Face Safety Management System

Abstract

This system provides a tunnel face safety management system that estimates restricted access areas based on machine learning-based image recognition and uses these restricted areas to manage the safety of tunnel face operations. [Solution] The face safety management system 1 of the present invention comprises an image acquisition unit 10 that acquires a face front image A including the face surface A1, an area estimation unit 20 that estimates an access restriction area B within a predetermined range from the face surface A1 based on the face front image A, and an area display unit 30 that displays the face front image A and the access restriction area B. The area estimation unit 20 comprises a face estimation unit 21 and a distance calculation unit 22. The face estimation unit 21 estimates the face surface A1 from the face front image A based on image recognition by machine learning, and the distance calculation unit 22 applies the design cross section of the face surface A1 to the face front image A to calculate the distance from the face surface A1 to the tunnel entrance side, and estimates the area within a predetermined distance from the face surface A1 toward the tunnel entrance side as the access restriction area B.

Description

Technical Field

[0001] The present invention relates to a face safety management system, and particularly to a face safety management system that estimates a restricted access area based on image recognition by machine learning and manages the safety of the work in front of the face using the restricted access area.

Background Art

[0002] In mountain tunnel construction, a series of processes such as (1) drilling, (2) charging, (3) blasting, (4) mucking, (5) spraying, (6) erection of support works, (7) installation of rock bolts, etc. are regarded as one cycle, and the tunnel is excavated by repeating this cycle. Each of these processes is carried out in front of the face. However, during the work, if rock fragments peel off from the natural ground or the natural ground collapses, there is a risk that workers will be involved and a disaster (skin peeling disaster) will occur. The skin peeling disaster has a high severity when it occurs, with a mortality rate of 6% at the time of occurrence and 42% of the victims being forced to take more than one month off work. Therefore, in 2024, the "Guidelines on Measures for Preventing Skin Peeling Disasters at the Face of Mountain Tunnel Construction" (Non-Patent Document 1) of the Ministry of Health, Labour and Welfare was revised, defining the range within 45 degrees from the top edge of the face as an "area that requires special consideration", avoiding entry as much as possible, and when entry is unavoidable, thoroughly implementing safety measures such as wearing a back protector under the supervision of a supervisor, etc. were clearly stated. Based on the above background, at the construction site of mountain tunnels, an area within 45 degrees from the top edge of the face is demarcated with colored tape or the like, and a dedicated face supervisor is arranged for visual monitoring on a 24-hour basis (Non-Patent Document 2).

Prior Art Documents

Non-Patent Documents

[0003]

Non-Patent Document 1

Non-Patent Document 2

[0004] Conventional technology has the following problems: <1> The person in charge of monitoring the tunnel face has a heavy workload, including monitoring the behavior of tunnel face collapses and restricting access within a 45-degree radius from the top of the tunnel face. This constant monitoring puts them under immense mental strain. Furthermore, they must stand and monitor near the tunnel face where dust is swirling and visibility is poor, making it a physically and mentally demanding job. In addition, the need to assign a dedicated person in charge of monitoring the tunnel face day and night, constantly monitoring the tunnel face, results in significant labor costs. <2> Because the process relies on the visual inspection of the tunnel face manager, and requires simultaneous monitoring of both sides, there is a risk of human error, such as overlooking the presence of workers, which could prevent timely warnings and the implementation of safety measures (human error). <3> In situations where heavy machinery for excavation frequently enters and exits the site, it is not possible to mark restricted access areas with colored tape near the tunnel face. In this case, it is difficult for the tunnel face manager and workers to grasp the extent of the restricted access area, which is insufficient for safety management and increases the risk of workers mistakenly entering the restricted area (a management error). <4> Operators of heavy machinery may not be able to instantly assess the danger if they accidentally enter a restricted area (physical error). <5> As the tunnel face advances, the restricted access area extends, and the workers' work area becomes further away from the tunnel face, reducing work efficiency. To avoid this, the restricted access area needs to be updated in conjunction with the advancement of the tunnel face, but each update requires surveying and replacing the color tape, making the work time-consuming.

