怀. 皮棉杂质透射检测及成像目标增强方法. 物理学报, 2005, 54(9): 4058-4064. doi: 10.7498/aps.54.4058
    [20] 马兆勉, 陶纯堪. 区域分形与人工目标检测. 物理学报, 1999, 48(12): 2202-2207. doi: 10.7498/aps.48.2202
计量
  • 文章访问数:  5015
  • PDF下载量:  104
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-10-28
  • 修回日期:  2021-12-02
  • 上网日期:  2022-01-26
  • 刊出日期:  2022-04-05

/

返回文章
返回