JP2021174284A - Construction site image acquisition system, construction site image acquisition device, and construction site image acquisition program - Google Patents

Abstract

To acquire learning data for realizing highly efficient learning of a learning device for determining the quality of a construction situation shown in a construction site image.SOLUTION: A memory 42 of a server 16 stores guide information 48 indicating an appropriate image which is an appropriate construction site image as learning data of a learning device 44. Prior to the acquisition of the construction site image on a site terminal 14, a display control unit 54 transmits the guide information 48 to the site terminal 14 and displays it on a display of the site terminal 14.SELECTED DRAWING: Figure 3

Description

The present invention relates to a construction site image acquisition system, a construction site image acquisition device, and a construction site image acquisition program.

Conventionally, at various construction sites including construction sites and construction sites, it has been practiced to take a picture of a construction object with a camera and acquire a construction site image. Conventionally, construction site images have been used as targets for image analysis in addition to being attached to daily reports for site workers to report work to managers such as site supervisors and contractors. By analyzing the construction site image, it is possible to obtain various information regarding the status and state of the construction site (hereinafter referred to as "construction status") at the construction site.

For example, in Patent Document 1, by analyzing a construction site image obtained by photographing a construction site, a building element included in the construction site image is specified, and a construction process is specified based on the specified building element. The situation determination system is disclosed.

JP-A-2017-107443

By the way, in recent years, construction site images have been analyzed using a learning device. In particular, the quality of the construction situation shown in the construction site image has been judged by visual inspection by a human such as an administrator in the past, but by letting the learning device learn sufficiently, the learning device is made to judge this. Will be able to.

Various methods have been proposed as learning methods for the learner, but in general, a huge amount of learning data is required to train the learner. Specifically, a huge number of construction site images are required in order to learn the learning device for determining the quality of the construction status shown in the construction site image. As described above, since the construction site image is acquired by the site worker at the construction site, it is preferable to use the construction site image acquired at the site every day as learning data.

Here, the learning efficiency of the learning device may fluctuate depending on the content of the construction site image as the learning data. The learning efficiency of the learning device is an index showing the proficiency level of the learning device (accuracy of quality judgment of the construction situation based on the construction site image) with respect to the number of learning data (number of samples). In order to improve the learning efficiency of the learning device that judges the quality of the construction status shown in the construction site image, the learning data includes the construction site image (positive) in which the construction object showing the excellent construction status is properly photographed. A large number of construction site images (negative samples) are required, in which the construction object showing the poor construction status is properly photographed.

However, when acquiring a construction site image as learning data, for example, the construction object to be photographed is not properly captured or is out of focus, and the learning data (regardless of the quality of the construction situation). Inappropriate construction site images may be acquired. In this specification, an image of a construction site that is inappropriate as learning data is generally referred to as an inappropriate image regardless of whether the construction situation is good or bad. On the other hand, regardless of the quality of the construction situation, an image of the construction site that is appropriate as learning data is generally called an appropriate image.

Construction site images are acquired by an unspecified number of site workers. In particular, at construction sites, there are many changes in on-site workers, and it is not uncommon for on-site workers who are unfamiliar with photography to take images of construction sites. Further, when taking a picture of a construction site, it is necessary to take a picture of various construction objects according to, for example, the construction process. For this reason, it has been difficult for field workers to take an appropriate image in the past.

Even if an inappropriate image is used as learning data when learning a learning device that determines the quality of the construction situation shown in the construction site image, the learning of the learning device does not progress much. Therefore, it can be said that the more inappropriate images are included in the learning data, the lower the learning efficiency of the learning device. In other words, in order to train the learner efficiently, it is preferable that the learning data includes more appropriate images.

An object of the present invention is to acquire learning data that realizes highly efficient learning of a learning device that determines the quality of a construction situation shown in a construction site image.

The invention according to claim 1 is shown on the display unit of a camera that captures the construction situation at the construction site and acquires the construction site image on the construction site image prior to the site worker taking the construction site image. The learning device uses a display control unit that displays guide information indicating an appropriate image, which is an appropriate construction site image, as learning data when learning a learning device that determines the quality of the construction situation, and a construction site image acquired by the camera. It is a construction site image acquisition system characterized by having a learning data storage unit for storing as learning data of the above.

