Flaw detection device for waterproof sheet

Abstract

To provide a flaw detection device for waterproof sheet capable of preventing increase of construction cost of a tunnel and efficiently carrying out flaw confirmation work and repair work at a flaw position.SOLUTION: A flaw detection device 1 for waterproof sheet is provided with: a plurality of cameras 16 imaging respective imaging area IR of each of waterproof sheets in charge so as to cover a waterproof sheet developed area formed with a plurality of the waterproof sheets 3 installed along an inner periphery of a tunnel; a developed area image generation section 111 generating developed area images exhibiting all of the water proof developed areas through composing each of camera images of the waterproof sheets; a flaw position image extraction section 112 extracting a flaw position from the camera images or the developed area images using a learned model having learned a group of images having mode information of flaws of the waterproof sheets; and a flaw position image generation section 113 generating a flaw position image with a mark of the flaw position DA extracted in the developed area image to output it on an image display section.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates to a waterproof sheet flaw detector that detects damaged areas in a waterproof sheet laid along the inner periphery of a tunnel. [Background technology]

[0002] In general tunnel construction, as shown in Fig. 8, support 201 and shotcrete 202 are constructed near the tip of the tunnel in the direction of tunnel excavation E, then rock bolts 203 are cast, and then a waterproof sheet 205 is constructed using a waterproof sheet construction dolly 204, and then a rebar assembly dolly (not shown) is used to install lining rebars 206. After the work of installing the lining rebars 206 is completed, a mobile form dolly 207 is placed to install formwork 208 and pour lining concrete 209.

[0003] When constructing the waterproof sheet 205 along the inner circumference of the upper half of the tunnel UT, for example, a waterproof sheet 205 having a waterproof layer and a cushioning material such as a nonwoven fabric laminated thereon is used, and the waterproof sheet 205 is wound in a roll with a length equal to the circumference of the laying and a width of about 2 m. Then, the waterproof sheet 205 is placed on the arch-shaped sheet placement platform 2041 of the waterproof sheet construction cart 204, and the waterproof sheet 205 is spread out along the inner surface of the sprayed concrete 202, and the cushioning material portion is fixed to the sprayed concrete 202 by nailing or the like, and the waterproof layer portion is welded and joined to the front end of the adjacent waterproof sheet 205 that has already been laid.

[0004] Thus, the waterproof sheet 205 installed in the tunnel may be damaged during the tunnel construction. If the damage to the waterproof sheet 205 is not repaired, it will lead to future water leakage, so it is important to recognize and repair the damaged part of the waterproof sheet 205. However, it is not easy to confirm the presence or absence of damage and the damaged part of the waterproof sheet 205 installed in the tunnel. For this reason, Patent Document 1 proposes an inspection method in which a waterproof sheet in which a thermoplastic resin and a luminescent layer are provided between a water permeable layer such as a nonwoven fabric and a waterproof layer is used, and electromagnetic waves are irradiated to the waterproof layer side of the waterproof sheet installed with the lights in the tunnel turned off, causing the luminescent material in the luminescent layer to luminesce through the damaged part of the waterproof layer, thereby discovering the damaged part. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 4182361 Summary of the Invention [Problem to be solved by the invention]

[0006] However, the method of Patent Document 1 requires the use of a special waterproof sheet, which increases the construction cost of the tunnel. Furthermore, when performing damage confirmation work, it is necessary to move a waterproof sheet construction cart having an arch-shaped sheet placement platform from the site where the waterproof sheet is laid to perform the damage confirmation work, and to irradiate the electromagnetic waves with the lights in the tunnel turned off, which makes the damage confirmation work very time-consuming. Furthermore, when repairing the damaged area after identifying the damaged area, it is necessary to move the waterproof sheet construction cart back to the site where the waterproof sheet is laid to perform the repair, making it difficult to efficiently perform the repair work on the damaged area.

