KR20210144390A - Method for generating tunnel construction documents automatically - Google Patents

Abstract

The present invention provides various construction documents such as construction record management record book, inspection request form, inspection result notification form, work report and construction waste consignment processing work request form using an automated system in which the construction record book, the inspection management section, the work daily report management section, and the waste treatment management section are automated. Provided are a method capable of automatically generating and a program recorded in a storage medium to perform such a method. In particular, automatic calculation of material quantity is possible through image analysis, and DFM technology is applied to create more effective construction documents.

Description

Method for generating tunnel construction documents automatically}

The present invention relates to a method for automatically generating various documents related to the construction of a tunnel, and a program recorded in a storage medium to perform the method.

Tunnel construction is made by repeating one round of processes such as drilling, blasting, burping, sealing and installing shotcrete, support, shotcrete, and rock bolts. On average, information blasting is performed twice a day. Much attention is required as in other construction sites, but especially careful attention is required considering blasting.

More than 60 related construction documents are required for tunnel construction. In the actual field, the site manager takes pictures for each of the above processes, requests for inspection or records the actual measured data, and then returns to the office after the day's work is completed and performs various document tasks with a computer to perform paper work. The construction document is prepared by printing the document.

In this case, various problems arise.

Reliance on records by site managers may result in inaccurate documentation. For example, if the site manager incorrectly confirms the number of rock bolts, it is documented as it is, and work based on this is made in the next round of drilling and blasting. In other words, there is a problem in that the document accuracy varies depending on the skill level of the site manager.

If the site manager concentrates on the document work, the construction site management supervision, which is the original task of the site manager, may be relaxed for various reasons, such as concentrating on acquiring information, measuring, and taking pictures. This may cause a major accident.

Site managers work overtime even after a day's work is completed and have to spend a lot of time writing paper documents. Paper documents prepared in this way may be lost or lost. If this pattern continues, problems may arise in the construction site management due to the accumulation of fatigue of the site manager.

(Patent Document 1) KR 10-1675034

(Patent Document 2) KR 10-2074578

(Patent Document 3) JP 2019-023392A

(Patent Document 4) JP 2018-071165A

(Patent Document 5) KR 10-2009-0020359A

(Patent Document 6) JP 4094770B2

(Patent Document 7) EP 1176393B1

(Patent Document 8) JP 2019-065648A

The present invention has been devised to solve the above problems.

It is intended to provide a method to automate the creation of various construction documents.

To provide a method that enables automatic creation of documents with high accuracy and efficiency, such as automatically counting the number of specific materials, automatically synchronizing the inspection request form and the inspection result notification do.

It is intended to provide a method for generating high-accuracy automatic construction documents regardless of the skill level of the site manager. It should be possible to generate high-accuracy documents within the system even for low-skilled site managers.

We want to apply the DFM processing method. Through this, the image included in the construction document will quickly provide a variety of information beyond a simple photographic image.

An embodiment of the present invention for solving the above problems, (a1) the construction record unit 200 to check the general information; - The general information of step (a1) includes the project group, construction name, tool, location, direction and date and time, (b1) the construction type of any one of the construction types stored in the construction record book 200 and the construction type database 810 inputted and confirmed; (c1) an image is input to the image management unit 620, and the construction record unit 200 calls and confirms the input image; (d1) the construction record unit 200 confirming the materials, specifications, and the quantity; and (e1) the construction record book 200 uses the general information, work type, image, and material identified in the steps (a1), (b1), (c1) and (d1), and the standard and quantity thereof. and generating a construction record management record by further using the created moving board board, wherein the (d1) step is, (d11) When automatic input and AI identification are selected, the construction record book 200 is Confirms the material and standard set in advance as corresponding to the input work type, and the AI identification module 622 of the image management unit 620 identifies the material of the confirmed standard through the input image analysis, and the quantity to confirm; (d12) When automatic input and RFID identification are selected, the construction record unit 200 confirms the materials and specifications set in advance to correspond to the input type of work, and the RFID identification module 623 in the image management unit 620 receives confirming the quantity of the material of the identified standard based on one RFID signal; and (d13) when manual input is selected, materials, specifications, and quantities thereof are input to and confirmed in the construction record book 200 .

In addition, in step (c1), an image is input to the image management unit 620, the input image is DFM-processed by a digital face mapping (DFM) module 621, and the construction management unit 200 performs the Including the step of confirming the DFM-processed image, wherein the DFM module 621 DFM processing the input image, the scene image information as an input layer, the analysis information to be DFM-processed on the scene image information As the steps of automatically outputting analysis information for the inputted scene image information through an automatic mak scene analysis system using learning as an output layer, - Here, the analysis information includes information on carcinoma, information on cancer quality, and joints Including information on - (A1) the learning scene image information input unit loads the learning scene image information prepared in advance for learning, and the learning analysis information input unit loads the learning analysis information corresponding to the closing scene image information, mapping the loaded scene image information for learning with the analysis information for learning, and setting learning information; (A2) generating, by the learning information dividing unit, a plurality of divided learning information for each piece of learning information by cropping the learning information as a preset unit; (A3) converting the format conversion unit into a format for learning by loading the plurality of divided learning information; (A4) performing, by the learning performing unit, learning multiple times for each single scene image information using the format-converted plurality of divided learning information; (A5) constructing an automatic scene analysis system by performing the steps (A1) to (A4) on a plurality of scenes for learning image information; and (B) when the scene image information is input to the input unit, the inference unit infers the analysis information using the automatic analysis system for the scene scene, the output unit outputs it, and DFM processing the input image. do.

