JP7752864B2 - Tunnel face warning system - Google Patents

Description

This invention relates to technology that detects risks such as collapse by detecting instantaneous changes in the tunnel excavation surface (face) in real time.

In tunnel construction, the tunnel excavation face (face) may collapse during excavation work, which may lead to a serious disaster. It is known that the face may suddenly displace by several millimeters to several centimeters before a collapse. Therefore, a crack detection device is known that detects the risk of face collapse during excavation work in real time and prevents disasters (see Patent Document 1).
The crack detection device in Patent Document 1 focuses on the fact that when "mirror-sprayed concrete" is applied to a tunnel face, cracks of a few millimeters in width appear on the sprayed surface as a sign of impending collapse, and collapse occurs along the cracks a few minutes later, and is able to detect these cracks. Crack detection is performed by inputting the tunnel face image to be judged into a trained model that has learned the relationship between the tunnel face image and pixels corresponding to cracks contained in the tunnel face image.
However, the crack detection device of Patent Document 1 has a problem in that it cannot achieve sufficient accuracy because it detects cracks using only the face image captured by the camera.

In recent years, LiDAR (Light Detection and Ranging) has been used in autonomous driving systems for automobiles and other vehicles. LiDAR measures distance by emitting laser light from a sensor and measuring the time it takes for the light to hit an object and bounce back.
Therefore, a technology is known that attempts to detect slight sudden displacements, which are a sign of collapse, in real time at mountain tunnel excavation construction sites using a high-speed 3D laser scanner (see Non-Patent Document 1 ). The technology in Non-Patent Document 1 uses LiDAR as the high-speed 3D laser scanner, but since it detects the risk of collapse etc. using only LiDAR, there is a problem that sufficient accuracy cannot be obtained.

Japanese Patent Application Laid-Open No. 2021-184163

Takahiro Sanpei, Tomohiro Mizoguchi, "Basic Study on Real-time Detection of Sudden Displacement of Tunnel Face Using 3D Scanner", Proceedings of the 2017 Autumn Meeting of the Japan Society for Precision Engineering, pp.527-528

In light of this situation, the present invention aims to provide a system that can detect instantaneous changes in the tunnel excavation surface (face) in real time with high accuracy and that also has an alarm function.

To solve the above problems, the tunnel face warning system of the present invention comprises a camera means for capturing images of the tunnel face in real time, a three-dimensional distance sensor for measuring the three-dimensional position of the tunnel face in real time, an analysis unit for calculating the amount of extrusion from the previous and current positions in the time series of the three-dimensional position of the tunnel face input from the three-dimensional distance sensor at the timing when a change is detected from the difference between the previous and current images in the time series of the tunnel face images input from the camera means, outputting the collapse timing and collapse amount, determining whether the collapse amount exceeds a predetermined threshold and outputting an alarm signal if it is determined that the predetermined threshold has been exceeded, and an alarm means for receiving the alarm signal from the analysis unit and issuing an alarm.

By using a camera means and a 3D ranging sensor in combination, changes in the tunnel excavation surface (face) (signs of collapse) are captured by analyzing images captured by the camera means, and the 3D ranging sensor is configured to calculate the specific amount of extrusion of the face, instantaneous changes in the tunnel excavation surface (face) can be detected in real time with high accuracy. In addition, the alarm means can properly communicate the condition of the face to workers at the construction site, preventing the occurrence of serious disasters. Rotating lights and buzzers are preferably used as alarm means.
Here, the previous image in the time series of images of the face input from the camera means does not mean the image one frame before in the group of images taken in real time by the camera means, but an image in a steady state with no change, such as an image at the start of measurement. Also, the collapse timing refers to the timing at which a change is detected from the difference between the previous image and the current image in the time series of images of the face, and specifically, an artificial intelligence model (AI model) analyzes the images and makes the judgment.
Furthermore, the previous position on the time series of the three-dimensional position of the working face input from the three-dimensional distance sensor is the position of the working face when there is no change and it is steady, and the amount of push-out calculated from the previous position and the current position is the amount of movement (distance) of each point cloud. The amount of push-out is calculated from the amount of movement of the point cloud, and the amount of collapse (volume) is calculated from the area of the moved point cloud and the amount of push-out. It is a sign of collapse regardless of whether a collapse has actually occurred or not.

