摘要
针对地下管道背景复杂、缺陷种类繁多、特征模糊的问题,提出了一种基于立体视觉感知的地下管道异物检测与测量方法。首先创新性地将双目视觉与深度学习技术深度融合,通过双目可见光相机采集图像,结合张氏标定法确定相机参数,利用极线校正消除误差。然后采用多层级特征引导的异物检测方法,对图像进行初步框选,再通过IGEV(Iterative Geometry Encoding Volume)网络进行立体匹配,生成双目视差图,快速定位异物轮廓及坐标。最后,将边缘坐标转换至世界坐标系,利用欧几里得距离计算公式获取异物的三维信息。实验结果表明,该方法在地下管道异物检测与测量领域表现出色,提供了一种创新的解决方案。
To address the issues of complex background,diverse defect types,and fuzzy features in underground pipelines,a foreign object detection and measurement method for underground pipelines based on stereo vision perception was proposed.First,binocular vision and deep learning technologies were innovatively integrated in depth.Images were collected by binocular visible-light camera,and the camera parameters were determined by Zhang’s calibration method,and the epipolar correction was used to eliminate errors.Then,a foreign object detection method guided by multi-level features was used to perform preliminary frame selection of the image,and then stereo matching was performed through iterative geometry encoding volume(IGEV)network to generate binocular disparity map,and the contour and coordinates of the foreign object were quickly located.Finally,the edge coordinates were converted into the world coordinate system,and the Euclidean distance calculation formula was used to obtain the three-dimensional information of the foreign object.Experimental results showed that this method may realize the detection and measurement of foreign objects in underground pipelines,providing an innovative solution.
作者
王鑫
柏晶晶
王瀚
冯兴明
倪一峰
WANG Xin;BAI Jingjing;WANG Han;FENG Xingming;NI Yifeng(Yancheng Power Supply Branch,State Grid Jiangsu Power Supply Co.,Ltd.,Yancheng 224000,China)
出处
《国外电子测量技术》
2025年第6期86-92,共7页
Foreign Electronic Measurement Technology
基金
国网江苏省电力有限公司科技项目(J2024052)。
关键词
地下管道异物检测
目标测量
立体视觉感知
detection of foreign objects in underground pipelines
object measurement
stereo vision perception
作者简介
王鑫,硕士,高级工程师。E-mail:19313141293@163.com。