In addition, (a2) the step of the detection management unit 300 to check the general information; - The general information of step (a2) includes the name of the construction, the tool, the date and time of the inspection request and the detection number, (b2) in the inspection management unit 300, the construction type database 810 of any one of the types of construction inputted and confirmed; (c2) confirming that the special items and related drawings are input to the detection management unit 300; (d2) selecting any one or more detailed work types from among the detailed work types stored in the work type database 810 in the detection management unit 300 and confirming it as a detection checklist; (e2) an image is input to the image management unit 620, and the detection management unit 300 calls and confirms the input image; and (f2) the detection management unit 300 checks the general information, work type, special items, related drawings and detection checklist confirmed in steps (a2), (b2), (c2), (d2) and (e2). It is preferable to further include the step of generating a detection request using the.

In addition, (a3) the step of the detection management unit 300 to call and confirm the general information; - The general information of step (a3) includes the name of the construction, the tool, the date of receipt, and the recipient, (b3) the detection and management unit 300, the detector information, the detection result and action requirements are entered and confirmed; And (c3) the detection management unit 300 using the general information, the detector information, the detection result and the action requirements confirmed in the steps (a3) and (b3) to generate a detection result notice further comprising it is preferable

In addition, (a4) the step of checking the work daily report management unit 400 by calling the general information; - The general information of step (a4) includes the name of the construction, the tool, the date and time of the current day and the last day, and the weather condition, (b4) the step of confirming that the today's plan and the last day's plan are input to the work daily report management unit 400; (c4) confirming that the tunnel excavation status, process status, input number, planned number of people, input equipment and scheduled equipment are input to the work daily report management unit 400; (d4) the operation daily report management unit 400 calculating and confirming the accumulated tunnel excavation status, personnel input, and equipment input; and (e4) the work daily report management unit 400 confirms the general information in the steps (a4), (b4), (c4) and (d4), today's plan, myongil plan, tunnel excavation status, process status, input number , further comprising the step of generating a work report using the expected number of people, input equipment, scheduled equipment, tunnel excavation status total, personnel input total and equipment input total, wherein (d4) step is, (d41) If there is, the work daily report management unit 400 checks the accumulated tunnel excavation status, the input number of people and the input equipment on the previous day's work daily report, and the tunnel excavation status, input personnel and input equipment input in step (c4) above. calculating the cumulative tunnel excavation status, personnel input, and equipment input by summing each; and (d42) if there is no work report for the previous day, calculating the tunnel excavation status, input number of people and input equipment input in step (c4) as the cumulative tunnel excavation status, the cumulative personnel input, and the cumulative equipment input, respectively. desirable.

In addition, (a5) the waste treatment management unit 500 to check the general information; - The general information of step (a5) includes the name of the project, the tool and the consignment number, (b5) the type of consigned waste, the consignment treatment request date and time, the place of occurrence and the preparation status are input to the waste treatment management unit 500; (c5) an image is input to the image management unit 620, and the waste treatment management unit 500 calls and confirms the input image; (d5) The waste treatment management unit 500 constructs using the general information, the type of consigned waste, the date and time of the consignment treatment request, the place of occurrence, the preparation status and the image confirmed in the steps (a5), (b5) and (c5) It is preferable to further include the step of generating a waste consignment treatment work request form.

Another embodiment of the present invention for solving the above problems provides a program, recorded in a storage medium, so that the above-described methods are performed.

The present invention has the following advantages.

First, by automatically generating construction documents that require a lot of writing time at the actual construction site, it is possible to reduce the heavy work time of the site manager and focus on the site management, which is the original task of the site manager, so that safe site management is possible.

Second, the accuracy of various construction documents increases. For example, the material quantity inaccuracy problem is solved by automatically checking the material quantity by AI or RFID technology. Since the detailed work types described in the inspection request form are applied to the inspection result notification, the problem that some of the checklists are not detected is solved. When the construction documentation is accurate, the construction and construction stability also increases.

Third, through the provision of DFM-processed images, various information can be provided in all kinds of construction documents that require images. For example, in all documents, carcinoma, sarcoid zone, etc. are provided, and RQD can also be provided. By sharing these documents with field workers, it is possible to prevent unexpected accidents in advance and provide inspiration for various ideas.

Fourth, information that is advantageous to be input by the site manager induces input, and all other information is automatically processed, providing a high-convenience UI.

1 schematically shows a system for carrying out a method according to the invention;
2 is a flowchart illustrating a first embodiment of the method according to the present invention.
3 is a flowchart illustrating DFM processing according to the present invention.
4 is a flowchart illustrating a second embodiment of the method according to the present invention.
5 is a flowchart illustrating a third embodiment of the method according to the present invention.
6 is a flowchart illustrating a fourth embodiment of the method according to the present invention;
7 is a flowchart illustrating a fifth embodiment of the method according to the present invention.
8A and 8B show an example of a construction record management record book automatically generated by the first embodiment according to the present invention.
9A to 9H are diagrams for explaining DFM processing according to the present invention.
10 shows an example of a detection request form automatically generated by the second embodiment according to the present invention.
11 shows an example of a detection result notification automatically generated by the third embodiment according to the present invention.
12 shows an example of a job report automatically generated by the fourth embodiment according to the present invention.
13 shows an example of a construction waste consignment processing work request form automatically generated by the fifth embodiment according to the present invention.
14 to 20 show a UI of a program recorded in a storage medium to perform the method according to the present invention.

