摘要
在隧道施工过程中,岩体参数的获取是实现隧道掘进机参数调整和智能决策的前提,因此,要对掘进过程中获得的岩块进行采样和检测,而岩块识别和检测目前主要由人工完成。本文针对岩块的自动化识别和检测问题,提出了一种基于机器视觉的岩块自动化识别检测方法,通过融合岩块区域检测和语义分割算法能够快速准确获取岩块的形心坐标和过形心最小直径。首先,使用YOLOv3网络对岩块进行识别,实现岩块的区域检测。其次,针对每个区域的岩块采用FCN-DenseNet网络进行语义分割和图像处理,并对全卷积神经网络进行改进,减少了语义分割模型的参数量,提高了语义分割效率,提升了岩块轮廓获取的精度和速度。最后,根据获得的岩块轮廓点,计算其形心坐标及过形心的最小直径,为机械臂抓取和岩块点荷载强度的计算提供支持。搭建实验平台,完成机械臂手眼标定和深度相机坐标下岩块图像与岩块点云对齐,获取岩块形心坐标在机械臂坐标下的位置。实验结果表明,本文所提算法能够快速准确地获取岩块的形位参数,对10次实验中的102块岩块识别检测成功率为91.18%,在所有完成识别检测岩块中的吸取成功率为92.47%,可以应用于岩体的自动化检测,提高岩体检测的效率和智能化水平。
In the process of tunnel construction,the acquisition of rock mass parameters is the premise of parameter adjustment of tunnel boring machine and intelligent decision-making,so necessary to sample and detect the rock blocks in tunneling.However,the recognition and detection of rock blocks are mainly done manually at present.In order to solve the problem of automatic recognition and detection of rock blocks,this paper presents an automatic recognition and detection of rock block parameters based on machine vision,which can quickly and accurately obtain the centroid coordinates and the minimum diameter of rock block through centroid by fusing rock block region detection and semantic segmentation algorithm.Firstly,the YOLOv3 network is used to identify the rock block and realize the region detection of the rock block.Then,FCN-DenseNet network is used for semantic segmentation and image processing of rock blocks in each area.By improving the total convolutional neural network,the number of parameters of semantic segmentation model is reduced,the efficiency of semantic segmentation is improved,and the accuracy and speed of rock block contour acquisition are also improved.Finally,the centroid coordinates and the minimum diameter passing through the centroid are calculated according to the contour points of the rock block,which provides support for grasping by the mechanical arm and calculating the load strength of the rock block point.The experimental platform is built,the hand-eye calibration of the robot arm is completed,the rock image is aligned with the point cloud of rock block under the coordinates of the depth camera,and the position of the rock centroid coordinates under the coordinates of the robot arm is obtained.The experimental results show that the proposed algorithm can quickly and accurately obtain the shape and position parameters of rock blocks,and the successful rate of recognition and detection of 102 rock blocks in 10 experiments is 91.18%,and the successful rate of suction of all rock blocks is 92.
作者
薛山
段岳飞
胡天亮
马嵩华
XUE Shan;DUAN Yuefei;HU Tianliang;MA Songhua(School of Mechanical Engineering,Shandong University,Jinan 250061,China;Key Laboratory of High Efficiency and Clean Mechanical Manufacture,Ministry of Education,Jinan 250061,China;National Experimental Teaching Demonstration Center of Mechanical Engineering,Jinan 250061,China)
出处
《中国矿业》
北大核心
2024年第6期129-136,共8页
China Mining Magazine
基金
泰山学者工程专项经费项目资助(编号:tsqn202211024)。
关键词
岩块识别
区域检测
语义分割
岩块定位
点云对齐
rock block recognition
area detection
semantic segmentation
rock block location
point cloud alignment