Desirably, the camera acquires a construction site image for each construction process, and the display control unit learns the learner for determining the quality of the construction status of the selection process, which is the selected process. The guide information indicating an appropriate image is displayed, and the learning data storage unit associates the selection process with the construction site image acquired by the camera and stores the learning data as learning data of the learning device.

Desirably, the guide information is characterized in that the main points for obtaining the proper image are highlighted.

Desirably, the guide information is characterized in that it indicates an appropriate composition which is the composition of the appropriate image.

Desirably, the display control unit is characterized in that NG information indicating an improper image, which is an improper construction site image, is displayed on the display unit of the camera as learning data when learning the learning device. ..

Desirably, the learning processing unit that learns the learning device so as to output the quality of the construction situation indicated by the input construction site image using the learning data, and has a composition suitable for the proper composition. It is characterized by further including a learning processing unit for learning the learning device by using an appropriate image as the learning data.

Desirably, the determination processing unit for determining whether or not the construction site image acquired by the camera is the proper image, and the construction site image having a composition conforming to the proper composition is the proper image. The learning processing unit further includes a determination processing unit for determining, and the learning processing unit trains the learner using the learning data including the construction site image determined by the determination processing unit to be the proper image. It is a feature.

Desirably, it is a notification processing unit that notifies the site worker of the determination result by the determination processing unit, and the construction site image acquired by the camera is inappropriate as learning data when learning the learner. When the determination processing unit determines that the image is appropriate, the display unit of the camera displays NG information indicating the inappropriate image determined by the determination processing unit in the past.

Further, according to the present invention, the construction status shown in the construction site image is shown on the display unit of the camera that captures the construction status at the construction site and acquires the construction site image before the site worker captures the construction site image. A display control unit that displays guide information indicating an appropriate image, which is an appropriate construction site image as learning data when learning the learning device for determining the quality of the camera, and learning of the learning device using the construction site image acquired by the camera. It is a construction site image acquisition device characterized by including a learning data storage unit for storing as data.

Further, according to the present invention, the computer is shown on the display unit of the camera that captures the construction situation at the construction site and acquires the construction site image, and is shown on the construction site image prior to the site worker taking the construction site image. The display control unit that displays guide information indicating an appropriate image, which is an appropriate image of the construction site as learning data when learning the learning device that determines the quality of the construction situation, and the construction site image acquired by the camera are learned. It is a construction site image acquisition program characterized by functioning as a learning data storage unit that stores as learning data of a vessel.

According to the present invention, it is possible to acquire learning data that realizes highly efficient learning of a learning device that determines the quality of a construction situation shown in a construction site image.

It is a block diagram of the construction site photography system which concerns on this embodiment. It is a block diagram of the site terminal which concerns on this embodiment. It is a block diagram of the server which concerns on this embodiment. It is a figure which shows the example of the process selection screen. It is a figure which shows the example of the shooting screen. It is a figure which shows the display example of a guide image. It is a figure which shows the display example of the guide image which emphasized the main point. It is a figure which shows the display example of an inappropriate image. It is a figure which shows the example of a proper image. It is a figure which shows the example of an inappropriate image. It is a flowchart which shows the flow of the learning data acquisition processing in the construction site photography system which concerns on this embodiment.

FIG. 1 is a schematic configuration diagram of a construction site image acquisition system 10 according to the present embodiment. The construction site image acquisition system 10 is a management terminal 12 used by a manager (site supervisor, construction owner, etc.) who manages site workers who work at the construction site or who manages the construction site itself, and 1 at the construction site. Alternatively, it is configured to include a field terminal 14 and a server 16 used by a plurality of field workers. The management terminal 12, the field terminal 14, and the server 16 are connected to each other so as to be able to communicate with each other by a communication line 18 such as the Internet or a mobile phone line. Although only one management terminal 12 is shown in FIG. 1, the construction site image acquisition system 10 may include a plurality of management terminals 12 used by a plurality of managers. Further, although only one site terminal 14 is shown in FIG. 1, the construction site image acquisition system 10 may include a plurality of site terminals 14 used at a plurality of construction sites.

