JP6989884B1 - Electronic small blackboard generator and electronic small blackboard generation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an electronic small blackboard generator for automatically generating an electronic small blackboard image from a design drawing. SOLUTION: A structural information input unit for inputting a structural diagram used for construction work, an identifier for analyzing the structural diagram and uniquely identifying a structure included in the structural diagram, and the structure distinguished by the identifier. A cross-sectional view illustrating the structure of an object, a learning unit that generates a trained model that is machine-learned using a layout pattern as an array pattern as training data for element information that is distinguished by the identifier and indicates the specifications of the structure. The layout pattern is a layout pattern that distinguishes the layout pattern for the input of the structural drawing by the trained model and superimposes it on the image that records the construction status of the structure for each structure to be constructed in the construction work. Electronic small blackboard image output unit that outputs based on, and electronic small blackboard generator including. [Selection diagram] Fig. 1

Description

The present invention relates to an electronic small blackboard generator and an electronic small blackboard generation method.

In the construction work, it is necessary to make a record that the construction is completed according to the design drawing created in advance. In particular, for construction sites that cannot be visually recognized from the outside after the completion of construction, it is particularly required to record the construction status of each site as an image. As this recording method, a method of taking a picture of the construction site mainly using a camera is used. At that time, along with the construction site, a small blackboard for construction with information for identifying the construction site should also be reflected.

The small blackboard for construction had to be rewritten to the contents of each construction site recorded as an image, and this rewriting was done manually. In addition, since it is necessary to record accurately, it is necessary to fill in with clarity that can be seen in the photograph. In this respect, since the number of construction sites to be recorded is large depending on the scale of construction work, it is very complicated to accurately rewrite the small blackboard for construction each time. Therefore, in recent years, in order to eliminate the complexity of small blackboards for construction work, the introduction of electronic small blackboards is progressing.

Conventionally, it has been necessary for an operator to identify and extract a necessary part from the characters and figures described in a design drawing. As an example of the prior art, there is known a technique of manually specifying characters and figures described in a design drawing and extracting those specified parts to generate an image of an electronic small blackboard (see Patent Document 1). ).

Japanese Unexamined Patent Publication No. 2019-174882

In particular, in reinforced concrete construction, the image of reinforcing bars is to be recorded in order to record that the arrangement state of the reinforcing bars buried inside the concrete wall at the time of completion (type, arrangement, number of reinforcing bars used, etc.) is as designed. The figure showing is useful.

According to the technique described in Patent Document 1, in order to manually generate an image of an electronic small blackboard in advance, a person specifies and extracts a necessary drawing part from a bar arrangement diagram, so that the work for advance preparation is complicated. There is.

In addition, in order to respond to changes in the working environment in recent years, regulations on the upper limit of working hours are planned in Japan, and there is a demand to improve the work efficiency of personnel involved in construction work, so electronic work will be carried out more and more in the future. It is expected that the use of blackboards will be promoted.

However, in the design drawings of building works, for example, the description styles such as bar arrangement drawings are not completely unified, and even if the information necessary for the structural drawings is determined, the styles may fluctuate. Therefore, in order to automatically identify and extract the figures and characters required to generate an electronic small blackboard from a bar arrangement diagram, it is necessary to deal with fluctuations in expressions (variations in style) in the drawings. be. That is, there is a problem in generating an electronic small blackboard accurately and efficiently in order to improve the efficiency of creating construction records required for construction work by using the conventional technology.

An object of the present invention is to provide an electronic small blackboard generator and an electronic small blackboard generation method for automatically generating an electronic small blackboard image from a design drawing.

In order to solve the above problems, the present invention relates to an electronic small blackboard generator, a structural information input unit for inputting a structural diagram used for construction work, and a structure included in the structural diagram by analyzing the structural diagram. A layout pattern as an arrangement pattern of an identifier for uniquely identifying the above, a cross-sectional view showing the structure of the structure distinguished by the identifier, and element information distinguished by the identifier and showing the specifications of the structure. The layout pattern is distinguished for the input of the structural diagram by the learning unit that generates a trained model that has been machine-learned as training data, and the trained model is used for each structure constructed in the construction work. It is characterized by including an electronic small blackboard image output unit that outputs an electronic small blackboard image to be superimposed on an image for recording the construction status of the structure based on the layout pattern.

