.png){kind=icon}
.png){kind=icon}
Return
Research on automatic identification and measurement technology of tunnel rock settlement
.png){kind=icon}
PRE
AI
S
Y
Y
G
R
DOI：10.1177/00202940261417228.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Aiming at the problem of tunnel settlement monitoring in complex construction environments, a visual tunnel settlement monitoring system is developed. Due to the poor lighting conditions inside the tunnel, direct measurement of the tunnel's surrounding rock settlement was challenging. It was necessary to install targets at the monitoring points to indirectly reflect the surrounding rock settlement through the displacement of these targets. Consequently, a self-luminous hanging target specifically designed for engineering tunnels was developed. As the main feature, the two-digit digital tube of the target, became a prerequisite for realizing the tunnel surrounding rock detection targets. Then, a high-precision industrial camera is utilized to collect target images, and then a tunnel target dataset is constructed. Additionally, the improved RT-DETR target detection model is proposed for target recognition. To further increase the model's ability to capture target features, the EfficientVit is introduced as the backbone network, the attention mechanism ECA-Net is added, and the loss function is improved to SIoU loss to enhance the model's application efficiency in the industrialized direction. The target displacement is calculated by the ratio between the image pixel value and the measured value, which indirectly reflects the settlement of the tunnel. Finally, the interactive visual interface of the tunnel settlement monitoring platform is also developed to further display various information in the monitoring process. The measurement error of the proposed method was evaluated in a controlled laboratory environment. Within a tunnel viewing distance of 25 m, the relative error was found to be less than 8.7%. A small-scale comparison with total station measurements was conducted, demonstrating that the proposed algorithm meets the accuracy requirements for practical engineering applications.
Keywords:
tunnel rock
settlement monitoring
object detection
deep learning
attention mechanism
Journal
M
IF:
2
Papers:
52
Citations:
0
Organization
Cited Papers