1.MOE Key Lab of Transportation Tunnel Engineering,Southwest Jiaotong University,Chengdu 610031,China;
2.National Engineering Research Center of Geological Disaster Prevention Technology in Land Transportation,Southwest Jiaotong University,Chengdu 610031,China
Abstract：Aiming at the demand for monitoring model box displacement, surrounding rock movement, and tunnel lining deformation in vibration table experiments, a self-correction technology for tilted images is proposed by combining Yolo-V2, improved Canny edge detection, Hough transform, and tilt correction technology. The technology corrects the tilted image to an image where the camera axis is perpendicular to the motion plane of the measuring point for analysis, solving the problem caused by the limitations of the size of the vibration table and the tunnel lining model, which makes it difficult to set up the camera position vertically above the motion plane of the measuring point for shooting, affecting the accuracy of the subsequent processing results, and improving the field of use for target tracking algorithms in structural displacement and deformation monitoring. At the same time, the tracking of the corrected image is completed through the kernel correlation filtering technology to form a complete monitoring system. The system was used for the seismic action test of a fault-crossing tunnel's toughened structure with earthquake resistance and anti-seismic performance. The displacement of the model box, the movement of the surrounding rock, and the deformation of the tunnel lining were monitored. The reliability of the system was verified by comparing the data collected using high-precision wire-type displacement meters and displacement sensors. The results show that when there are markers at the monitoring points, and after correcting the acquired results for the tilted image, the 2-norm of the monitoring results compared with the displacement sensor is less than 0.005; when it is difficult to post markers, the precision of the tracking of the surface feature on the structure reaches 97.5%. The experimental results prove that the automatic correction technology for tilted images has high reliability in tracking images without markers for non-standard labels.
常铭宇1,2，申玉生1,2，张熙1,2，高登1,2，骆阳1,2，王浩鱇1,2. 基于数字图像技术振动台试验结构位移信息捕捉技术研究[J]. 振动与冲击, 2024, 43(5): 139-148.
CHANG Mingyu1,2,SHEN Yusheng1,2,ZHANG Xi1,2,GAO Deng1,2,LUO Yang1,2,WANG Haokang1,2. Capturing technology for displacement information of shaking table test structure based on digital image technique. JOURNAL OF VIBRATION AND SHOCK, 2024, 43(5): 139-148.
[1] 高春华,纪金豹,闫维明等. 地震模拟振动台技术在中国的发展[J].土木工程学报,2014,47(08):9-19
Gao Chunhua, Ji Jinbao, Yan Weiming,et al Developments of shaking table technology in China [J]. Civil Engineering Journal,2014, 47(8): 9－19.
[2] 叶肖伟,董传智. 基于计算机视觉的结构位移监测综述[J]. 中国公路学报,2019,32(11):21-39.
YE Xiaowei, DONG Chuanzhi. Review of structural displacement monitoring based on computer vision[J]. China Journal of Highway and Transport, 2019，32(11)：21–39.
[3] 晏班夫,陈泽楚,朱子纲. 基于非接触摄影测量的拉索索力测试[J]. 湖南大学学报(自然科学版),2015,42(11):105-110.
Yan Banfu, CHEN Zechu, ZHU Zigang. Cable force identification Based on non-contact photogrammetry system[J]. Journal of Hunan University (Natural Sciences), 2015, 42(11): 105-110.
[4] 邵新星,钱帅宇,王文波. 基于自标定数字图像相关技术的榫卯构件大面积全场变形测量[J]. 东南大学学报(自然科学版),2018,48(02):337-341.
SHAO XX, QIAN S Y. Larae-area ful-field deformation measurement of mortise-enon structures based on self-calibration diaital image corelation technologyl[J]. Journal of Southeast University(Natural Science Edition).2018, 48(2): 337-341.
[5] 周云,程依婷. 基于数字图像相关理论的非接触式结构位移测量方法[J]. 湖南大学学报(自然科学版),2021,48(05):1-9.
ZHOU Yun, CHENG Yiting. Non-contact structural displacement measurement based on digital image correlation method[J]. Journal of Hunan University：Natural Science, 2021, 48(5): 1-9.
[6] 杨娜,汪德佳,李天昊. 基于计算机视觉方法的古建筑变形监测[J]. 建筑结构学报,2023,44(01):192-202.
Yang Na, Wang Dejia, Li Tianhao. Deformation monitoring of an Cient buildings based on computer vision method [J/OL]. Journal of Building Structures, 2023, 44(1): 192-202.