According to the present invention, an electronic small blackboard can be automatically generated from a design drawing.

The block diagram of the system used for carrying out the electronic small blackboard generation method which concerns on this invention. The hardware block diagram which shows the embodiment of the electronic small blackboard generator which concerns on this invention. The hardware configuration diagram of the information input device and the image output device used in the embodiment of the electronic small blackboard generator according to the present invention. The figure which shows the usage form of the said image output device. The flowchart which illustrates the process flow of the electronic small blackboard generation method which concerns on this invention. The figure which shows the example of the structural diagram as the input data to the said electronic small blackboard generator. The figure explaining the flow of machine learning for the structural diagram input to the said electronic small blackboard generator. The figure explaining the flow of machine learning for the structural diagram input to the said electronic small blackboard generator. The figure explaining the flow of machine learning for the structural diagram input to the said electronic small blackboard generator. The figure which shows the example of the data table generated by processing the structural diagram input to the said electronic small blackboard generator by a trained model, and the example of the output electronic small blackboard image. The figure which shows another example of the data table generated by processing the structural diagram input to the said electronic small blackboard generator by a trained model. The figure which shows another example of the electronic small blackboard image which is output based on the data table which processed the structural diagram input to the said electronic small blackboard generator by a trained model, and was output. The figure which shows the example of the network structure used in the machine learning of the said electronic small blackboard generator.

[Summary of the embodiment according to the present invention]
Hereinafter, the electronic small blackboard generator and the electronic small blackboard generation method according to the present invention will be described with reference to the drawings. FIG. 1 is a system configuration diagram showing an example of an electronic small blackboard generator using cloud computing.

The electronic small blackboard automatic generation device 10 as an electronic small blackboard generation device is a device that realizes a predetermined function by using the hardware resources of a computer arranged on the Internet and using software technology executed in the hardware resources. Is. The details of the electronic small blackboard automatic generation device 10 will be described later, but among the design drawings for building work (hereinafter, simply referred to as “design drawings”) input from the information input device 20, the structure such as a pillar. It is equipped with a trained model by machine learning that uses the design drawing of an object (hereinafter referred to as "structural drawing") as teacher data. In the following description, an information processing device typified by a so-called personal computer is exemplified as hardware that realizes an information input function to the electronic small blackboard automatic generation device 10. Further, as an information output device (including an image display), a so-called tablet-type information terminal is exemplified. However, the hardware resources responsible for input / output of information to the electronic small blackboard automatic generation device 10 according to the present embodiment are not limited to these. Any form may be used as long as the information processing function described below can be realized.

Then, the electronic small blackboard automatic generation device 10 analyzes the structural diagram in the trained model to automatically and automatically generate an appropriate electronic small blackboard image even if the input structural diagram styles are various. It can be accurately generated and managed in association with construction sites that require recording.

Further, the electronic small blackboard automatic generation device 10 also has a function of appropriately outputting the managed and stored electronic small blackboard image based on the request of the image display device 30 and storing the recorded image acquired by the image display device 30. Be prepared.

[Hardware configuration example of electronic small blackboard automatic generation device 10]
FIG. 2 is a diagram showing a hardware configuration example of an information processing device constituting an electronic small blackboard automatic generation device 10 using cloud computing technology. For example, the electronic small blackboard automatic generation device 10 has the following hardware resources.

As shown in FIG. 2, the electronic small blackboard automatic generation device 10 has an arithmetic unit 101 configured by a CPU (Central Processing Unit). Further, it has a storage device 102 composed of a RAM (Random Access Memory), a ROM (Read Only Memory), an SSD (Solid State Drive), and the like.