[0005] The object of the present invention is to provide a tunnel face safety management system that solves the problems of the conventional technology described above. [Means for solving the problem]

[0006] The present invention provides a tunnel face safety management system comprising: an image acquisition unit that acquires a front-facing image of the tunnel face including the tunnel face; an area estimation unit that estimates an access restriction area within a predetermined range from the tunnel face based on the front-facing image of the tunnel face; and an area display unit that displays the front-facing image of the tunnel face and the access restriction area. The area estimation unit comprises a tunnel face estimation unit and a distance calculation unit. The tunnel face estimation unit estimates the tunnel face from the front-facing image based on image recognition using machine learning, and the distance calculation unit applies the design cross-section of the tunnel face to the front-facing image of the tunnel face to calculate the distance from the tunnel face to the tunnel entrance, and estimates the area within a predetermined distance from the tunnel face toward the tunnel entrance as an access restriction area.

[0007] The present invention provides a tunnel face safety management system comprising: an image acquisition unit that acquires a front-facing image of the tunnel face including the tunnel face; an area estimation unit that estimates an access restriction area within a predetermined range from the tunnel face based on the front-facing image of the tunnel face; and an area display unit that displays the front-facing image of the tunnel face and the access restriction area. The area estimation unit comprises a tunnel face estimation unit, a support structure estimation unit, and a distance calculation unit. The tunnel face estimation unit estimates the tunnel face from the front-facing image based on image recognition by machine learning, the support structure estimation unit estimates the support structure installed on the borehole wall from the front-facing image based on image recognition by machine learning, and the distance calculation unit calculates the distance from the tunnel face to the tunnel entrance by applying the design dimensions of the support structure to the front-facing image of the tunnel face, and estimates the area within a predetermined distance from the tunnel face towards the tunnel entrance as an access restriction area.

[0008] The present invention provides a tunnel face safety management system comprising: an image acquisition unit that acquires a front-facing image of the tunnel face including the tunnel face; an area estimation unit that estimates an access restriction area within a predetermined range from the tunnel face based on the front-facing image of the tunnel face; and an area display unit that displays the front-facing image of the tunnel face and the access restriction area. The area estimation unit comprises a tunnel face estimation unit and a material estimation unit, wherein the tunnel face estimation unit estimates the tunnel face from the front-facing image based on image recognition by machine learning, and the material estimation unit estimates materials placed inside the tunnel from the front-facing image based on image recognition by machine learning, and estimates the area from the tunnel face to the materials as an access restriction area.

[0009] The tunnel face safety management system of the present invention further comprises a hazard determination unit, which estimates workers or heavy machinery from a front-facing image of the tunnel face based on machine learning-based image recognition, and determines that a state in which workers or heavy machinery are present in a restricted access area, or in which workers or heavy machinery are approaching a restricted access area, is a hazardous state.

[0010] The face safety management system of the present invention includes an alarm unit that issues an alarm signal in conjunction with the determination of a dangerous condition by a hazard determination unit, and the alarm signal may include at least one of the following: illumination of a warning light, generation of an alarm sound, vibration of heavy machinery, and vibration of the worker's body.

[0011] The tunnel face safety management system of the present invention may have a warning light that is a red flashlight, and the flashlight may illuminate the front of the tunnel face when a warning signal is issued.

[0012] The face safety management system of the present invention may include an area projection unit that projects an area of ​​restricted access into the tunnel. [Effects of the Invention]

[0013] The face safety management system of the present invention has at least one of the following effects based on the above configuration. <1> Strengthening monitoring of access to restricted areas can improve safety at the tunnel face and reduce the risk of work-related accidents. <2> By simply installing a camera (image acquisition unit) inside the tunnel, the system can automatically estimate restricted access areas and monitor worker entry into those areas. This eliminates the need for a dedicated face management supervisor, freeing workers from arduous tasks. <3> Based on fixed-point observations using cameras, the restricted access area can be automatically updated as the tunnel face advances. This eliminates the need for cumbersome tasks such as surveying and changing colored tape, resulting in high work efficiency. <4>If it includes a danger determination unit, it can automatically detect the entry of workers into the restricted access area without relying on the human judgment of the face management responsible person. Therefore, the judgment is accurate and safety measures can be taken in a timely manner, resulting in high safety. <5>If it includes a reporting unit, not only workers but also operators of large heavy machinery can intuitively recognize the entry into the restricted access area.

Brief Description of the Drawings

[0014] [Figure 1] Configuration diagram of the face safety management system [Figure 2] Explanation diagram of the face safety management system [Figure 3] Explanation diagram of the distance calculation by the distance calculation unit [Figure 4] Explanation diagram of the reporting by the reporting unit [Figure 5] Explanation diagram of Example 2 [Figure 6] Explanation diagram of Example 4

[0015] Hereinafter, the face safety management system of the present invention will be described in detail with reference to the drawings.