Hereinafter, "tunnel construction document" means various documents to be generated in the tunnel construction process. In the present invention, it is described as including a construction record management record book, inspection request form, inspection result notification form, work daily report, and construction waste consignment processing work request form and described as an example, but it is not limited to these documents, and the method according to the present invention It should be understood as a concept that includes all the various documents that can be created using

Hereinafter, "learning" refers to when an information processing device such as a computer applies input information to an input layer, it undergoes predetermined training to automatically infer the corresponding output information to build an algorithm that outputs it to the output layer. It refers to the data processing technique performed and is a kind of artificial intelligence. Machine learning may be included. Learning consists of training using a plurality of training sets in which an input layer and an output layer are determined. Learning is performed by using a predetermined algorithm as the subject of the information processing device, and when an algorithm is built by performing a plurality of learning executions, inference can be automatically performed in a predetermined manner according to the information input by the user. do. The algorithm used here may vary. Below, as a learning algorithm to be used, inception is exemplarily proposed.

Hereinafter, "digital face mapping (DFM)" means mapping digitized information on a photographed image. By DFM according to the present invention, information on carcinoma, cancer quality, joint, etc. is digitally mapped on a scene image taken by a user with a camera or the like. In other words, when a user inputs an image of a scene using the method according to the present invention, analysis information is automatically inferred and superimposed on the image and output, and this analysis result is a DFM result. As shown in the upper part of FIG. 9G, joints may be shown as line segments, or as shown in the lower part of FIG. 9G, areas of carcinoma (eg, granite, etc.) and rock quality (eg, severe weathering, etc.) may be separated and shown.

Description of the system

The system for performing the method according to the present invention includes a construction record unit 200 , a detection management unit 300 , a work daily report management unit 400 , a waste treatment management unit 500 , a general information management unit 610 , an image analysis unit 620 . ), a document output unit 710 , a payment management unit 720 , a work type database 810 , and a weather condition database 820 .

The construction record book 200 performs a function related to the automatic generation of the construction record management record book, the detection management unit 300 performs a function related to the automatic generation of the detection request form and the detection result notification, and the work daily report management unit 400 performs the work It performs a function related to the automatic generation of the daily report, and the waste treatment management unit 500 performs a function related to the automatic generation of the construction waste consignment processing work request form. Specific details will be described in Examples 1 to 5 below.

The general information management unit 610 manages general information included in each document. General information refers to information other than specific construction-related information as information written on a predetermined basis for each entire construction or section construction (tool). For example, it may include a project group (constructor), construction name (tunnel name), tool (section), location, direction, date and time, weather status, creator, recipient, and the like. The general information may be different for each type of document. General information described in the example of the present invention is shown in the following table. However, it should be noted that this is only an example, and if the form or format of the construction document to be described is changed, addition or deletion is possible.

Construction record management record book Project group, construction name, tool, location, direction, date Inspection request form Construction name, tool, inspection request date and time, inspection number Inspection result report Construction name, tool, date of receipt, recipient job report Construction name, tool, date and time, weather conditions Construction waste consignment processing work request form Construction name, tool, consignment number

It is preferable that the user input in advance, except for the date and time and weather conditions that require real-time detection among such general information.

In another embodiment, information may be processed together with an individual unique code for the work type and managed together in the work type database 810 . In other words, if a specific work type is input or selected together with the date and time and number of times, or an individual unique code for the work type is input or selected, general information related to it (eg, project group, construction name, tool, location, direction, etc.) is called together. can get This will be described later in relation to the work type database 810 .

The image analysis unit 620 performs various analyzes by using the camera 20 connected by wire or wirelessly to directly acquire an image by the system, or by using an image uploaded by a user.

The image analysis unit 620 includes a DFM module 621 , an AI identification module 622 , and an RFID identification module 623 .

The DFM identification module 621 performs digital mapping of information on carcinoma, cancer quality, joint, etc. on the input image. A specific implementation method will be described below.

The AI identification module 622 automatically calculates the number of objects in the image by applying AI to find a specific object to the input image. For example, when a final image is input, it is possible to automatically calculate how many rock bolts 10 of a specific standard are identified in the image. Any AI used is free.

The RFID identification module 623 automatically calculates the number of corresponding objects in the corresponding image (precisely at the time of photographing) by identifying the RFID signal obtained when photographing the input image. To this end, the RFID tag 14 is already attached to the object, and the RFID identification module 623 is connected to the RFID reader 15 that reads it by wire or wirelessly. For example, referring to FIG. 1 , at the time of acquiring an image after a plurality of rock bolts 10 consisting of a rock bolt body 11 , acupressure plate 12 and nut 13 are embedded, the RFID reader 15 is the corresponding block. may be read and the read RFID signals may be synchronized with the image. In this case, when an image is input to the RFID identification module 623, the synchronized RFID signal is confirmed. When the RFID identification module 623 analyzes the identified RFID signal, it is possible to identify the signal transmitted from which object, so that it is checked which material is provided in the corresponding image and how many.

The document output unit 710 performs this when it is necessary to output the document created by the method according to the present invention on paper.

The payment management unit 720 performs a function of paying the document created by the method according to the present invention or transmitting the paid document, and the payment in this case may be electronic approval.

The work type database 810 stores the type of work and its detailed work type. By using the name of the construction type in the database, the name of the construction type is unified in any document and input is free even when a new construction method is developed. The detailed work type is information necessary for the selection of the inspection checklist to be recorded at the time of the inspection request.