Further, the electronic small blackboard automatic generation device 10 has an input device 103 having a function of inputting data received via a communication network to the arithmetic unit 101 and an output having a function of outputting the result processed by the arithmetic unit 101. It has a device 104.

Further, the electronic small blackboard automatic generation device 10 has an interface (hereinafter referred to as “I / F105”) and the like. A communication module 11 is connected to the I / F 105. The communication module 11 has a function of connecting the Internet to the electronic small blackboard automatic generation device 10 so as to be communicable.

In FIG. 2, the electronic small blackboard automatic generation device 10 is represented as a hardware resource housed in one housing, but the present invention is not limited to this. For example, the electronic small blackboard automatic generation device 10 according to the present embodiment may be configured to combine a plurality of information processing devices to realize the same hardware resources as described above.

Further, the electronic small blackboard automatic generation device 10 may include an arithmetic unit 101, a storage device 102, an input device 103, an output device 104, and an auxiliary device. Specifically, the information processing device may have an auxiliary device such as a GPU (Graphics Processing Unit) outside or inside.

[Hardware configuration example of information input device 20]
FIG. 3A is a diagram showing a hardware configuration example of an information input device 20 as an information processing device for inputting a structural diagram to the electronic small blackboard automatic generation device 10.

As shown in FIG. 3A, the information input device 20 has the same hardware resources as the electronic small blackboard automatic generation device 10. That is, it has a CPU 201 as an arithmetic processing unit, a ROM 202 as a non-volatile storage unit, a RAM 203 as a volatile storage unit, an SSD 204 for storing and storing input information, and the like. Further, it has an I / F 205 as an external interface, and the I / F 205 has a communication module 22 and an image reading module 21.

The image reading module 21 has a function of optically reading a structural drawing to generate image information and generating input data to the information input device 20. For example, the image reading module 21 corresponds to a scanner or the like.

The communication module 22 has a function of transmitting the read data to the electronic small blackboard automatic generation device 10. The data (input data) transmitted by the information input device 20 to the electronic small blackboard automatic generation device 10 is not limited to the data optically read by the image reading module 21, but is a data file including a structural drawing. But it may be.

In this embodiment, it is illustrated that the information input device 20 transmits the input data as a structural drawing to the electronic small blackboard automatic generation device 10 as a data file. Various types of architectural design drawings, which are input data to be transmitted to the electronic small blackboard automatic generation device 10, are assumed. As an example, the input data is a bar arrangement diagram showing the bar arrangement status of the reinforced concrete member. The bar arrangement diagram as input data may be PDF (Portable Document Form) data.

[Hardware configuration example of image display device 30]
FIG. 3B is a diagram showing a hardware configuration example of an image display device 30 as an information processing device for displaying and outputting an electronic small blackboard image automatically generated by the electronic small blackboard automatic generation device 10. .. Since there are many things in common with the hardware resources described so far, what is unique to the image display device 30 will be described.

As shown in FIG. 3B, a monitor 31 as an image display unit and a camera 32 as an image acquisition unit are connected to the I / F 305.

The monitor 31 displays the electronic small blackboard image generated by the electronic small blackboard automatic generation device 10. In addition, the subject image acquired at the angle of view of the camera 32 is displayed. Further, the monitor 31 has a touch panel function, and also has a function of accepting an imaging trigger operation.

The camera 32 is an imaging means for recording the state (construction status) of the construction site as an image, takes a picture according to the operation of the shutter provided on the monitor 31, and automatically generates the acquired image on the electronic small blackboard. Send to 10.

Here, a usage pattern of the image display device 30 will be described with reference to FIG. FIG. 4 is a front view showing an example of the appearance of the image display device 30. In the image display device 30, the monitor 31 is arranged on the front side and the camera 32 is arranged on the back side.