Example 1

[0016] <1>The face safety management system The face safety management system 1 is a system for managing the safety of face front work in mountain tunnel construction. The face safety management system 1 includes at least an image acquisition unit 10, an area estimation unit 20, and an area display unit 30. In this example, it further includes a danger determination unit 40 and a reporting unit 50, and each component is connected so as to be communicable with each other (Fig. 1). Here, the connection may be a wired connection or a wireless connection. The face safety management system 1 has one feature in its configuration in which the restricted access area B is estimated by applying image recognition by machine learning to the face front image A acquired by the image acquisition unit 10. The face safety management system 1 can be configured, for example, with a camera, an alarm device, a server, a display, and an antenna. In detail, a frame is installed near the top of the tunnel at a predetermined distance from the tunnel face A1 toward the tunnel entrance. The camera of the image acquisition unit 10 and the alarm device of the alarm signaling unit 50 are placed on the frame. A server containing the area estimation unit 20 and the danger determination unit 40 is installed at a further distance toward the tunnel entrance from the frame, and the display of the area display unit 30 is installed in a remote monitoring room outside the tunnel (Figure 2). However, the configuration of the tunnel face safety management system 1 is not limited to the above. For example, the area estimation unit 20, the area display unit 30, and the hazard determination unit 40 may be integrated as a personal computer (PC) or tablet terminal. Also, the area display unit 30 may be located inside the tunnel instead of in the remote monitoring room.

[0017] <1.1> Restricted Access Area Restricted Access Area B is an area where access by workers is controlled. In this example, restricted access area B is defined as the "area requiring special consideration" in the Ministry of Health, Labour and Welfare's "Guidelines for Preventing Skin Collapse Disasters at the Face of Mountain Tunnel Construction," specifically the area within a 45-degree angle from the top of face A1 toward the tunnel entrance. The specific extent of restricted access area B varies depending on the design cross-section of the tunnel face A1. However, in the case of the full-section construction method, if the height of the center of the tunnel face A1 is 10m, then the restricted access area B will be approximately 10m from the tunnel face A1 towards the tunnel entrance. However, the restricted access area B is not limited to the above and can be set as an appropriate range depending on the purpose. For example, it may be the area within a predetermined distance to the safe side from the 45-degree line from the top of the tunnel face A1.

[0018] <2> Image acquisition unit The image acquisition unit 10 is a component that acquires an image A of the front of the tunnel face, including the tunnel face A1. In this example, a 4K camera with waterproof and dustproof functions is used as the image acquisition unit 10. The 4K camera has a high resolution of 3840 x 2160 pixels and can capture the front face image A from a distance with high precision. However, the image acquisition unit 10 is not limited to a 4K camera; for example, a night vision camera or an infrared camera may be used in low-light underground environments. The image acquisition unit 10 is positioned with its front facing the tunnel face A1. As excavation of the tunnel face A1 progresses, the installation position of the image acquisition unit 10 moves further away from the tunnel face A1, but as long as it does not interfere with image analysis, there is no need to move the image acquisition unit 10. It is desirable to attach an openable and closable shield to the front of the image acquisition unit 10 to protect it from flying debris during blasting.

[0019] <3> Area Estimation Unit The area estimation unit 20 is a component that estimates the restricted access area B. The area estimation unit 20 comprises at least a face estimation unit 21, a distance calculation unit 22, and a storage unit 23. The area estimation unit 20 acquires the tunnel face front image A from the image acquisition unit 10 and stores it in the storage unit 23. It applies the tunnel face front image A to the tunnel face estimation unit 21 to estimate the tunnel face A1. It applies the tunnel face A1 to the distance calculation unit 22 to calculate the distance within the tunnel face front image A, and estimates the area within a predetermined distance from the tunnel face A1 as the restricted access area B.

[0020] <3.1> Storage section The memory unit 23 is a memory element that stores various types of data. The memory unit 23 stores data of the tunnel face front image A acquired from the image acquisition unit 10, as well as trained models related to the estimation of the tunnel face A1 and worker E, and the design cross-section of the tunnel face A1.

[0021] <3.2> Estimated section of the tunnel face The face estimation unit 21 is a component that estimates the face surface A1 in the face front image A. The tunnel face estimation unit 21 estimates the tunnel face A1 by applying a trained model in the memory unit 23 to the tunnel face front image A in the memory unit 23. In detail, for example, the tunnel face A1 can be estimated by the following procedure. The face estimation unit 21 performs preprocessing on the face front image A in the storage unit 23, such as noise reduction, scaling, image resizing, missing value imputation, and feature extraction. The face estimation unit 21 inputs the pre-processed face front image A to a trained model for face extraction, and the trained model extracts feature quantities such as image edges, texture, color, and shape from the face front image A, and identifies the region of the face A1 within the face front image A based on these feature quantities. The face estimation unit 21 detects the boundary lines of the face surface A1 as edges, displays them as contours in the face front image A, verifies and corrects the boundary lines as necessary, and then stores them in the storage unit 23.