In the tunnel face warning system of the present invention, the analysis unit is preferably equipped with a learning model that identifies the timing of collapse in images of the face, and detects the timing of collapse from the difference between the previous image and the current image in the time series of images of the face.
The analysis unit is equipped with an artificial intelligence model (AI model), i.e., a learning model, that predicts the timing of a collapse from changes in images of the face, making it possible to predict the timing of a collapse.

In the tunnel face warning system of the present invention, it is preferable that the analysis unit calculates the collapsed area of the tunnel face based on the amount of push-out calculated from the previous and current positions on the time series of the three-dimensional position of the tunnel face input from the three-dimensional distance sensor, and determines whether or not the collapsed area exceeds a predetermined area and outputs a second warning signal. With this configuration, it is also possible to output a different warning if the collapsed area is large.

It is preferable that the tunnel face warning system of the present invention further comprises a display means for inputting and displaying a group of time-series images of the tunnel face including the timing of collapse and the amount of collapse from the analysis unit.
By providing the display means, a user can visually check the condition of the face and the degree of danger. The display means may further display data input from the three-dimensional distance measuring sensor and data related to warnings.

The tunnel face warning system of the present invention has the advantage of being able to detect instantaneous deformations in the tunnel excavation surface (face) in real time with high accuracy, and also being able to alert those in the vicinity to the risk of collapse through its warning function.

Functional block diagram of the tunnel face warning system of the first embodiment Configuration image of the tunnel face warning system of Example 1 Schematic flow diagram of the tunnel face warning system of Example 1 Display image Functional block diagram of a tunnel face warning system according to a second embodiment

An example of an embodiment of the present invention will be described in detail below, with reference to the drawings. Note that the scope of the present invention is not limited to the following examples and illustrated examples, and many modifications and variations are possible.

Fig. 1 shows a configuration image diagram of a tunnel face warning system of Example 1. As shown in Fig. 1, the tunnel face warning system 1 includes a camera means 2, a three-dimensional distance measuring sensor 3, an analysis unit 4, a warning means 5, and a display means 6, and the analysis unit 4 includes a learning model 41 and a determination means 42.
The camera means 2 captures images of the face in real time, and a wide range of known cameras can be used. The three-dimensional ranging sensor 3 measures the three-dimensional position of the face in real time, allowing the amount of extrusion of the mirror surface to be grasped in real time. A known LiDAR scanner is preferably used as the three-dimensional ranging sensor 3. The analysis unit 4 calculates the amount of extrusion from the previous and current positions in the time series of the three-dimensional position of the face input from the three-dimensional ranging sensor 3 at the timing when a change is detected from the difference between the previous and current images in the time series of the face images input from the camera means 2, outputs the collapse timing and collapse amount, determines whether the collapse amount exceeds a predetermined threshold, and outputs an alarm signal if it is determined that the predetermined threshold has been exceeded.

The learning model 41 identifies the timing of collapse in images of the tunnel face, and uses the learning model 41 to detect the timing of collapse from the difference between the previous and current images in the time series of images of the tunnel face. In other words, the learning model 41 is not used to compare the three-dimensional position of the tunnel face measured by the three-dimensional ranging sensor 3, but is only used to compare images of the tunnel face captured by the camera means 2. The determination means 42 determines whether the amount of collapse exceeds a predetermined threshold. The alarm means 5 receives an alarm signal from the analysis unit 4 and issues an alarm, and is composed of, for example, a rotating light or buzzer. The display means 6 receives and displays a group of images of the tunnel face in a time series, including the timing of collapse, and the amount of collapse from the analysis unit 4; a known display is preferably used. The display means 6 can visualize the amount of collapse, the amount of extrusion, display contours, highlight weak parts of the tunnel face, and perform other displays. A touch panel may also be used for not only displaying the data but also starting, stopping, changing settings, and other operations.