When the user records an image of a construction site using the image display device 30, the electronic small blackboard 311 generated and managed by the electronic small blackboard automatic generation device 10 is read out by a predetermined operation. At this time, since the camera 32 faces the construction site, the scenery of the construction site as the subject 312 is projected. Further, the electronic small blackboard 311 can be superimposed on a position that does not interfere with the projected subject 312. The position of the electronic small blackboard 311 can be moved to an arbitrary position by the user operating the monitor 31. Further, in one or both of the functions of the electronic small blackboard automatic generation device 10 and the image display device 30, an appropriate position can be automatically specified and superimposed.

In addition, the electronic small blackboard automatic generation device 10 analyzes the information contained in the subject 312 to individually identify the position of the reinforcement arrangement under the construction situation, and based on the identification result, each reinforcement arrangement. An image in which the bar arrangement marker 313 is superimposed may be generated.

The image display device 30 operates the shutter 314 at the stage where the display as shown in FIG. 4 is established, so that the part of the subject 312 that needs to be recorded and the position to avoid it are the electronic small blackboards 311. The superimposed image is recorded. At this time, in addition to the electronic small blackboard 311 as information for identifying the subject 312, the bar arrangement marker 313 may be superimposed and the image may be recorded.

The image recorded in the image display device 30 may be stored in the SSD 304 included in the image display device 30, or may be transmitted to the electronic small blackboard automatic generation device 10 and stored in the electronic small blackboard automatic generation device 10. It may be stored in the device 102.

[Embodiment of Electronic Small Blackboard Generation Method]
Next, the electronic small blackboard generation method realized by the functional processing executed in the electronic small blackboard automatic generation device 10 will be described with reference to the flowchart of FIG.

FIG. 5 illustrates the entire process executed as an electronic small blackboard generation method. For example, the electronic small blackboard automatic generation device 10 executes each process in the order of "learning process" and "execution process" as shown in the figure. The learning process is a process that is executed in advance of the execution process. That is, the learning process is a process of generating and learning a learning model by machine learning. The execution process is a process using the trained model after training.

The electronic small blackboard automatic generation device 10 does not have to execute the learning process and the execution process as continuous processes as illustrated in FIG. Therefore, the electronic small blackboard automatic generation device 10 does not have to learn the learning model by the learning process and then continuously execute the learning model. For example, the electronic small blackboard automatic generation device 10 may once create a trained model and then perform an execution process using the trained model as another opportunity.

Further, once the trained model has been generated, the trained model may be diverted and the electronic small blackboard automatic generation device 10 may start from the execution process. Hereinafter, a case where the electronic small blackboard automatic generation device 10 continuously executes the learning process and the execution process will be described as an example. That is, the learning model and the trained model may be configured to perform transfer learning (Transfer Learning), fine tuning (Fine tuning), or the like.

[Example of learning process]
First, the electronic small blackboard automatic generation device 10 accepts data input of the bar arrangement diagram Dt as teacher data for automatically generating an electronic small blackboard image by the information input device 20 (S501). As the data of the bar arrangement diagram input in S501, for example, the data generated by PDF may be acquired.

That is, in S501, the structural information input step is executed by the structural information input unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

Subsequently, the input bar arrangement diagram data is analyzed to analyze the table structure included in the bar arrangement data. Then, each cell can be individually identified from the framework included in the bar arrangement diagram formed by the table structure (S502).

Here, the structure of the bar arrangement diagram data input as the teacher data will be described. As illustrated in FIGS. 6A, 6B, and 6C, the bar arrangement diagram Dt as teacher data has various layouts. By performing learning processing using these as teacher data, characters and figures can be distinguished from the bar arrangement diagram Di input in the execution process, and identifiers and information elements of information necessary for generating an electronic small blackboard image can be obtained. Identify. Then, by associating each information, it is possible to generate a trained model for generating table data in order to generate an electronic small blackboard for each structure.