[0007] The present invention has been proposed in consideration of the above-mentioned problems, and aims to provide a waterproof sheet flaw detection device that can prevent an increase in tunnel construction costs and enable efficient damage confirmation work and repair work of damaged areas. [Means for solving the problem]

[0008] The waterproof sheet flaw detection device of the present invention is characterized in that it comprises a plurality of cameras that photograph the respective waterproof sheet photographing areas so as to cover the waterproof sheet unfolded area consisting of a plurality of waterproof sheets laid along the inner circumference of a tunnel, an unfolded area image generation unit that combines the camera images from each of the cameras to generate an unfolded area image that shows the entire waterproof sheet unfolded area, a damage area extraction unit that extracts damaged areas from the camera images or the unfolded area image using a trained model that has been trained using a group of images having information on the state of damage to the waterproof sheets as training data, and a damage area image generation unit that generates a damage area image in which the extracted damaged areas are marked in the unfolded area image and outputs the damage area image to an image display unit. According to this, since there is no need to use a special waterproof sheet and the damage check of the waterproof sheet can be performed without being limited by the type of waterproof sheet, it is possible to prevent an increase in the construction cost related to the damage check of the waterproof sheet, and therefore to prevent an increase in the construction cost of the tunnel. In addition, for example, the damage check of the waterproof sheet and the repair work of the damaged part can be performed while the waterproof sheet construction cart or the like is placed at the site where the waterproof sheet is laid, so that the damage check of the waterproof sheet and the repair work of the damaged part can be performed efficiently. In addition, by generating and outputting a damaged part image in which the extracted damaged part is marked on the development area image of the waterproof sheet, the position of the damaged part to be repaired in the development area of ​​the waterproof sheet can be quickly grasped, and the damaged part can be repaired more quickly.

[0009] The waterproof sheet flaw detection device of the present invention is characterized in that the damage extraction unit uses the trained model, which has been trained using a group of images having information on the appearance of simple wrinkles, sagging, and overlapping areas of the waterproof sheet as training data, to extract damaged areas and poorly welded areas between the waterproof sheets from the camera-captured images or the unfolded area images, and the damage area image generation unit generates a damage area image in which the extracted damaged areas and poorly welded areas are respectively marked in the unfolded area image and outputs it to the image display unit. This makes it possible to accurately distinguish between simple wrinkles or sagging in the waterproof sheet, or overlapping of waterproof sheets, and damage to the waterproof sheet or poor welding between waterproof sheets, and allows the work of checking for damage to the waterproof sheet and poor welding, as well as the work of repairing damaged and poorly welded areas, to be carried out more efficiently and reliably.

[0010] The waterproof sheet flaw detection device of the present invention is characterized in that it is equipped with a learning processing unit that corrects the trained model using condition information of a non-damaged area that is not actual damage that is input when a part of the extracted damaged area or the extracted damaged area is not actual damage, and the camera-captured image or the expanded area image including the non-damaged area as training data. This makes it possible to continuously improve the accuracy of the trained model and continuously increase the accuracy of waterproof sheet damage inspection work.

[0011] The waterproof sheet flaw detection device of the present invention is characterized in that a guide member formed in an arch shape that follows the inner circumference of the tunnel is provided at the rear of the waterproof sheet construction cart, and each of the cameras is installed at intervals on the guide member so that they can photograph the waterproof sheet photography area located forward in the direction of tunnel excavation. This allows damaged areas of the waterproof sheet at the site where the waterproof sheet has been laid to be discovered while the waterproof sheet construction cart is positioned at the site and before the waterproof sheet construction cart begins to move forward, so that repair work for the damaged areas can be carried out quickly and efficiently on the scaffolding of the waterproof sheet construction cart.

[0012] The waterproof sheet flaw detection device of the present invention is characterized in that each of the cameras photographs its own waterproof sheet photography area so as to cover the waterproof sheet deployment area with the covering rebar assembled on the inside. With this, even if damage occurs to the waterproof sheet during the assembly work of the lining rebars, the location of the damage can be identified and repair work to the damaged area can be carried out quickly and efficiently, for example, on the scaffolding of a rebar assembly cart.