[7] 赵永强,饶元,董世鹏,张君毅. 深度学习目标检测方法综述[J]. 中国图像图形学报，2020，25(04): 629-654.
Zhao Y Q, RAO Y, Dong S P. Survey on deep learning object detection. Journal of Image and Graphics, 25(4): 629-654.
[8] 周登科,杨颖,朱杰. 基于深度学习的指针式仪表倾斜校正方法[J]. 计算机辅助设计与图形学学报,2020,32(12):1976-1984.
Zhou D K, Yang Y, Zhu J. Tilt Correction Method of Pointer Meter Based on Deep Learning [J], Journal of Computer-Aided Design & Computer Graphics, 2020, 32(12): 1976-1984
[9] 张微,康宝生. 相关滤波目标跟踪进展综述[J]. 中国图象图形学报,2017,22(08):1017-1033.
Zhang W, Kang B SH. Overview of the progress of correlated filtering target track[J]. Journal of Image and Graphics, 2017, 22(8): 1017-1033.
[10] 张熙,申玉生,周鹏发. 基于能量原理的隧道结构地震影响区划分及节段式衬砌响应特性（英文）[J].Journal of Central South University,2023,30(01):227-242.
Zhang X, Shen Y SH, Zhou P F. Partions of seismic affected zone and characteristics of segmental lining based on energy principle[J]. Journal of Central South University, 2023, 30(01): 227-242.
[11] 倪麟, 龚劬, 曹莉,等. 基于自适应加权中止滤波的二维Otsu图像分割算法[J]. 计算机应用研究, 2013, 30(2): 598-600.
Ni Lin, Gong Qu, Cao Li, et al. Two-dimensional Otsu image seggmential algorithm based on adaptive weighted median filter[J]. Aoolication Research of Computers, 2013, 30(2): 598-600.
[12] CHEN J，SHI X，LI J. Shaking table test of utility tunnel under
non-uniform earthquake wave excitation[J]. Soil Dynamics and Earthquake Engineering，2010，30(11)：1400–1416.
[13] 陈红娟，李小军，闫维明，等. 锯末混合土场地模型振动台试验研究[J]. 岩土工程学报，2017，39(11)：1–8.
CHEN Hongjuan，LI Xiaojun，YAN Weiming，et al. Study on shaking table test of sawdust mixed clay site model[J]. Chinese Journal of Geotechnical Engineering，2017，39(11)：1–8.(in Chinese)
[14] 张涛，高波，范凯祥，等. 地震作用下振动台试验刚性模型箱侧壁柔性材料研究[J]. 岩石力学与工程学报，2018，37(10) ： 2415 –2424.
ZHANG Tao， GAO Bo， FAN Kaixiang， et al. Study on flexible material in the sidewall of rigid model box in shaking table test[J].Chinese Journal of Rock Mechanics and Engineering， 2018，37(10)：2415 – 2424.(in Chinese)
[15] 禹海涛，袁 勇. 长大隧道地震响应分析与试验方法新进展[J]. 中国公路学报，2018，31(10)：19–35.
YU Haitao，YUAN Yong. Review on seismic response analysis and test methods for long and large tunnels[J]. China Journal of Highway and Transport，2018，31(10)：19–35.(in Chinese))
[16] 禹海涛, 张敬华, 袁 勇, 等. 盾构隧道-工作井节点振动
台试验设计与分析[J]. 中国公路学报, 2017, 30(8): 183–YU Hai-tao, ZHANG Jing-hua, YUAN Yong, et al. Design and analysis of shaking table test on shield tunnel-shaft junction[J]. China Journal of Highway and Transport, 2017, 30(8): 183–192. (in Chinese)
[17] 吴佳奇,孙华生.一种倾斜影像几何纠正的有效方法[J].遥感技术与应用,2015,30(05):1006-1011.
.J.Wu, H.Sun. Efficient Method of Oblique Image Correction, Remote Sensing Technology and Application, 30, 5(2015)
[18] 瞿洋,杨利平.Hough变换OCR图象倾斜矫正方法[J].中国图象图形学报,2001(02):78-81.
Qu Yang, Yang Liping. Hough transform OCR image slantcorection method [J]. Journal of Image and Graphices, 2001, 6(2):178-181(in Chinese)
[19] 周登科,杨颖,朱杰.基于深度学习的指针式仪表倾斜校正方法[J].计算机辅助设计与图形学学报,2020,32(12):1976-1984.
ZHOU D K,YANG Y, ZHU J,et al. Tilt correction method of pointer meter based on deep learning [J]. Journal of Computer-Aided Design & Computer Graphics, 2020, 32(12):1976-1984.