FIG. 7 is an example of a bar arrangement diagram Dt as teacher data having a certain layout. As shown in FIG. 7, the bar arrangement diagram Dt as teacher data contains information for individually distinguishing the column structures to be constructed in the construction work using the reinforced concrete structure. A cross-sectional view showing the number and arrangement of reinforcing bars used for each column (hereinafter referred to as "bean diagram P1"), and identification information for identifying the arrangement (number of floors) and position of the column structure in the building. M11 is included. In addition, element information M22 showing information about "upper end muscle", "lower end muscle", "star wrap", and "abdominal muscle" as information for showing the structure of each pillar associated with the identification information M11 is included.

In S502, a well-known image analysis processing technique is used to analyze architectural drawing data, and all vertical lines and horizontal lines are detected from the bar arrangement diagram Dt as shown in FIG. 7. Then, the intersection of each vertical line and the horizontal line is specified, and all the components P11 including the bean diagram P1 are specified. That is, in S502, the component specifying step is executed by the component specifying unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

Further, in S502, the characters included in each component P11 are analyzed by using the image analysis technique, and the size of the character area is extracted.

Further, in S502, the contour of the potential table included in each component P11 is acquired and defined by the coordinate position. Analysis is performed along the vertical and horizontal lines that make up the potential table to analyze whether or not the component P11 contains a plurality of cells.

FIG. 8 is a diagram illustrating an image of a potential table included in component P11. In the latent table P111 illustrated in FIG. 8, the frame line is not represented in the bar arrangement diagram Dt, but the table structure is analyzed by analyzing the structure of the component P11, and the information is obtained as a specific table structure as the analysis result. To be acquired. Then, each element included in the component P11 is specified based on the structure of the latent table P111.

Further, the latent table P111 includes a latent cell P112. In the example of FIG. 8, one latent table P111 includes a plurality of latent cells P112. Therefore, in FIG. 8, one component P11 includes one latent table P111, and the latent table P111 includes two latent cells P112.

That is, in S502, the cell extraction step is executed by the cell extraction unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

Further, in S502, the table structure identification step is executed by the table structure identification unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

As described above, the bar arrangement diagram Dt is analyzed, the table contour and the cell contour are specified in various styles and layouts, and the correlation between them is learned. That is, the identification of the bean diagram P1 included in the bar arrangement diagram Dt (teacher data), the identification of the identification information M11 and the identification of the element information M22, and the correlation between them are machine-learned (S503). That is, in S503, the learning step is executed from the learning unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

Then, in S503, an array state (layout) of information including an identifier for uniquely identifying the structure included in the bar arrangement diagram Dt, a drawing related to the identifier, and element information related to the identifier. Is used as training data to execute machine learning processing. As a result, the layout pattern as the data arrangement pattern used when automatically generating the electronic small blackboard 311 described later is distinguished based on the arrangement (position) of each data included in the bar arrangement diagram Dt and its contents. Generate a possible trained model. By inputting predetermined data to this trained model, data including each data listed above is automatically identified and extracted, and applied to the layout of the electronic small blackboard 311 to be electronic as an image. Output the small blackboard 311.

For example, a DB scan algorithm is used to perform clustering, etc., to generate a trained model that classifies into similar table groups even if the layouts are different in different bar arrangement diagrams D. Then, the trained model that makes the classified table group identifiable is stored in the electronic small blackboard automatic generation device 10, and the bar arrangement diagram Di input at the execution processing stage is included in the table group in the trained model. It is possible to identify which table style corresponds to.

That is, in S503, the table structure identification step is executed by the table structure identification unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

Based on the above processing, as a result, a trained model applicable for automatically generating an electronic small blackboard from the structural diagram data can be obtained.

For example, as illustrated in the region R of FIG. 7, even if the information associated with one identification information M11 on a one-to-one basis includes a plurality of cells, a learning model for distinguishing them is generated. .. As a result, it is possible to separately acquire the bean diagram P1 included in each component P11 and the element information M22 associated with the bean diagram P1 and execute a process of associating and storing them.