[0013] The waterproof sheet flaw detector of the present invention is characterized in that the generated image of the damaged area is stored in a construction management record storage unit. This makes it possible to store the generated images of damaged areas and use them in tunnel construction management. Effect of the Invention

[0014] According to the waterproof sheet flaw detector of the present invention, it is possible to prevent an increase in tunnel construction costs and to efficiently carry out damage confirmation work and repair work on damaged areas. [Brief description of the drawings]

[0015] [Figure 1] 1 is a block diagram showing a waterproof sheet flaw detector according to an embodiment of the present invention. [Diagram 2] FIG. 2 is an explanatory diagram of a waterproof sheet construction cart equipped with a camera in the waterproof sheet flaw detection device of the embodiment, viewed from the tunnel crossing direction. [Diagram 3] FIG. 2 is an explanatory diagram of a waterproof sheet construction cart equipped with a camera in the waterproof sheet flaw detection device of the embodiment, viewed from the longitudinal direction of a tunnel. [Figure 4] FIG. 2 is a conceptual diagram illustrating the waterproof sheet imaging areas of each of a plurality of cameras in the waterproof sheet flaw detection device of the embodiment. [Diagram 5] 4 is a flowchart showing the processing steps of the waterproof sheet flaw detector of the embodiment. [Figure 6] 13 is a diagram showing an example of a damaged area image in which the damaged area is marked and output by the waterproof sheet flaw detector of the embodiment. FIG. [Figure 7] FIG. 11 is an explanatory diagram showing a rebar assembly cart on which a camera is installed in a modified example of the waterproof sheet flaw detector of the embodiment, viewed from the tunnel longitudinal direction. [Figure 8] FIG. 1 is an explanatory perspective view illustrating a typical tunnel construction procedure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] [Waterproof sheet flaw detector according to an embodiment] As shown in Fig. 1, a waterproof sheet flaw detection device 1 according to an embodiment of the present invention is composed of a flaw detection processing device 10 and a plurality of cameras 16 that capture images of the respective waterproof sheet capture areas IR (see Fig. 4). The flaw detection processing device 10 can be composed of an operation terminal such as a server, a personal computer, or a smartphone, and includes a control unit 11 such as an MPU or CPU, a storage unit 12 composed of an HDD, SSD, ROM, RAM, etc., an input unit 13 such as a mouse, keyboard, touch panel, etc., an image display unit 14 such as a display, and a communication interface 15.

[0017] The memory unit 12 stores the required control programs that work in cooperation with the control unit 11 to cause the flaw detection processing device 10 to execute predetermined processing. In particular, the memory unit 12 in this embodiment has a flaw detection processing program storage unit 121 that stores a flaw detection processing program that causes the control unit 11 to execute flaw detection processing, and a learning processing program storage unit 122 that stores a learning processing program that causes the control unit 11 to execute learning processing and correction processing of a trained model.

[0018] Furthermore, the memory unit 12 has a trained model storage unit 123 that stores a trained model in a state where a required number of image groups having information on the state of damage to the waterproof sheet have been trained as teacher data T, and a construction management record storage unit 124 that stores the generated damaged area images and the developed area images after the damaged areas have been repaired as construction management records in association with information on the date and time and place of image generation. In this embodiment, the trained model in a pre-trained state stored in the trained model storage unit 123 is in a state where a required number of image groups having information on the state of damage to the waterproof sheet, an image group having information on the state of simple wrinkles in the waterproof sheet, an image group having information on the state of slack in the waterproof sheet, an image group having information on the state of poor welding of the waterproof sheet due to breakage, derailment, or mismatch of the welding line with the adjacent waterproof sheet, an image group having information on the state of overlapping areas of the waterproof sheets, or an image group having appropriate multiple of these state information as teacher data T.