[0022] <3.2.1> Pre-trained models A trained model for extracting tunnel face includes an algorithm and parameters for extracting tunnel face A1 from tunnel face front image A. Such a trained model can be obtained, for example, by following the procedure below. A large number of images of the tunnel face are prepared, and ground truth data is added to include features such as edges, textures, and color differences of the tunnel face A1 within the image. This data is then accumulated to construct a training dataset. Next, a training model for cutting face A1 is constructed by training the training dataset based on a known machine learning algorithm. Machine learning algorithms can utilize a variety of known techniques, including convolutional neural networks (CNNs), a type of deep learning, semantic segmentation, object detection, decision trees, support vector machines (SVMs), and clustering.

[0023] <3.3>Distance calculation section The distance calculation unit 22 is a component that calculates the distance from the tunnel face A1 to the tunnel entrance. The distance calculation unit 22 calculates the distance from the tunnel face A1 to the tunnel entrance by applying the design cross section of the tunnel face A1 to the tunnel face A1 in the tunnel face front image A. The design cross-section of tunnel face A1 is data that includes the cross-sectional shape and dimensions (height, width, radius, etc.) of tunnel face A1 at the applicable construction site. In this example, the calculation is performed using the maximum width of the design cross-section of tunnel face A1 as the basis, following the procedure below. The distance calculation unit 22 measures the number of pixels that make up the maximum width occupied by the tunnel face A1 on the tunnel face front image A (Figure 3). The distance calculation unit 22 calculates the actual dimension per pixel from the number of pixels on the measured face front image A and the design dimensions of the design cross-section of the face A1, and uses this pixel scale to calculate the distance within the image. For example, if the face A1 with a maximum width of 10m occupies 2000 pixels on the face front image A, the actual dimension per pixel will be 5mm. Furthermore, if parallax due to the installation position of the image acquisition unit 10 or geometric distortion of the image due to lens distortion exists, more accurate distance measurement can be achieved by performing calibration in advance, acquiring distortion parameters, and reflecting them. However, the objective of the present invention is not precise surveying, but rather to set the restricted access area B to a safe side at least relative to the tunnel face A1, so the need for correction is not high.

[0024] <3.4> Estimation of restricted access areas The area estimation unit 20 identifies a point at a predetermined distance from the tunnel face A1 toward the tunnel entrance based on the pixel scale calculated by the distance calculation unit 22, and estimates the area from the tunnel face A1 to that point as the restricted access area B.

[0025] <4> Area display section The area display unit 30 is a component that displays the restricted access area B as an image. The area display unit 30 displays the tunnel face front image A and the restricted access area B on the display. Specifically, for example, the restricted access area B is displayed in different colors within the tunnel face front image A. By installing the area display unit 30 in a remote monitoring room outside the tunnel, the construction status near the tunnel face can be checked in real time from the remote monitoring room, and unsafe actions by worker E, such as entering the restricted access area B, can be monitored. Furthermore, the area display unit 30 may also display the determination of a dangerous state F by the dangerous determination unit 40 (described later), or a warning signal from the alarm unit 50, etc.

[0026] <5> Risk Assessment Department The hazard determination unit 40 is a component that determines the hazardous state F. In detail, the hazardous state F is determined as follows, for example. The risk determination unit 40 applies a trained model for worker extraction, also stored in the memory unit 23, to the face front image A in the memory unit 23 to estimate the worker E in the face front image A. The trained model for worker extraction can be obtained by the same known procedure as the trained model for face extraction described above. The hazard determination unit 40 refers to the front face image A and determines that the presence of worker E within the restricted access area B constitutes a hazardous state F. However, hazardous state F is not limited to the presence of worker E within the restricted access area B; for example, the state in which worker E approaches within a predetermined distance from the restricted access area B may also be determined as a hazardous state F. Furthermore, the determination of hazardous state F by the hazard determination unit 40 may be based not only on the worker E, but also on the heavy machinery. In this case, the hazard determination unit 40 applies a trained model for heavy machinery extraction using machine learning to determine that hazardous state F is the state in which heavy machinery is present within the restricted entry area B, or the state in which heavy machinery is approaching within a predetermined distance from the restricted entry area B. The determination of a dangerous condition F can be used not only for issuing a warning signal G by the alarm unit 50 described later, but also for notifying a remote monitoring room and controlling heavy machinery (for example, stopping the engine of a backhoe).