[Example of execution process]
Subsequently, an example of the execution process will be described with reference to FIG. After the trained model is generated in S503, the bar arrangement diagram Di as input data is newly input from the information input device 20 to the electronic small blackboard automatic generation device 10 (S504). The electronic small blackboard automatic generation device 10 executes a content specifying process using the trained model for the input bar arrangement diagram Di (S505). As a result, the table data used for generating the electronic small blackboard is generated.

For example, FIG. 9 illustrates the bar arrangement diagram Di. Using the trained model as the bar arrangement diagram Di as the input data, each component P11 as illustrated in FIG. 8 and the like is specified. Then, the positions of the latent table P111 and the latent cell P112 are acquired for each component P11. Then, from the acquired latent table P111 and latent cell P112, text information (character information) and figure information (image information) necessary for embedding in the electronic small blackboard image are acquired and classified.

Further, the trained model is used to specify the information corresponding to the identification information M11, and further, the unit for generating and distinguishing the electronic small blackboard image is also specified. For example, the character information corresponding to the head position of the column of the information specified as the identification information M11 is specified as "information indicating the floor". In addition, the character information corresponding to the head position of the line is specified as "information indicating the position on the same floor". Based on these specified information, table data having an information structure that distinguishes "electronic small blackboard information corresponding to a structure constructed as a structure on the same floor" for each position of each structure is generated.

That is, in S505, the information identification step is executed by the information identification unit realized in the processing program executed by the electronic small blackboard automatic generation device 10. In the information identification step, the above data table is generated.

Then, based on the generated table data, an electronic small blackboard image specified from the information included in the bar arrangement diagram Di is generated (S506). For example, the electronic small blackboard 311 illustrated in FIG. 10 (b) is generated from the bar arrangement diagram Di exemplified in FIG. 10 (a).

Based on the information about the identification information M11 specified from the bar arrangement diagram Di and the information of the component P11 specified as the element information M22, the electronic small blackboard 311 as illustrated in FIG. 10A is generated.

The reference numeral M111 in FIG. 10A is a portion corresponding to the identification information M11, is specified as character information, and corresponds to an information identifier included in the electronic small blackboard 311. The layout of this information identifier shall also be appropriately selected according to the layout of the bar arrangement diagram Di.

Further, the reference numeral M221 is a portion corresponding to the element information M22, and changes for each component P11.

Further, FIG. 11 is another example showing a part of the data table specified and generated from the bar arrangement diagram Di. Based on the data table of FIG. 11, the electronic small blackboards 311 illustrated in FIGS. 12 (a) and 12 (b) are generated.

In S506, the electronic small blackboard 311 generated as shown in FIGS. 10B and 12 is stored in the storage device 102. That is, in S506, the image storage step is executed by the image storage unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

After that, when the recording operation of the construction site is performed using the image display device 30, the stored electronic small blackboard 311 is read out and superimposed on the image of the construction site and recorded (S507).

That is, in S507, the electronic small blackboard image output step is executed by the electronic small blackboard image output unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

Subsequently, it is determined whether or not any of the generated electronic small blackboards 311 is not used for recording the construction site (S508). That is, it is determined that there is a structure whose construction status to be recorded is not recorded. As a result of the determination, when a structure in an unrecorded state exists (S508), the electronic small blackboard automatic generation device 10 outputs a notification regarding the unrecorded structure to the image display device 30 (S509). As a result, recording omission can be suppressed.

That is, in S509, the information notification step is executed by the information notification unit realized in the processing program executed by the electronic small blackboard automatic generation device 10.

Return to FIG. In S510, the electronic small blackboard automatic generation device 10 determines whether or not to additionally perform the learning process. For example, when it is determined that the accuracy needs to be further improved as a result of executing the execution process, the electronic small blackboard automatic generation device 10 additionally performs a learning process.

Next, when the learning process is additionally performed (S510: YES), the process of the electronic small blackboard automatic generation device 10 returns to S501, that is, the learning process is additionally performed. On the other hand, when the learning process is not additionally performed (S510: NO), the electronic small blackboard automatic generation device 10 ends the entire process.