The work type database 810 may be nested within the system in which the method according to the present invention is performed, but may be associated with a separate external system.

As described above, the type of work stored in the work type database 810 (eg, the seven types of work described in the prior art) may be stored in a format that a user can directly select in a drop-down manner, but in another embodiment, it is performed daily and by time. Each may be further divided, and an individual unique code may be assigned and stored together. For example, in the case of the second rock bolt construction on May 19, a code such as "051902A07" may be assigned. Therefore, when the user accesses the construction type database 810 and selects or inputs the construction type, date, and number of times, or selects or inputs an individual unique code, all information related to the construction type is called up, and as will be described later, construction documents Each is recorded in the necessary parts for writing. Here, the information related to the type of work includes the general information described above. For example, even if only a code such as "051902A07" is entered or selected, the project group (constructor), construction name (tunnel name), tool (section), location, direction, date and time, weather conditions, Information is automatically retrieved and recorded in the required location.

Due to the information storage and processing method of the construction type database 810, if a variable occurs in any one construction type and affects the entire construction type, the entire data stored in the database can be modified, so that other users who do not know the variable occurrence By selecting or entering the relevant work type or code, the modified data output is possible. In addition, it may be updated in real time, such as the process status to be described later. In other embodiments, notifications may be automatically sent to relevant users when data modifications are made.

The meteorological status database 820 is connected to a database such as the Meteorological Administration by wire or wirelessly to retrieve and record and store the weather of a desired date, which is used for various construction documents as general information.

Example 1 - Construction record management record book

The first embodiment will be described with reference to the flowchart of FIG. 2 . 8A and 8B show an actual construction record management record book prepared according to the present invention, which is referred to together.

First, the date and time may be checked based on the time point. The general information necessary for the construction record management record book is checked by the construction record book 200 by calling the general information from the general information management unit 610 (S210). The general information management unit 610 already includes basic information, for example, project group, project name, tool, location, and direction, and includes the recalled, project group, project name, tool, location, direction and date and time, but addition or deletion is not possible. Freedom is as described.

Next, any one of the types of work stored in the type database 810 is input to the construction record book 200 and confirmed (S220). As described above, the method in which the name of the work type itself is input or selected, or the method in which an individual unique code for the work type is input or selected can be confirmed. In this case, steps S210 and S220 of calling general information will proceed simultaneously.

Now, an image is input to the image management unit 620 . It may be taken directly by the camera 20 of the system, or it may be uploaded by a user. The construction recorder 200 calls and confirms the input image (S230).

Next, the quantity of materials of a specific standard corresponding to the type of work should be checked. Automatic input or manual input may be selected by the user, and in case of automatic input, AI identification or RFID identification may be selected (S240).

When automatic input and AI identification are selected, the construction record unit 200 checks the materials and specifications set in advance to correspond to the input type of work (S250), and the AI identification module 622 of the image management unit 620 displays the input image Through the analysis, the material of the previously confirmed standard is identified and the quantity is checked (S260, S261).

When automatic input and RFID identification are selected, the construction record unit 200 confirms the materials and specifications set in advance as corresponding to the entered work type (S250), and the RFID identification module 623 in the image management unit 620 takes an image Based on the received and synchronized RFID signal, the quantity of the material of the confirmed standard is checked by analyzing it (S260, S262).

When the manual input is selected, the material, the standard, and the quantity thereof are directly input to the construction record book 200 and confirmed (S263).

Next, the construction record book 200 creates a moving mountain board using the confirmed general information, work type, image, and material and its specifications and quantity (S270), and further uses the created moving mountain board to automatically record the construction record management record. to generate (S280).

An example of an automatically generated construction record management record book is shown in FIGS. 8A and 8B . Figure 8a is a case where the AI identification module 622 is used, and 41 114mm steel pipes are automatically identified (red circle) and the number is described, and Figure 8b is a case using the RFID identification module 623, 50X20X30 support is 1 The dog installation was confirmed and described through the RFID signal. It is confirmed that the Dongsan board is added to the lower left corner of each image.

Meanwhile, in another embodiment of the present invention, a DFM-processed image rather than a general image may be included in the construction document, which will be described.

DFM module

DFM is based on AI, and we first describe how to prepare learning information to do this.

Prepare the input layer, the scene image information for learning, and the output layer, which is the analysis information, for learning. It is preferable that the analysis information for learning is information verified by an expert. The scene image information for learning and the analysis information for learning are mapped to each other.

If there are N pieces of training scene image information and corresponding analysis information for learning, some (A pieces) of them are used for learning and the rest (N-A pieces) are used for verification (N>A).

In the actual implementation system, 900 out of 1000 learning scenes image information was used for learning, and the remaining 100 were used for verification.

The learning scene image information input unit loads the learning scene image information prepared in advance for learning, and the learning analysis information input unit loads the learning analysis information corresponding to the final scene image information. The prepared scene image information for learning and the analysis information for learning are referred to as learning information (S311) (FIG. 9A).

It is important not to learn it as it is, but to learn it by dividing it. This is to learn a lot of learning information quickly by reducing the amount of computation. To this end, the learning information dividing unit generates a plurality of divided learning information for each single scene image information by cropping the prepared learning information as a preset unit (S312).

In the actual implementation system, the pixel unit was divided into 30X30 and stored (FIG. 9b). As learning information, a plurality of scene image information and corresponding analysis information are mapped and acquired, and the divided information is stored in the learning information database as individual files.

Next, the learning method will be described.