[0019] The control unit 11 executes processing as an unfolded area image generating unit 111, a damaged area extracting unit 112, and a damaged area image generating unit 113 in accordance with the flaw detection processing program. The unfolded area image generating unit 111 synthesizes the camera images of the waterproof sheet assigned shooting area IR by each camera 16 to generate an unfolded area image showing the entire waterproof sheet unfolded area.

[0020] The damaged area extraction unit 112 uses the learned model stored in the learned model storage unit 123 to extract damaged areas DA and poorly welded areas BW from the camera-captured image of the waterproof sheet assigned shooting area IR, or from the unfolded area image generated by combining the camera-captured images of the waterproof sheet assigned shooting area IR. The damaged area extraction unit 112 of this embodiment uses a trained model that has been pre-trained using as training data T both a group of images having simple wrinkles in the waterproof sheet and modal information indicating they are simple wrinkles, slackness in the waterproof sheet and modal information indicating they are slackness, and overlapping areas of the waterproof sheets and modal information indicating the overlapping areas, i.e., modal information of the proper welding line with an adjacent waterproof sheet, and a group of images having damaged areas DA of the waterproof sheet and modal information indicating the damaged areas DA, and modal information indicating poor welding areas BW of poor welding of the waterproof sheet due to breakage, derailment, or misalignment of the welding line with an adjacent waterproof sheet, and accurately distinguishes between simple wrinkles in the waterproof sheet, slackness in the waterproof sheet, and overlapping of waterproof sheets, and extracts the damaged areas DA and poor welding areas BW from the camera-captured images or unfolded area images of the waterproof sheet's assigned shooting area IR.

[0021] The damaged area image generating unit 113 generates a damaged area image by applying marking M to the damaged area DA and the poor welding area BW extracted in the expanded area image, and outputs the generated damaged area image to the image display unit 14 as shown in Fig. 6. The damaged area image generating unit 113 also stores the generated damaged area image in the construction management record storage unit 124 as a construction management record, in association with information on the date and time and place where the image was generated.

[0022] The control unit 11 executes further learning and correction processes of the learned model as the learning processing unit 114 according to the learning processing program. In the embodiment, when a part of the damaged area DA extracted in the damaged area image or the extracted damaged area DA is not actual damage, for example, by visual confirmation, the learning processing unit 114 updates and corrects the learned model stored in the learned model storage unit 123 using the non-damaged area that is not actually damaged and the state information indicating that it is a non-damaged area or indicating the actual state of the area judged to be non-damaged, which are input from the input unit 13, and the camera-captured image of the waterproof sheet assigned shooting area IR including the non-damaged area or the deployment area image including the non-damaged area as teacher data, and stores the learned model in the learned model storage unit 123.

[0023] In addition, in this embodiment, when, for example, by visually checking, a part of the poorly welded area BW extracted in the damaged area image or the extracted poorly welded area BW is not an actual poorly welded area, the learning processing unit 114 updates and modifies the learned model stored in the learned model storage unit 123 using, as teacher data, non-poorly welded areas that are not actual poorly welded areas input from the input unit 13, condition information indicating that the area is a non-poorly welded area or indicating the actual condition of the area determined to be a non-poorly welded area, and a camera-captured image of the waterproof sheet assigned shooting area IR including the non-poorly welded area or an unfolded area image including the non-poorly welded area, and stores the learned model in the learned model storage unit 123.

[0024] The multiple cameras 16 installed in the waterproof sheet flaw detection device 1 of this embodiment are positioned so as to be able to capture the respective waterproof sheet capture areas IR so as to cover the waterproof sheet deployment area consisting of multiple waterproof sheets 3 laid along the inner circumference of the tunnel, as shown in Figures 2 to 4, and are installed so that the camera images captured by each camera 16 are imported into the flaw detection processing device 10 via the communication interface 15 (see Figure 1).

