水利学报2024，Vol.55Issue(9)：1110-1122,13.DOI:10.13243/j.cnki.slxb.20240008
基于机器视觉的闸坝表面位移非接触式监测方法
Machine vision-based non-contact monitoring method for gate and dam surface displacements
摘要
Abstract
Aiming at the problems of conventional monitoring method such as high labor density,low monitoring frequency,and difficulty in achieving long-term stable monitoring of the surface displacement of locks and dams,a non-contact intelligent monitoring method integrating spatio-temporal features is proposed.The method adopts an artificial target as a marker,takes a camera as an acquisition device,transmits image information wirelessly,and makes use of Adaptive Gamma Correction Weighted Distribution(AGCWD)and Weighted Guided Image Filter(WGIF)with improved edge-awareness factor.WGIF with edge-awareness factor to enhance the feature expression ability of low illumination images,and the Spatio-Temporal Context(STC)algorithm based on Bayesian framework on board the computer to deeply mine the contextual spatio-temporal information of the target image,and further introduce surface fitting to obtain sub-pixel level displacement information of the target to achieve the sub-pixel displacement information of both horizontal and vertical bi-directional surface displacements of the gate and dam.Further,surface fitting is introduced to obtain subpixel-level displacement information to achieve non-pixel-level non-contact monitoring of horizontal and vertical bi-directional surface displacements of the lock and dam.The la-boratory and field test results show that the displacement monitoring data under different experimental scenarios are highly consistent with the calibration data,and the error is less than 0.05 mm;compared with the existing image optimization processing methods,the image optimization processing methods based on AGCWD and WGIF increase the Peak Signal-to-Noise Ratio(PSNR)by 2.70％,increase the information entropy by 4.91％,and reduce the standard deviation by 2.63％;compared with the established target tracking algorithms,the STC target tracking al-gorithm based on surface fitting is more effective in obtaining sub-pixel-level displacement information.Compared with the existing target tracking algorithms,the field monitoring data of the STC target tracking algorithm based on surface fitting improves the accuracy of the similar target tracking algorithms by 48％,which can provide a high-precision solution for the monitoring of the gate and dam surface displacements.关键词
闸坝/表面位移监测/图像序列时空特征/数字图像优化/亚像素级目标追踪Key words
lock and dam/surface displacement monitoring/spatio-temporal characteristics of image sequences/digital image optimisation/sub-pixel level target tracking分类
建筑与水利引用本文复制引用
陈波,何梦佳,刘伟琪,马聪..基于机器视觉的闸坝表面位移非接触式监测方法[J].水利学报,2024,55(9):1110-1122,13.基金项目
国家自然科学基金面上项目(52079049) （52079049）
国家自然科学基金重点项目(52239009) （52239009）
全国重点实验室基本科研业务费(522012272,5230248A2) （522012272,5230248A2）