The learning module performs learning using the divided learning information prepared in the learning information generation module.

First, the format conversion unit loads a plurality of divided learning information and converts it into a learning format (S313).

In the actual implementation embodiment, the learning information was converted to .tfrecord so that learning can proceed using TF-Slim of tensorflow, which includes the width, height, depth, and label of the image identified in the scene image information. (label), raw image (image raw), etc. are included.

Next, the learning executor uses the format-converted plurality of divided learning information, but performs learning a plurality of times for each piece of divided learning information while converting a learning rate and a training step for a preset number of times ( S314). In this process, the learning parameter setting unit may set an optimal parameter so that the amount of computation is minimized (S315). The optimal parameters will be described below.

Through this process, learning is performed several times for each piece of divided learning information, and learning is performed several times for each piece of learning information in which a plurality of pieces of divided learning information are gathered.

In the actual implementation embodiment, the learning method used by the learning execution unit is Inception. In a deep learning network, the deeper the layer, the better the performance, but the amount of computation also increases exponentially. Inception reduces the number of connections between nodes in order to reduce the amount of computation, and makes the matrix operation a dense operation. The inception algorithm is adopted in consideration of the specificity of the image analysis of the scene as in the present invention (the size of the image itself is wide, it is important to understand the boundary between carcinoma and rock quality, and segmentation such as joints is important).

Inception is a method of reducing the number of parameters in 3X3 and 5X5 convolutional layers by reducing the image channel by 1X1 convolution operation. As a result, the number of parameters is reduced while constructing a network that is deeper than CNN (Convolutional Neural Networks), and the amount of computation is reduced.

Specifically, in general inception, (1) 1X1 convolution for the previous layer, (2) 1X1 convolution and 3X3 convolution, (3) 1X1 convolution and 5X5 convolution, and (4) 3X3 max pooling and 1X1 convolution, 4 types of operations are performed, and the learning is carried out by combining them into one.

In the present invention, while using inception, the need to perform 7X7 convolution operation in the first layer (stem layer) is replaced with 3x3 convolution operation three times, the optimizer is set to RMSProp, and the last layer (fully connected) layer) was applied with batch normalization (BN). In addition to this, by applying factorization to further reduce the amount of computation and time by improving the kernel, perform 2 3X3 convolution operations instead of 1 5X5 convolution operation, and 2 3X1 convolution operations instead of 1 3X3 convolution. improved. As a result, it was confirmed that the calculation amount was reduced by about 2.78 times. Figure 9c schematically shows a network used in this way.

Meanwhile, the learning parameter setting unit sets the final parameter according to the result of performing learning while changing the learning rate and the training stage. The parameter set here can be divided into an optimization parameter and a learning parameter.

Examples of the optimizer parameter include weight_decay, Optimizer, opt_epsilon, rmsprop_momentum, rmsprop_decay, and the like.

weight_decay means the weight for optimizing the loss function of the model. Here, it was set to 0.00004 according to the optimization result.

opt_epsilon is the smallest value to prevent the case where the calculated value becomes 0. Here, it was set to 1 according to the optimization result.

rmsprop_momentum is the momentum value of the RMSPropOptimizer. Here, it was set to 0.9 according to the optimization result.

rmsprop_decay is the rate of decrease of the probability applied for each update. Here, it was set to 0.9 according to the optimization result.

As a learning parameter, learning_rate, max_number_of_steps, learning_rate_decay_type, num_epochs_per_decay, replicas_to_aggregate, moving_average_decay, etc. may be mentioned.

learning_rate is the initial learning rate. Here, it was set to 0.0001 according to the optimization result.

max_number_of_steps is the maximum number of training steps. Here, it was set to 1000 according to the optimization result.

Here, learning_rate and max_number_of_steps are the main parameters. In order to find an appropriate learning speed and learning stage, learning was carried out by changing the learning_rate to 0.0001, 0.001, 0.01, and 0.1, and in the case of max_number_of_steps, learning was carried out while changing from 1,000 to 10,000 (increasing by 1,000).

learning_rate_decay_type is a parameter that specifies how to decrease the learning rate. These include fixed, exponential, and polynomial. Here, it was set to fixed according to the optimization result.

num_epochs_per_decay is the number of epochs after the learning rate is reduced. Here, it was set to 2 according to the optimization result.

replicas_to_aggregate is the number of gradients to collect before updating the parameter. Here, it was set to 1 according to the optimization result.

moving_average_decay is the decay rate used for the moving average. Here, it is set to None according to the optimization result. In this way, learning is performed several times for each divided learning information in one final scene image information for learning, and by performing this learning on a plurality of learning scenes image information, an automatic final scene analysis system is built (S316).

The validation unit checks the established system by validating the validity by using other unused scene image information among the prepared learning information (S317).

As a result of checking in the actual implementation example, it was confirmed that the accuracy was 70% or more, and thus the accuracy was relatively high. It showed a big difference compared to the case where a learning algorithm other than Inception was used.

learning algorithm accuracy Softmax classification 30.6% SVM 47.8% Inception 66.8% Inception (actual application) 71.8%

Now, the user can automatically acquire analysis information such as carcinoma, rock quality, joint, etc. by inputting the image information of the scene using the built-up automatic analysis system for the scene.

First, the makjang scene image information is input to the input unit (S321). The user can directly obtain and input the scene image information by shooting with a camera while using an information processing device equipped with a communication function and a camera, or select an image stored in the memory in the information processing device and input it through the input unit. may be In the field, only an information processing device is sufficient, and no separate equipment is required.