[0025] In the example of Fig. 2 and Fig. 3, in which a plurality of cameras 16 are arranged so as to be able to photograph the respective photographing areas IR of the waterproof sheet, a waterproof sheet construction cart 5 capable of moving back and forth along a temporary rail 4 is provided in the tunnel space T. The waterproof sheet construction cart 5 has a sheet placement platform 52 supported by a scaffolding 51, and is used so that the waterproof sheet 3 is placed on the sheet placement platform 52 when laying the waterproof sheet 3 along the inner circumference of the sprayed concrete 21 of the tunnel. The waterproof sheet 3 in this example is composed of a waterproof sheet part 31 composed of a waterproof layer and a buffer material 32 laminated and fixed to the waterproof sheet part 31, and is laid by fixing the buffer material 32 to the sprayed concrete 21 by nailing or the like. In addition, the ends of the waterproof sheet parts 31 adjacent to each other in the front and rear of the tunnel excavation direction E are arranged to overlap and fixed to each other.

[0026] A guide member 53 formed in an arch shape that follows the inner circumference of the tunnel is provided at the rear of the waterproof sheet installation cart 5, and each camera 16 is installed at intervals on the guide member 53 so that it can photograph the waterproof sheet assigned shooting area IR located ahead in the tunnel excavation direction E.

[0027] 7, in which multiple cameras 16 are arranged so that they can each capture an image of their assigned capture area IR of the waterproof sheet, a rebar assembly cart 6 that can move back and forth along a temporary rail 4 is provided in the tunnel space T, and the lining rebars 7 are assembled on the inner periphery of the waterproof sheet 3 using the rebar assembly cart 6. After the lining rebars 7 are erected, a mobile form cart is placed to install the form 8 and pour the lining concrete.

[0028] At the rear of the rebar assembly cart 6, a guide member 61 formed in an arch shape that follows the inner circumference of the tunnel is provided, and each camera 16 is installed at intervals on the guide member 61 so that it can photograph the waterproof sheet's assigned photography area IR of the waterproof sheet 3 with the lining rebar 7 assembled on the inside, which is located forward in the tunnel excavation direction E.

[0029] In the processing by the waterproof sheet flaw detection device 1 of this embodiment, as shown in Fig. 5, for example, in response to an image capture request input from the input unit 13, the cameras 16 capture images of the respective waterproof sheet capture areas IR so as to cover the waterproof sheet deployment area consisting of the multiple waterproof sheets 3 laid along the inner circumference of the tunnel (S1). At this time, it is preferable to set the adjacent waterproof sheet capture areas IR so that their edges overlap each other, so that a deployment area image showing the entire waterproof sheet deployment area described later can be generated. The flaw detection processing device 10 imports the camera images corresponding to the respective waterproof sheet capture areas IR captured by the cameras 16 via the communication interface 15 (S2).

[0030] The unfolded area image generating unit 111 of the flaw detection processing device 10 generates an unfolded area image showing the entire waterproof sheet unfolded area by synthesizing the camera images of the respective waterproof sheet assigned shooting areas IR by the respective cameras 16 (S3). Any appropriate image synthesis technology can be used as the image synthesis technology for synthesizing the camera images of the respective waterproof sheet assigned shooting areas IR to generate an unfolded area image showing the entire waterproof sheet unfolded area.

[0031] The damaged area extraction unit 112 of the flaw detection processing device 10 extracts the damaged area DA and the defective welding area BW from the camera image of the waterproof sheet assigned shooting area IR or the developed area image generated by combining the camera images of the waterproof sheet assigned shooting area IR using the learned model stored in the learned model storage unit 123 (S4). Then, the damaged area image generation unit 113 generates a damaged area image in which markings M are applied to the damaged area DA and the defective welding area BW extracted in the developed area image, and outputs the damaged area image to the image display unit 14 (S5).

[0032] In addition, the damaged area image is visually inspected, and if the extracted damaged area BA, part of the poor welding area BW, or the extracted damaged area BA and poor welding area BW are not actual damage or poor welding, the trained model is updated and corrected as described above, and the updated trained model is stored in the trained model storage unit 123 (S6).

