基于计算机视觉的输电塔位移监测ROI关键点法
ROI Method for Displacement Identification of Power Transmission Tower Based on Computer Vision
张楷 1孙超 2刘家豪 2李玉学 3田利4
作者信息
- 1. 石家庄铁道大学土木工程学院 石家庄,050043;中国电建集团河北省电力勘测设计研究院有限公司 石家庄,050031;河北省电力勘测设计技术创新中心 石家庄,050031
- 2. 北京理工大学机械与车辆学院 北京,100081
- 3. 石家庄铁道大学土木工程学院 石家庄,050043
- 4. 山东大学土建与水利学院 济南,250100
- 折叠
摘要
为了实现输电塔远距离位移监测,同时满足低成本、无接触、易实施及精确度高等要求,结合输电塔的内轮廓特征和计算机视觉位移识别技术,提出感兴趣区域(region of interest,简称ROI)关键点法.首先,利用N近邻最小能量法进行ROI轮廓搜索提取,并与Harris角点检测算法相结合;其次,通过输电塔台架实验与灰度模板匹配法相比,ROI关键点法位移识别结果的平均误差、均方根误差分别降低了56%和45%,绝对误差小于5 mm和10 mm的准确率提高了61%和3%,计算效率提高了11倍,稳定性及抗噪性能较高;最后,在实验塔对比验证中,ROI关键点法的位移测量值与实际位移的差值百分比在0.0%～11.1%之间.结果表明,ROI关键点法在输电塔结构位移监测中具有较高的准确率、精细度、计算效率、稳定性及鲁棒性.
Abstract
In order to monitor the displacement of power transmission towers over long distances and to fulfill the requirements for cost-effectiveness,non-contact monitoring,ease of implementation and high precision,the ROI method has been developed.This method ingeniously integrates the inner contour features of power trans-mission towers with computer vision-based displacement recognition techniques.Firstly,N-nearest-neighbor&minimum-energy method is used to extract the region of interest(ROI)contour,combined with Harris corner detection algorithm.Secondly,comparing the displacement recognition results with the gray template matching method in the tower platform experiments,the average error and the root mean square error of the ROI method are reduced by 56%and 45%respectively,and the accuracy rates of absolute error less than 5 mm and 10 mm are increased by 61%and 3%.The computational efficiency is improved by 11 times,and the ROI method shows high stability and noise resistance.Finally,in actual tower tests,the percentage difference between the displacement measurement values of the ROI method and the actual displacement is between 0.0%and 11.1%.The results show that the ROI method has high accuracy,precision,computational efficiency,stability,and ro-bustness in the displacement monitoring of transmission tower.
关键词
计算机视觉/输电塔/位移监测/无标靶/N近邻最小能量法Key words
computer vision/power transmission tower/displacement identification/non-marker/N-nearest-neighbor&minimum-energy method引用本文复制引用
基金项目
国家自然科学基金资助项目(52178489)
出版年
2024
NETL
NSTL
万方数据