The image correction unit corrects the image to be input in several ways.

The input image taken directly by the user further includes information on boundary objects such as surrounding walls as well as the final scene (see the lower left of FIG. 9D). . That is, an image is input (top of FIG. 9d), a tunnel boundary node may be input on the scene image information input by the image correction unit (bottom left of FIG. 9d), and the input tunnel boundary node is connected and connected By extracting only the area (bottom right of FIG. 9D), the scene image information is corrected (S322). Most of the prior art selects a method of correcting an image rather than a method of selecting a part from the image. The present invention selects an area extraction method by nodes.

Also, as shown in FIG. 9E , the image correction unit may divide the input scene image information into a plurality of regions by using a watershed technique ( S323 ). This reduces the amount of computation when inputting the scene image information input as the input layer, thereby enabling rapid computation and inference in the reasoning unit to be described later. As described in the prior art, speed is very important in analyzing a scene image.

Next, the inference unit infers the analysis information by using a system constructed for each divided region, and infers the digital face-mapped analysis information for the input scene image information by integrating the analysis information (S324). That is, the inference unit inputs the scene image information into the input layer of the built-up automatic analysis system, and when the output layer by artificial intelligence is checked, it is inferred as analysis information.

On the other hand, in another embodiment of the present invention, the joint extraction may be made separately. The joint extraction unit extracts a line segment from the input scene image information using a canny edge detection technique (FIG. 9F (a)). Next, the joint extraction unit may remove noise by performing a predetermined test for each extracted line segment (FIG. 9F(b)) It may also be possible ((c) of FIG. 9F). Now, the remaining line segments will be output as joints (S325).

When all analysis information is arranged through this process, the output unit outputs the analysis information as a result ( FIG. 9G ). The result of this method may be stored in the DFM database, may undergo a user modification process described later, and may be reused as learning information to increase accuracy. It may be output in the form of a report as shown in the lower right of FIG. 9G .

In another embodiment of the present invention, the RQD (rock quality designation; rock quality index) may be calculated and further output. The RQD calculation can be performed by any well-known method, but the important thing is to automatically set the RQD radiation center. That is, the RQD calculation unit includes the steps of setting the center of the RQD radiation line, and calculating and outputting the RQD using the center of the RQD radiation line ( S326 ).

As a method for the RQD calculation unit to set the RQD radiation center, the present invention provides four methods. First, the center of the RQD radiation line can be set as the center of the points constituting the joint among the analysis information (top left of FIG. 9h ). Second, the center of the RQD radiation line may be set as the center of the center point of the line segments constituting the joint among the analysis information (top right of FIG. 9h ). That is, it is set based on the densely populated area. Third, the center of the RQD radiation line may be set as the center of the line segment with the longest length among the line segments constituting the joint among the analysis information (bottom left of FIG. 9h ). Fourth, the center of the RQD radiation line may be set as the center of the horizontal radiation line that most intersects the joint in the analysis information among a plurality of horizontal radiation lines set on the scene image information (bottom right of FIG. 9h ).

When the DFM process is finished, the output information may be partially modified (S327), and the information thus confirmed may be utilized again as learning information in step S311 (S328).

In this way, the DFM module 621 may DFM process the input image, so that the result may be, for example, as shown in FIG. 9G, and the DFM-processed image in this way is the construction shown in FIGS. It can be applied as an image on the record management record book.

Example 2 - Inspection Request Form

A second embodiment will be described with reference to the flowchart of FIG. 4 . Figure 10 shows an actual detection request prepared according to the present invention, together with reference to.

First, the detection management unit 300 checks the general information (S410). As described above, the general information of the inspection request form may include the construction name, tool, inspection request date and time and inspection number, but may be added or modified. The detection number may be automatically generated according to a set rule as a serial number.

Next, any one of the types of work stored in the type database 810 is input to the detection management unit 300 and confirmed (S420). As described above, the method in which the name of the work type itself is input or selected, or the method in which an individual unique code for the work type is input or selected can be confirmed. In this case, steps S410 and S420 of calling general information will proceed simultaneously. In addition, special items and related drawings are input and confirmed (S430).

In the inspection request form, the detailed work type should be selected and used as a detection checklist, and the detailed work type based on the work type database 810 is selected and input to the detection management unit 300 and confirmed as a detection checklist (S440) (FIG. 17) Reference).

Next, an image is input to the image management unit 620, and the detection management unit 300 calls and confirms the input image (S450). The DFM-processed image described above may be utilized.

Now, the detection management unit 300 generates a detection request using the confirmed general information, work type, special note, related drawings and detection checklist (S460).

10 shows an example of a request for detection automatically generated by the present invention.

Example 3 - Inspection result notification

A third embodiment will be described with reference to the flowchart of FIG. 5 . 11 shows an actual detection result notification written in accordance with the present invention, which is referred to together.

The detection management unit 300 calls and confirms the general information (S510). The general information of the inspection result notification may include the name of the construction, the tool, the date of receipt, and the recipient, but may be added or deleted.

Next, the detector information, the detection result, and the action requirements are input to the detection management unit 300 and confirmed (S520).

Now, the detection management unit 300 generates a detection result notification using the confirmed general information, the detector information, the detection result and the action requirements (S530).

11 shows an example of a detection result notification automatically generated by the present invention. It can be confirmed that the detection checklist is written with the code name of the detailed process, and the result for each detailed process will be described as the detection result next to it.