[0033] According to the waterproof sheet flaw detector 1 of the present embodiment, it is not necessary to use a special waterproof sheet, and the damage and welding defects of the waterproof sheet 3 can be confirmed without being limited by the type of waterproof sheet, so that the increase in construction costs related to the damage and welding defects of the waterproof sheet 3 can be prevented, and the increase in the construction cost of the tunnel can be prevented. In addition, for example, the waterproof sheet construction cart 5 or the like can be placed at the site where the waterproof sheet 3 is laid, and the damage and welding defects of the waterproof sheet 3 can be confirmed and repaired efficiently. In addition, by generating and outputting a damaged area image in which markings M are applied to the damaged area DA and the welding defects BW extracted in the development area image of the waterproof sheet 3, the positions of the damaged area DA and the welding defects BW to be repaired in the waterproof sheet development area can be quickly grasped, and the repair can be performed more quickly.

[0034] In addition, the damage extraction unit 112 extracts the damaged areas DA and the poorly welded areas BW using a trained model that has been trained using a group of images having information on simple wrinkles, sagging, and overlapping areas of the waterproof sheet as training data T, thereby making it possible to accurately distinguish between simple wrinkles and sagging of the waterproof sheet, overlapping of waterproof sheets, and damage to the waterproof sheet and poor welding between waterproof sheets, and thereby making it possible to more efficiently and reliably check for damage and poor welding of the waterproof sheet 3, and to repair the damaged areas and poorly welded areas.

[0035] In addition, when a part of the extracted damaged area DA or poor welding area DW or the extracted damaged area DA or poor welding area BW is not actual damage or poor welding, the input information on the condition of the non-damaged area or non-poor welding area that is not actual damage or poor welding, and the camera-captured image or unfolded area image including the area are used as training data to correct the trained model, thereby continuously improving the accuracy of the trained model and continuously improving the accuracy of the damage confirmation work for the waterproof sheet 3 and the accuracy of the poor welding confirmation work.

[0036] In addition, when multiple cameras 16 are installed on the arch-shaped guide member 53 provided at the rear of the waterproof sheet construction cart 5, damaged areas DA and poorly welded areas BW of the waterproof sheet 3 at the site where the waterproof sheet 3 has been laid can be discovered while the waterproof sheet construction cart 5 is positioned at the site before the waterproof sheet construction cart 5 moves forward, so that repair work on the damaged areas DA and poorly welded areas BW can be carried out quickly and efficiently on the scaffolding 51 of the waterproof sheet construction cart 5.

[0037] Furthermore, if each camera 16 photographs its own waterproof sheet photographing area IR so as to cover the waterproof sheet deployment area with the lining rebars 7 assembled on the inside, even if damage occurs to the waterproof sheet 3 during the assembly work of the lining rebars 7, the location of the damage can be identified and repair work to the damaged area can be carried out quickly and efficiently, for example, on the scaffolding of the rebar assembly cart 6.

[0038] In addition, by storing the generated images of damaged areas, etc. in the construction management record storage unit 124, the images of damaged areas, etc. can be used to help manage the tunnel construction.

[0039] [Scope of the invention disclosed herein] The inventions disclosed in this specification include, in addition to the inventions and embodiments listed as inventions, those specified by changing partial contents of these to other contents disclosed in this specification, those specified by adding other contents disclosed in this specification to these contents, or those specified by deleting partial contents of these to the extent that partial effects can be obtained, and specifying them as a higher-level concept. The inventions disclosed in this specification also include the following contents and modifications.

[0040] For example, in the above embodiment, a normal machine learning learning process program and a trained model by normal machine learning were used for the learning process program and the trained model, but it is also suitable to use a deep learning learning process program and a trained model by deep learning, and the trained model may be updated and corrected by deep learning.