The camera means 2 compares the most recent images from the start to the end of the camera footage, and the analysis unit 4 uses the learning model 41 to make judgments based on these, detecting the timing of collapse from the difference between the previous and current images in the time series of face images. The camera means 2 can also judge weathering, extract weak areas, and determine the presence or absence of spring water or cracks.
Meanwhile, the three-dimensional ranging sensor 3 measures the three-dimensional position of the working face in real time and inputs the data to the analysis unit 4. When the analysis unit 4 detects the timing of a collapse using the learning model 41, it calculates the amount of push-out from the previous and current positions in the time series of the three-dimensional position of the working face input from the three-dimensional ranging sensor 3, and calculates the amount of collapse from the amount of push-out. The determination means 42 determines whether the amount of collapse exceeds a predetermined threshold, and if it is determined that the predetermined threshold has been exceeded, it can output an alarm signal, issue an alarm using the alarm means 5, or display on the display means 6. By using the camera means 2 and the three-dimensional ranging sensor 3 in combination, it is possible to accurately determine the risk of collapse and issue an alarm.

Fig. 2 shows a configuration image diagram of the tunnel face warning system of Example 1. As shown in Fig. 2, the tunnel face warning system 1 is installed and used inside a tunnel excavation work site 7. The tunnel face warning system 1 is composed of a camera 20 and a device main body 10, and the camera 20 and the device main body 10 are connected by a communication cable 11. Alternatively, the camera 20 and the device main body 10 may be configured to communicate wirelessly.
The camera 20 is equipped with a known camera as the camera means 2 and a LiDAR scanner as the three-dimensional distance measuring sensor 3. The device main body 10 is provided with a rotating light 50a and a buzzer (not shown) as the alarm means 5, and a buzzer sound is emitted from a speaker 50b. The device main body 10 is also provided with a display 60 as the display means 6.

Figure 4 shows an image of the display. As shown in Figure 4, an image 61 captured in real time of the tunnel excavation site 7 is displayed on the display 60. Here, the tunnel face 71 is the monitored object, and a monitored area 72 is displayed. In addition to images captured in real time, images in steady state may also be displayed on the display 60, and three-dimensional position data of point clouds in steady state and three-dimensional position data of point clouds measured in real time may also be displayed on the screen. The display 60 is a touch panel, and the monitored area 72 can be changed by operating the cross key 62a, or the image can be enlarged or reduced by touching the enlarge button 62b or reduce button 62c.

When using the tunnel face warning system of the first embodiment, first, as a prerequisite, the face positions and installation orientations of the camera means 2 and the three-dimensional distance measuring sensor 3 are registered.
FIG. 3 shows a schematic flow diagram of the tunnel face warning system of the first embodiment. As shown in FIG. 3, first, steady-state image data of the tunnel face 71 is input from the camera means 2 to the analysis unit 4 (step S01). Next, three-dimensional position data of the point cloud of the tunnel face 71 in steady state is input from the three-dimensional ranging sensor 3 to the analysis unit 4 (step S02). The camera means 2 captures images of the tunnel face in real time, and inputs the image data of the tunnel face 71 from the camera means 2 to the analysis unit 4 (step S03). The input images are displayed on the display means 6. When the analysis unit 4 detects a change from the difference between the previous image and the current image in the time series using the learning model 41, that is, when a change is detected from the difference between the steady-state image data and the image data captured and input in real time (step S04), the three-dimensional position data of the point cloud of the tunnel face 71 is input to the analysis unit 4 (step S05). On the other hand, if no change is detected from the difference between the previous image and the current image in the time series, that is, if no change is detected from the difference between the image data in a steady state and the image data captured and input in real time (step S04), image data of the face 71 is input again from the camera means 2 (step S03).