Embodiment 4 - Daily report

A fourth embodiment will be described with reference to the flowchart of FIG. 6 . 12 shows an actual work report prepared according to the present invention, which is referred to together.

The work daily report management unit 400 calls and confirms general information (S610). The general information of the work report may include the name of the work, tools, date and time of the current day and day, and weather conditions, but may be added or deleted.

Next, the today's plan and the next day's plan are input to the work daily report management unit 400 and confirmed (S620).

In addition, the tunnel excavation status, the process status, the input number, the planned number of people, the input equipment and the scheduled equipment are input to the work daily report management unit 400 and confirmed (S630, S640, S650). Here, the number of input personnel and input equipment are the personnel and equipment of the construction performed today, and the expected number of personnel and equipment are the personnel and equipment of the construction planned for the day.

Next, the work daily report management unit 400 calculates and checks the accumulated tunnel excavation status, personnel input, and equipment input, which can be distinguished through whether or not a work daily report exists.

If there is a daily work report for the previous day, the work daily report management unit 400 checks the accumulated tunnel excavation status, the total number of input personnel and the accumulated input equipment on the previous day's work report, and the tunnel excavation status and input number input in steps S630 and S650 here. And by summing the input equipment respectively, the accumulated tunnel excavation status, personnel input, and equipment input total are calculated (S660, S661).

If there is no work report for the previous day, that is, in the case of the first day of construction, the tunnel excavation status, input personnel and input equipment input in steps S630 and S650 are calculated as the cumulative tunnel excavation status, personnel input, and equipment input, respectively (S660) , S662).

Next, the work daily report management unit 400 confirms general information, today's plan, myongil plan, tunnel excavation status, process status, input number, expected number of people, input equipment, planned equipment, tunnel excavation status cumulative, personnel input cumulative and equipment input A work report is generated using the accumulated total (S670).

12 shows an example of a work report automatically generated by the present invention.

5th embodiment - Construction waste consignment treatment work request form

A fifth embodiment will be described with reference to the flowchart of FIG. 7 . 13 shows an actual construction waste consignment processing work request prepared in accordance with the present invention, which is referred to together.

The waste treatment management unit 500 checks the general information (S710). The general information of the construction waste consignment processing work request includes the name of the construction, the tool, and the consignment number. The consignment number may be automatically generated according to a set rule as a serial number.

Next, the type of consignment waste, the date and time of the consignment treatment request, the place of occurrence and the preparation state are input to the waste treatment management unit 500 (S720).

A waste-related image is input to the image management unit 620, and the waste treatment management unit 500 calls and confirms the input image (S730).

Now, the waste treatment management unit 500 generates a construction waste consignment treatment work request form using the general information identified above, the type of consignment waste, the date and time of the consignment treatment request, the place of occurrence, the preparation state, and the image.

13 shows an example of a construction waste consignment processing work request form automatically generated by the present invention. This may include a calculation of the quantity of construction waste storage using images.

Description of the UI

A user interface (UI) of an embodiment of a program recorded in a storage medium to perform the method according to the present invention will be described with reference to FIGS. 14 to 20 .

14 is a main screen of the program. Process rate, input manpower, scheduled manpower, input equipment, scheduled equipment, etc. appear, which indicates information confirmed or calculated by the work daily report management unit 400 . In addition, the weather condition is confirmed in the weather condition database (820). Links and current states of various documents automatically generated according to the present invention are shown at the bottom of the screen.

15 is a screen for generating a construction record management record book by the construction record book 200. An inputable index appears on the left and an image appears on the right.

16 is a screen for generating a detection request by the detection management unit (300). In particular, as shown in FIG. 17, a screen for selecting a detailed construction type from the construction type database 810 may pop up, and when selected here, it is written in the detection request as a detection checklist as a code name, and the written code name is the detection result It will be used as it is in the notification. In particular, it can be seen that the table for inputting the detection result is deactivated when the detection request form is generated in FIG. 17 .

18 is a screen for generating a detection result notification by the detection management unit (300). Information according to the detection request such as the code name on the detection checklist as well as general information is loaded from the detection request form as it is.

FIG. 19 is a screen for generating a work report by the work daily report management unit 400 .

20 is a screen for generating a construction waste consignment processing work request by the waste treatment management unit 500 .

In the above, the present specification has been described with reference to the embodiments shown in the drawings so that those skilled in the art can easily understand and reproduce the present invention, but these are merely exemplary, and those skilled in the art may make various modifications and equivalent other modifications from the embodiments of the present invention. It will be appreciated that embodiments are possible. Therefore, the protection scope of the present invention should be defined by the claims.

10: Rock bolt
11: Rock bolt body
12: acupressure plate
13: nut
14: RFID tag
15: RFID reader
20: camera
200: construction record book
300: detection management unit
400: work daily report management department
500: waste treatment management department
610: General information management department
620: image analysis unit
621: DFM module
622: AI identification module
623: RFID identification module
710: document output unit
720: payment management unit
810: work database
820: weather database

Claims (7)

