一种基于YOLOv4的隧道吊柱位姿检测方法
吴越舟，邓斌，李荣铎，许冠麟
（西南交通大学 机械工程学院，四川 成都 610031）
摘要：针对隧道内吊柱实现自动抓取安装需要获取吊柱位姿信息的问题，提出一种将YOLOv4目标检测算法与多特征点提取算法融合的吊柱位姿检测算法。通过YOLOv4算法在双目视觉的左右图像上标出吊柱法兰的识别框，在扩大后的识别框范围内将图像从三原色（RGB）空间转换到HSV空间，利用检测目标的特有HSV信息分离出其区域，对该区域进行直线检测，通过canny和霍夫变换筛选出吊柱法兰的特征点。利用视差原理与面面交会原理还原出吊柱相对于相机的位姿信息。在不同距离进行了多组位姿检测实验，结果显示，检测速度为26 f/s，在1000～1500 mm内平均误差均在20 mm以内，具有较高的精确性和实时性。
关键词：隧道吊柱；YOLOv4；位姿检测；双目视觉
中图分类号：TP391.41 文献标志码：A doi：10.3969/j.issn.1006-0316.2023.12.002
文章编号：1006-0316 (2023) 12-0008-07
Pose Detection Method of Tunnel Davit Based on YOLOv4
WU Yuezhou，DENG Bin，LI Rongduo，XU Guanlin
( School of Mechanical Engineering, Southwest Jiaotong University, Chengdu 610031, China )
Abstract：In order to obtain the pose information of the davit to realize the automatic grasping and installation of the davit in the tunnel, this paper proposes a davit pose detection algorithm that combines YOLOv4 target detection algorithm and multi-feature extraction algorithm. The identification frame of the flange of the davit on the left and right images of binocular vision is marked through the YOLOv4 algorithm, and the image is converted from the three primary colors (RGB) space to the HSV space within the expanded identification frame range, and the specific HSV information of the detected object is used to separate its area and conduct the line detection on the area. And the features of the flange of the davit are selected through Canny and Hough Transform. Finally, the pose information of the davit relative to the camera is restored by using the parallax principle and the surface to surface intersection principle. Several groups of pose detection experiments are carried out at different distances. The results show that the detection speed is 26 f/s, and the average error within 1000~1500 mm is within 20 mm, which has high accuracy and real-time performance.
Key words：tunnel davit；YOLOv4；pose detection；binocular vision
———————————————
收稿日期：2023-03-02
作者简介：吴越舟（1998－），男，四川南充人，硕士研究生，主要研究方向为机器视觉，E-mail：934712683@qq.com；邓斌（1964－），男，湖北荆门人，博士研究生，教授，主要研究方向机电液一体化。
（西南交通大学 机械工程学院，四川 成都 610031）
摘要：针对隧道内吊柱实现自动抓取安装需要获取吊柱位姿信息的问题，提出一种将YOLOv4目标检测算法与多特征点提取算法融合的吊柱位姿检测算法。通过YOLOv4算法在双目视觉的左右图像上标出吊柱法兰的识别框，在扩大后的识别框范围内将图像从三原色（RGB）空间转换到HSV空间，利用检测目标的特有HSV信息分离出其区域，对该区域进行直线检测，通过canny和霍夫变换筛选出吊柱法兰的特征点。利用视差原理与面面交会原理还原出吊柱相对于相机的位姿信息。在不同距离进行了多组位姿检测实验，结果显示，检测速度为26 f/s，在1000～1500 mm内平均误差均在20 mm以内，具有较高的精确性和实时性。
关键词：隧道吊柱；YOLOv4；位姿检测；双目视觉
中图分类号：TP391.41 文献标志码：A doi：10.3969/j.issn.1006-0316.2023.12.002
文章编号：1006-0316 (2023) 12-0008-07
Pose Detection Method of Tunnel Davit Based on YOLOv4
WU Yuezhou，DENG Bin，LI Rongduo，XU Guanlin
( School of Mechanical Engineering, Southwest Jiaotong University, Chengdu 610031, China )
Abstract：In order to obtain the pose information of the davit to realize the automatic grasping and installation of the davit in the tunnel, this paper proposes a davit pose detection algorithm that combines YOLOv4 target detection algorithm and multi-feature extraction algorithm. The identification frame of the flange of the davit on the left and right images of binocular vision is marked through the YOLOv4 algorithm, and the image is converted from the three primary colors (RGB) space to the HSV space within the expanded identification frame range, and the specific HSV information of the detected object is used to separate its area and conduct the line detection on the area. And the features of the flange of the davit are selected through Canny and Hough Transform. Finally, the pose information of the davit relative to the camera is restored by using the parallax principle and the surface to surface intersection principle. Several groups of pose detection experiments are carried out at different distances. The results show that the detection speed is 26 f/s, and the average error within 1000~1500 mm is within 20 mm, which has high accuracy and real-time performance.
Key words：tunnel davit；YOLOv4；pose detection；binocular vision
———————————————
收稿日期：2023-03-02
作者简介：吴越舟（1998－），男，四川南充人，硕士研究生，主要研究方向为机器视觉，E-mail：934712683@qq.com；邓斌（1964－），男，湖北荆门人，博士研究生，教授，主要研究方向机电液一体化。
设为首页 | 加入收藏 | 免责条款
《机械》杂志版权所有 Copyright©2008-2012 Jixiezazhi.com All Rights Reserved
电话：028-85925070 传真：028-85925073 E-mail：jixie@vip.163.com
地址：四川省成都锦江工业开发区墨香路48号 邮编：610063
Powered by PageAdmin CMS