The analysis unit 4 calculates the amount of push-out from the previous and current positions on the time series of the three-dimensional position of the working face 71 input from the three-dimensional distance measurement sensor 3, i.e., the three-dimensional position data of the point cloud in a steady state, and the three-dimensional position data of the point cloud measured and input in real time (step S06). The analysis unit 4 calculates the amount of collapse from the amount of push-out, and outputs the collapse timing and amount to the determination means 42 (step S07).
The determining means 42 determines whether the amount of collapse has exceeded a predetermined threshold and determines the risk of collapse (step S08). If it is determined that the amount of collapse has exceeded the predetermined threshold, i.e., if it is determined that the alarm level has been exceeded (step S09), an alarm signal is output (step S10). On the other hand, if it is determined that the alarm level has not been exceeded (step S09), image data of the working face 71 is again input from the camera means 2 (step S03).

Based on the output alarm signal, the alarm means 5 issues an alarm using a rotating light 50a or a buzzer. The display means 6 displays the output data on the timing and amount of collapse on a screen, and also displays the result of the determination made by the determination means 42.
In this way, by using the camera means 2 and the three-dimensional distance measuring sensor 3 together, instantaneous deformation of the face can be detected in real time with high accuracy.

In addition to the above flow, it is also possible to calculate the collapsed area of the face from the amount of extrusion and output an alarm signal. Specifically, the collapsed area of the face 71 is further calculated based on the amount of extrusion calculated in step S06, and in step S07, the collapsed area is output in addition to the collapse timing and amount. Then, in step S08, it is determined whether the collapsed area exceeds a predetermined area, and if it is determined that the predetermined area has been exceeded, i.e., if it is determined that the alarm level has been exceeded (step S09), a second alarm signal is output in step S10.

Fig. 5 shows a configuration image diagram of a tunnel face warning system of Example 2. As shown in Fig. 5, a tunnel face warning system 1a includes a camera means 2, a three-dimensional distance measuring sensor 3, an analysis unit 4, and a warning means 5, and the analysis unit 4 includes a learning model 41 and a determination means 42.
Unlike the tunnel face warning system 1 of Example 1, the tunnel face warning system 1a is configured without a display means 6. By adopting such a configuration, the system can be constructed at lower cost and can be used as a simple warning system.

This invention is useful for systems that can grasp instantaneous changes in the tunnel excavation surface in real time and detect risks such as collapse.

REFERENCE SIGNS LIST 1, 1a Tunnel face warning system 2 Camera means 3 Three-dimensional distance measuring sensor 4 Analysis unit 5 Warning means 6 Display means 7 Tunnel excavation construction site 10 Device body 11 Communication cable 20 Camera 41 Learning model 42 Determination means 50a Rotating light 50b Speaker 60 Display 61 Image 62a Cross key 62b Zoom in button 62c Zoom out button 71 Face 72 Monitoring target range

Claims (3)

a camera means for capturing images of the working face in real time;
a three-dimensional distance measuring sensor that measures the three-dimensional position of the working face in real time;
an analysis unit that is equipped with a learning model that identifies the timing of collapse of the image of the face, and that , at the timing that the collapse timing is detected from the difference between the previous image and the current image in the time series of the image of the face input from the camera means, calculates the amount of push-out from the amount of movement of the point cloud between the previous position and the current position in the time series of the three-dimensional position of the face input from the three-dimensional distance measuring sensor, calculates the amount of collapse from the area of the moved point cloud and the amount of push-out, outputs the collapse timing and the amount of collapse , determines whether the amount of collapse has exceeded a predetermined threshold, and outputs an alarm signal if it is determined that the amount of collapse has exceeded the predetermined threshold;
an alarm means for receiving the alarm signal from the analysis unit and issuing an alarm;
A tunnel face warning system comprising: The tunnel face warning system described in claim 1, characterized in that the analysis unit calculates the collapse area of the face based on the area of the point cloud calculated from the movement of the point cloud between the previous position and the current position on the time series of the three-dimensional position of the face input from the three-dimensional ranging sensor, determines whether the collapse area exceeds a predetermined area, and outputs a second warning signal. A tunnel face warning system as described in claim 1 or 2 , further comprising a display means for inputting and displaying a group of images of the face in time series including the timing of collapse and the amount of collapse from the analysis unit.