(a1) the construction record book 200 confirming the general information; - General information of step (a1) includes project group, construction name, tool, location, direction and date and time
(b1) checking the construction record unit 200 by inputting any one of the construction types stored in the construction type database 810;
(c1) an image is input to the image management unit 620, and the construction record unit 200 calls and confirms the input image;
(d1) the construction record unit 200 confirming the materials, specifications, and the quantity; and
(e1) the construction record book 200 uses the general information, work type, image, and material identified in steps (a1), (b1), (c1) and (d1), and its specifications and quantity and creating a construction record management record by further using the created moving board,
The step (d1) is,
(d11) When automatic input and AI identification are selected, the construction record unit 200 confirms the materials and specifications set in advance to correspond to the input type of work, and the AI identification module 622 of the image management unit 620 is , identifying the material of the identified standard through the analysis of the input image and confirming the quantity;
(d12) When automatic input and RFID identification are selected, the construction record unit 200 confirms the materials and specifications set in advance to correspond to the input type of work, and the RFID identification module 623 in the image management unit 620 receives confirming the quantity of the material of the identified standard based on one RFID signal; and
(d13) when manual input is selected, including the step of confirming that the material, specification, and its quantity are input to the construction record book 200 ,
Way.
The method of claim 1,
In step (c1), an image is input to the image management unit 620, the input image is DFM-processed by a digital face mapping (DFM) module 621, and the construction management unit 200 is subjected to the DFM processing Including the step of checking the image,
The step of the DFM module 621 DFM processing the input image,
Through the automatic analysis system for the scene using learning that uses the scene image information as the input layer and the analysis information processed by DFM on the scene image information as the output layer, automatically outputs the analysis information for the input scene image information as steps, wherein the analysis information includes information about the carcinoma, information about the cancer quality, and information about the joint-
(A1) The final scene image information input unit for learning loads the learning scene image information prepared in advance for learning, the analysis information input unit for learning loads the analysis information for learning corresponding to the final scene image information, and the loaded learning scene image information mapping information and analysis information for learning to set learning information;
(A2) generating, by the learning information dividing unit, a plurality of divided learning information for each piece of learning information by cropping the learning information as a preset unit;
(A3) converting the format conversion unit into a format for learning by loading the plurality of divided learning information;
(A4) performing, by the learning performing unit, learning multiple times for each single scene image information using the format-converted plurality of divided learning information;
(A5) constructing an automatic scene analysis system by performing the steps (A1) to (A4) on a plurality of scenes for learning image information; and
(B) when the scene image information is input to the input unit, the inference unit infers the analysis information using the automatic analysis system for the scene scene, and the output unit outputs it, comprising the step of DFM processing the input image,
Way.
The method of claim 1,
(a2) the detection management unit 300 confirming the general information; - General information of step (a2) includes construction name, tool, inspection request date and time, and inspection number
(b2) confirming that any one of the work types stored in the work type database 810 is input to the detection management unit 300;
(c2) confirming that the special items and related drawings are input to the detection management unit 300;
(d2) selecting any one or more detailed work types from among the detailed work types stored in the work type database 810 in the detection management unit 300 and confirming it as a detection checklist;
(e2) an image is input to the image management unit 620, and the detection management unit 300 calls and confirms the input image; and
(f2) The detection and management unit 300 uses the general information, work type, special note, related drawings and detection checklist confirmed in steps (a2), (b2), (c2), (d2) and (e2) Further comprising the step of generating a request for detection by
Way.
4. The method of claim 3,
(a3) the detection and management unit 300 calling the general information and confirming; - The general information of step (a3) includes the name of the project, the tool, the date of receipt, and the recipient
(b3) confirming that the detector information, the detection result and the action requirements are input to the detection management unit 300; and
(c3) the detection management unit 300 using the general information, detector information, detection result and action requirements confirmed in the steps (a3) and (b3) further comprising the step of generating a detection result notification,
Way.
The method of claim 1,
(a4) the work daily report management unit 400 calling and confirming general information; - The general information of step (a4) includes the name of the construction, tools, date and time of the current and last days, and weather conditions
(b4) confirming that the today's plan and the next day's plan are input to the work daily report management unit 400;
(c4) confirming that the tunnel excavation status, process status, input number, planned number of people, input equipment and scheduled equipment are input to the work daily report management unit 400;
(d4) the operation daily report management unit 400 calculating and checking the accumulated tunnel excavation status, personnel input, and equipment input; and
(e4) the work daily report management unit 400 confirms in the steps (a4), (b4), (c4) and (d4) The method further includes the step of generating a work report using the expected number of people, input equipment, scheduled equipment, tunnel excavation status cumulative total, personnel input total and equipment input total,
The step (d4) is,
(d41) If there is a work report for the previous day, the work daily report management unit 400 checks the accumulated tunnel excavation status, the total number of input personnel and the accumulated equipment input in the work daily report, and the tunnel excavation input in step (c4) above calculating the cumulative tunnel excavation status, personnel input, and equipment input by adding up the current status, the number of people input and the input equipment, respectively; and
(d42) if there is no work report for the previous day, calculating the tunnel excavation status, input personnel and input equipment input in step (c4) as the total tunnel excavation status, the cumulative personnel input and the equipment input, respectively,
Way.
The method of claim 1,
(a5) the waste treatment management unit 500 confirming the general information; - General information of step (a5) includes construction name, tool and consignment number
(b5) inputting the type of consigned waste, the date and time of the consignment treatment request, the place of occurrence and the preparation state to the waste treatment management unit 500;
(c5) an image is input to the image management unit 620, and the waste treatment management unit 500 calls and confirms the input image;
(d5) The waste treatment management unit 500 constructs using the general information, the type of consigned waste, the date and time of the consignment treatment request, the place of occurrence, the preparation status and the image confirmed in the steps (a5), (b5) and (c5) Further comprising the step of generating a waste consignment processing work request form,
Way.
A program recorded on a storage medium so that the method according to any one of claims 1 to 6 is performed.

Images