[0027] <6> Alarm Department The alarm unit 50 is a component that emits the warning signal G. The alarm unit 50 issues a warning signal G when the danger determination unit 40 determines that a dangerous state F exists. Warning signal G is, for example, the activation of a warning light using a high-intensity flashlight. The alarm unit 50 activates its warning light in conjunction with the determination of a dangerous condition F, causing a red warning color to flash in the area including the front of the tunnel face (Figure 4). This allows workers E to instantly recognize the dangerous condition F even in the noisy front of the tunnel face, prompting them to take evasive action such as retreating from the restricted access area B. However, the warning signal G is not limited to the illumination of a warning light; for example, it may include the sounding of a loud alarm siren via a speaker, the vibration of a vibration device attached to the heavy machinery to attract the operator's attention, the vibration of a vibration device attached to the worker's body, or a combination of these. [Example 2]

[0028] [An embodiment equipped with an area projection unit] In this example, the face safety management system 1 includes an area projection unit 60 that projects an access restriction area B into the tunnel. The area projection unit 60 can, for example, employ a laser projector installed near the top of the hole wall. The area projection unit 60 irradiates a laser based on the area estimation unit 20's estimation of the restricted access area B, and displays the restricted access area B on the borehole wall or ground (Figure 5). The indication of restricted area B may be done by illuminating the area within restricted area B with a different color than the area outside of restricted area B, or by illuminating the boundary line with restricted area B in a linear fashion. In this example, worker E can perform their work while visually observing restricted access area B, thus further improving work safety. [Example 3]

[0029] [Example of using shoring to estimate restricted access areas] In this example, the area estimation unit 20 comprises a face estimation unit 21, a support structure estimation unit 24, and a distance calculation unit 22, and estimates the restricted access area B using the support structure D. Shoring D is a material used to support the borehole wall and prevent collapse after blasting excavation of the tunnel face A1, and includes structural steel erected in the borehole wall and bearing plates for rock bolts driven into the borehole wall. Since the support structure D has predetermined design dimensions such as the width of the structural steel, the dimensions of the bearing plates, the distance between structural steel, and the distance between bearing plates, in this example, the design dimensions of the support structure D are used to calculate the distance from the tunnel face A1 to the tunnel entrance and estimate the restricted access area B. In detail, for example, the restricted access area B is estimated using the following procedure. The face estimation unit 21 estimates the face surface A1 in the face front image A using the same procedure as in Example 1. The support structure estimation unit 24 applies a trained model for support structure extraction, also stored in the memory unit 23, to the face front image A in the memory unit 23 to estimate the support structures D in the face front image A. The trained model for support structure extraction can be obtained by the same known procedure as the trained model for face surface extraction described above. The distance calculation unit 22 applies the design dimensions of the support structure D to the tunnel face front image A, calculates the actual dimension per pixel from the number of pixels on the tunnel face front image A and the design dimensions of the support structure D, and uses this pixel scale to calculate the distance within the image. The area estimation unit 20 identifies a point at a predetermined distance from the tunnel face A1 toward the tunnel entrance based on the pixel scale calculated by the distance calculation unit 22, and estimates the area from the tunnel face A1 to that point as the restricted access area B. [Example 4]

[0030] [Example of using materials to estimate restricted access areas] In this example, the area estimation unit 20 comprises a face estimation unit 21 and a material estimation unit 25, and estimates the restricted access area B using material C. Specifically, the restricted access area B is estimated using the following procedure, for example. Worker E places material C at the boundary line on the tunnel entrance side of the area that Worker E deems unsafe to enter. Material C consists of components mainly used for on-site safety management, such as traffic cones, warning lights, barricades, and signs. The face estimation unit 21 estimates the face surface A1 in the face front image A using the same procedure as in Example 1. The material estimation unit 25 applies a trained model for material extraction, also stored in the storage unit 23, to the face front image A in the storage unit 23 to estimate the material C in the face front image A. The trained model for material extraction can be obtained by the same known procedure as the trained model for face extraction described above. The area estimation unit 20 estimates the area from the tunnel face A1 to the material C within the tunnel face front image A as the restricted access area B (Figure 6). Here, if two materials C are placed near both side walls inside the tunnel, the line connecting the two materials C can be estimated as the boundary line of the restricted access area B. If only one material C is placed inside the tunnel, the line passing through material C and crossing the tunnel can be estimated as the boundary line of the restricted access area B. This example is useful, for example, when observations during construction determine that the tunnel face A1 is unstable, and it is desired to set an access restriction area B to a safer position beyond a 45-degree angle from the top of the tunnel face A1 toward the tunnel entrance. By simply placing material C at the boundary line where access is to be restricted, access restriction area B can be easily set, and safety at the front of the tunnel can be managed using the area display unit 30, the hazard determination unit 40, and the alarm unit 50, etc. [Explanation of symbols]