As described above, according to the method for generating an electronic small blackboard according to the present embodiment, when recording an image of the state of a construction site from a bar arrangement diagram as an example of a structural drawing, information related to the construction site is displayed. The electronic small blackboard image of is automatically generated by software technology. In this electronic small blackboard image generation process, the influence of the fluctuation of the bar arrangement diagram Dt as the teacher data and the way in which the information is associated with the bean diagram P1 contained therein is eliminated, and the electronic small blackboard image of a predetermined format is used. Use a trained model that has performed machine learning to generate.

In addition, when the image of the construction site is photographed using the generated electronic small blackboard image, the corresponding electronic small blackboard image is automatically superimposed to record the "electronic blackboard recorded image", and based on the recording state. It can also be used to manage the progress of the entire construction.

In addition, instead of the bar arrangement diagram Dt as teacher data, the BIM model in BIM (Building Information Modeling) is used, and the BIM model is input instead of the bar arrangement diagram Di, and the required electronic small blackboard 311 is used. You may generate an image of.

Further, CAD data may be used instead of the BIM model.

[Example of network structure]
Here, FIG. 13 shows an example of the network structure of the trained model generated by the above-mentioned electronic small blackboard generation method. The learning model generated by the electronic small blackboard automatic generation device 10 and the AI based on the learned model process image data and the like as a bar arrangement diagram Di with the following network structure.

For example, the AI may have a network structure having an input layer L1, a hidden layer L2, and an output layer L3. Specifically, the AI may be a network structure having a Convolution Natural Network (convolutional neural network, CNN) or the like as shown in the figure.

Here, the input layer L1 is a layer for inputting the bar arrangement diagram Di as input data as an image. The hidden layer L2 is a layer that performs processing such as convolution, pooling, normalization, or a combination thereof with respect to the image input by the input layer L1. The output layer L3 is a layer that outputs the result processed by the hidden layer L2. For example, the output layer L3 is composed of a fully connected layer or the like.

Convolution is, for example, an image or a feature map generated by performing a predetermined process on an image based on a filter, a mask, a kernel (hereinafter, simply referred to as "filter"), or the like. On the other hand, it is a process to generate a feature map by performing a filter process.

Specifically, the filter is data used for calculating the filter coefficient (sometimes referred to as "weight" or "parameter") by multiplying the pixel value of the image or the feature map. The filter coefficient is a value determined by learning or setting.

The convolution process is a process of multiplying each pixel value of the pixels constituting the image or the feature map by a filter coefficient to generate a feature map having the calculation result as a component. When the convolution process is performed in this way, the features of the image or the feature map can be extracted. The feature is, for example, the edge component or the result of statistical processing around the target pixel.

Further, when the convolution process is performed, the target image or the subject indicated by the feature map is an image or feature map that shifts up and down, shifts left and right, shifts diagonally, rotates, or is a combination thereof. However, similar characteristics can be extracted.

Pooling is a process of extracting features and generating a feature map by performing processes such as average calculation, minimum value extraction, and maximum value extraction for a target range. That is, the pooling is max pooling, avg pooling, or the like.

For convolution and pooling, pretreatment such as zero padding may be performed.

By convolution, pooling, or a combination thereof as described above, a so-called data amount reduction effect, syntheticity, movement invariance, or the like can be obtained.

Normalization is, for example, a process of adjusting the variance and the mean value. The normalization includes the case where it is performed locally. Then, when normalization is performed, the data becomes a value or the like within a predetermined range. Therefore, the data can be easily handled in the subsequent processing.

Fully connected is a process of dropping data such as a feature map into an output.

For example, the output is in the form of a binary output, such as "YES" or "NO". In such an output format, full join is a process of joining nodes based on the features extracted by the hidden layer L2 so that one of the two types can be concluded.

On the other hand, when there are three or more types of outputs, total coupling is a process of performing a so-called softmax function or the like. In this way, by full coupling, classification (including the case where an output indicating a probability is performed) can be performed by a maximum likelihood estimation method or the like.