[0041] In addition, in the above embodiment, an example has been described in which the camera 16 takes photographs to cover the entire waterproof sheet deployment area when the lining rebars 7 are not assembled on the inside, and an example has been described in which the camera 16 takes photographs to cover the entire waterproof sheet deployment area when the lining rebars 7 are assembled on the inside, but it is also possible to use a configuration that combines both examples. [Industrial Applicability]

[0042] The present invention can be used to detect damaged areas in a waterproof sheet laid along the inner periphery of a tunnel. [Explanation of symbols]

[0043] 1...Waterproof sheet flaw detection device 10...Flaw detection processing device 11...Control unit 111...Expanded area image generation unit 112...Damage location extraction unit 113...Damage location image generation unit 114...Learning processing unit 12...Memory unit 121...Flaw detection processing program storage unit 122...Learning processing program storage unit 123...Learning model storage unit 124...Construction management record storage unit 13...Input unit 14...Image display unit 15...Communication I / F 16...Camera 20...Natural ground 21...Sprayed concrete 22...Ceiling 3...Waterproof sheet 31...Waterproof sheet part 32...Buffer material 4...Temporary rail 5...Waterproof sheet construction cart 51...Scaffolding 52...Sheet placement stand 53...Guide member 6...Reinforcing bar assembly cart 61...Guide member 7...Lining reinforcing bar 8...Formwork 201...Shoring 202...Sprayed concrete 203...Rock bolt 204...Waterproof sheet installation cart 2041...Sheet placement platform 205...Waterproof sheet 206...Lining rebar 207...Mobile formwork cart 208...Formwork 209...Lining concrete T...Teaching data T...Tunnel space E...Tunnel excavation direction IR...Waterproof sheet photography area OR...Overlap area M...Marking DA...Damaged area BW...Welding failure area UT...Upper half of tunnel

Claims

1. A plurality of cameras are installed along the inner circumference of the tunnel to cover the area where the waterproof sheets are laid out, each of which captures its own area. A deployment area image generating unit that generates a deployment area image showing the entire waterproof sheet deployment area by synthesizing the camera images of each of the cameras; A damaged area extraction unit that extracts damaged areas from the camera image or the deployment area image using a trained model trained using a group of images having information on the state of damage to the waterproof sheet as training data; A waterproof sheet flaw detection device characterized by comprising: a damage area image generation unit that generates a damage area image by marking the damaged area extracted in the expanded area image and outputs it to an image display unit.

2. The damaged area extraction unit extracts damaged areas and poor welding areas between the waterproof sheets from the camera-captured image or the deployed area image using the trained model trained with a group of images having state information of simple wrinkles, slackness, and overlapping areas of the waterproof sheets as training data, The waterproof sheet flaw detection device described in claim 1, characterized in that the damaged area image generation unit generates a damaged area image in which the damaged area and the poor welding area extracted in the expanded area image are each marked, and outputs the image to an image display unit.

3. The waterproof sheet flaw detection device of claim 1 or 2, characterized in that it is equipped with a learning processing unit that corrects the learned model using information on the state of a non-damaged area that is not actual damage and the camera-captured image or the expanded area image including the non-damaged area as training data when a part of the extracted damaged area or the extracted damaged area is not actual damage.

4. A guide member formed in an arch shape that follows the inner circumference of the tunnel is provided at the rear of the waterproof sheet installation cart, A waterproof sheet flaw detection device as described in claim 1 or 2, characterized in that each of the cameras is installed at intervals on the guide member so as to be able to photograph the waterproof sheet photography area located forward in the direction of tunnel excavation.

5. A waterproof sheet flaw detection device as described in claim 1 or 2, characterized in that each camera photographs its own waterproof sheet photography area so as to cover the waterproof sheet deployment area with the covering rebar assembled on the inside.

6. 3. The waterproof sheet flaw detector according to claim 1, wherein the generated image of the damaged area is stored in a construction management record storage unit.

Images