[0031] 1. Face Safety Management System 10 Image acquisition unit 20 Area Estimation Section 21 Estimated face section 22 Distance calculation section 23 Memory section 24 Shoring Estimation Department 25. Materials Estimation Department 30 Area display section 40. Hazard Assessment Department 50 Reporting Department 60 Area Projection Section A Front view of the tunnel face A1 Face B Restricted Access Area C Materials D Shoring E Worker F Dangerous condition G Warning Signal

Claims

1. A tunnel face safety management system for managing safety at the tunnel face during tunnel construction in mountainous areas, An image acquisition unit that acquires an image of the front of the tunnel face, including the tunnel face itself, An area estimation unit that estimates an access restriction area within a predetermined range from the face surface based on the aforementioned face front image, The system includes an area display unit that displays the aforementioned face front image and the aforementioned restricted access area, The area estimation unit comprises a face estimation unit and a distance calculation unit, The face estimation unit estimates the face surface from the face front image based on machine learning-based image recognition, The distance calculation unit applies the design cross-section of the tunnel face to the front image of the tunnel face to calculate the distance from the tunnel face to the tunnel entrance, and estimates the area within a predetermined distance from the tunnel face towards the tunnel entrance as the restricted access area. Work face safety management system.

2. A tunnel face safety management system for managing safety at the tunnel face during tunnel construction in mountainous areas, An image acquisition unit that acquires an image of the front of the tunnel face, including the tunnel face itself, An area estimation unit that estimates an access restriction area within a predetermined range from the face surface based on the aforementioned face front image, The system includes an area display unit that displays the aforementioned face front image and the aforementioned restricted access area, The area estimation unit comprises a face estimation unit, a support structure estimation unit, and a distance calculation unit. The face estimation unit estimates the face surface from the face front image based on machine learning-based image recognition, The support structure estimation unit estimates the support structures installed on the borehole wall from the face front image based on machine learning-based image recognition, The distance calculation unit applies the design dimensions of the support structure to the face front image to calculate the distance from the face to the tunnel entrance, and estimates the area within a predetermined distance from the face towards the tunnel entrance as the restricted access area. Work face safety management system.

3. A tunnel face safety management system for managing safety at the tunnel face during tunnel construction in mountainous areas, An image acquisition unit that acquires an image of the front of the tunnel face, including the tunnel face itself, An area estimation unit that estimates an access restriction area within a predetermined range from the face surface based on the aforementioned face front image, The system includes an area display unit that displays the aforementioned face front image and the aforementioned restricted access area, The aforementioned area estimation unit comprises a face estimation unit and a material estimation unit, The face estimation unit estimates the face surface from the face front image based on machine learning-based image recognition, The material estimation unit estimates the materials placed inside the tunnel based on machine learning-based image recognition from the tunnel face front image, and estimates the area from the tunnel face to the materials as the restricted access area. Work face safety management system.

4. It is further equipped with a hazard assessment unit, The hazard determination unit estimates workers or heavy machinery from the image of the front of the tunnel face based on machine learning-based image recognition, and determines that a state in which workers or heavy machinery are present in the restricted access area, or in which workers or heavy machinery are approaching the restricted access area, constitutes a hazardous state. The face safety management system according to any one of claims 1 to 3.

5. The system includes an alarm unit that issues a warning signal in conjunction with the determination of the dangerous state by the aforementioned danger determination unit, The warning signal is characterized by including at least one of the following: illumination of a warning light, generation of an alarm sound, vibration of heavy machinery, and vibration of the worker's body. The face safety management system according to claim 4.

6. The warning light is a red flashlight, and the flashlight illuminates the front of the tunnel face when the warning signal is activated. The face safety management system according to claim 5.

7. The system is characterized by having an area projection unit that projects the aforementioned restricted access area into the tunnel. The face safety management system according to any one of claims 1 to 3.

Images