It should be noted that the present invention is not limited to each of the above-described embodiments, and various modifications can be made without departing from the technical gist thereof, and the technical concept included in the claims is described. All of the matters are the subject of the present invention. Although the above embodiment shows a suitable example, those skilled in the art can realize various modified examples from the disclosed contents. Such modifications are also included in the technical scope described in the claims.

10: Electronic small blackboard automatic generation device 11: Communication module 20: Information input device 21: Image reading module 22: Communication module 30: Image display device 31: Monitor 32: Camera 101: Arithmetic device 102: Storage device 103: Input device 104 : Output device 105: I / F
201: CPU
202: ROM
203: RAM
205: I / F
305: I / F
311: Electronic small blackboard 312: Subject 313: Reinforcement marker 314: Shutter

Claims (8)

Structural information input unit for inputting structural drawings used for building work,
An identifier for uniquely identifying a structure included in the structural diagram by analyzing the structural diagram, a cross-sectional view showing the structure of the structure distinguished by the identifier, and a structural structure distinguished by the identifier. A learning unit that generates a trained model that has been machine-learned using the layout pattern as an array pattern as training data for the element information indicating the specifications of
The electronic small blackboard that distinguishes the layout pattern with respect to the input of the structural drawing by the trained model and superimposes it on the image that records the construction status of the structure for each structure constructed in the construction work. An electronic small blackboard image output unit that outputs an image based on the layout pattern,
An electronic small blackboard generator characterized by including. An image storage unit that stores a recorded image in which the electronic small blackboard image is superimposed on an image that records the construction status, and an image storage unit.
Among the structures for storing the recorded image, the information notification unit for outputting information for notifying the existence of the structure in which the recorded image is not stored is included.
The electronic small blackboard generator according to claim 1. The learning unit identifies a component included in the structural diagram, identifies a table structure included in the component, and performs machine learning of the layout pattern.
The electronic small blackboard generator according to claim 1 or 2. The learning unit
A table structure identification unit that identifies the table structure in the structural diagram,
The component identification part that identifies the component held in the identified table structure, and the component identification part.
A cell extractor that extracts cells included in the component,
Including an information specifying unit for specifying image information and character information included in the cell,
The electronic small blackboard generator according to claim 1 or 2. It is an electronic small blackboard generation method realized by the electronic small blackboard image generation processing by cloud computing provided by a computer.
The computer
An input step that accepts the input of structural drawings used for building work,
An identifier for uniquely identifying a structure included in the structural diagram by executing machine learning using the structural diagram as teacher data, a sectional view showing the structure of the structure distinguished by the identifier, and the identifier. A learning step to generate a trained model for analyzing a layout pattern as an array pattern of element information distinguished in the above and showing the specifications of the structure.
For the input of the newly input structural drawing, the layout pattern is distinguished by using the trained model, and each structure to be constructed in the construction work is superimposed on the image recording the construction status of the structure. An electronic small blackboard image output step that outputs an electronic small blackboard image based on the layout pattern, and
An electronic small blackboard generation method characterized by including. An image storage step for storing a recorded image in which the electronic small blackboard image is superimposed on an image for recording the construction status, and an image storage step.
Among the structures in which the recorded image should be stored, the information notification step for outputting information notifying the existence of the structure in which the recorded image is not stored is further included.
The method for generating an electronic small blackboard according to claim 5. The learning step includes a step of specifying a component included in the structural diagram, specifying a table structure included in the component, and performing machine learning of the layout pattern.
The method for generating an electronic small blackboard according to claim 5 or 6. In the learning step
A table structure identification step for identifying the table structure in the structure diagram,
A component identification step that identifies the components held in the identified table structure, and
A cell extraction step that extracts cells contained in the component, and
Including an information specifying step for specifying image information and character information contained in the cell.
The method for generating an electronic small blackboard according to claim 5 or 6.

Images