The site terminal 14 has a camera function, and the site terminal 14 photographs the construction situation at the construction site by the operation of the site worker, acquires the construction site image, and transmits the construction site image to the server 16. The construction status represents the status and state of the construction site. As an example, the construction status is the construction result, and the construction object showing the construction result is photographed as a construction site image. However, the construction situation is not limited to the construction result, but may include various situations at the construction site. The construction site image is managed by the server 16 and analyzed. By accessing the server 16 from the management terminal 12, the administrator can view the construction site image or the analysis result thereof. The construction site image can be used for various purposes, but in particular, it is used as learning data for learning the learning device described later. Therefore, the construction site image acquisition system 10 exerts the function of the construction site image acquisition system for learning data.

The management terminal 12 is a mobile terminal such as a personal computer or a tablet terminal. The management terminal 12 has a communication interface for communicating with a server 16 and the like, a memory composed of a hard disk and RAM, a display composed of a liquid crystal display and the like, an input interface composed of a mouse, a keyboard or a touch panel, and the like. , A processor composed of a CPU, a microcomputer, and the like.

FIG. 2 is a schematic configuration diagram of the field terminal 14. The field terminal 14 is a mobile terminal such as a tablet terminal. In this embodiment, the site terminal 14 is permanently installed at each construction site.

By using the site terminal 14, it is possible to manage the entry / exit of each site worker entering / exiting the construction site. Specifically, first, the field worker logs in (authenticates) to the field terminal 14 by inputting identification information such as a user ID and biometric information. Site worker DB (database) in which site worker information (for example, the name of the site worker and the name of the contractor to which the site worker belongs) related to each site worker and the identification information of each site worker are associated with each other. Is stored in the site terminal 14 or the server 16, and the site terminal 14 identifies and authenticates the site worker based on the input identification information while referring to the site worker DB. Then, the site worker inputs the entry time and the exit time to the construction site to the site terminal 14. The entry time and exit time are transmitted to the server 16 and stored in the server 16.

Further, the site terminal 14 may receive the construction content information indicating the construction content to be performed at the construction site where the site terminal 14 is permanently installed from the server 16 and store it in the memory 26 described later. As the construction content information, the construction content information indicating the construction content to be performed on that day may be received every morning, or the construction content information indicating the construction content to be performed in the week may be received every Monday morning. The construction content information may indicate, for example, the construction process (for example, foundation construction, foundation curing, external scaffolding, etc.) at the construction site. As a result, the on-site worker can grasp the construction contents to be performed on that day by browsing the on-site terminal 14.

The communication interface 20 includes, for example, a network adapter and the like. The communication interface 20 exhibits a function of communicating with a device such as a server 16 via a communication line 18.

The input interface 22 includes, for example, a touch panel and buttons. The input interface 22 exerts a function of receiving input of various information from field workers.

The display 24 is a display unit composed of, for example, a liquid crystal panel. Various screens are displayed on the display 24.

The memory 26 includes, for example, an eMMC (embedded Multi Media Card), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like. A program for operating each part of the field terminal 14 is stored in the memory 26. Further, the memory 26 stores the above-mentioned construction content information received from the server 16, and also temporarily stores the field worker information of the field worker who is currently authenticated (currently using the field terminal 14). Will be done.

The camera 28 includes a lens, an image sensor such as a CCD (Charge-Coupled Device) image sensor, and a DSP (Digital Signal Processor) that performs various image processing. The camera 28 captures the construction situation at the construction site and acquires the construction site image. In particular, in the present embodiment, the on-site worker takes a picture of a construction object showing the construction status of the construction every time the work of one process of the construction is completed. That is, in the present embodiment, the camera 28 acquires the construction site image for each process. Further, by processing the DSP included in the camera 28, the construction site name, the construction process, the shooting date and time, and the construction were performed based on the construction content information and the site worker information stored in the memory 26 (construction site image). The name of the person in charge of construction (photographed) and the name of the contractor to which the name of the person in charge of construction belongs are superimposed on the construction site image. This information may be given as metadata (attribute data) of the construction site image. To reiterate, the construction site image acquired by the camera 28 is used as learning data for learning the learning device described later.

The processor 30 is composed of, for example, a microcontroller or the like. The processor 30 controls each part of the field terminal 14 according to the program stored in the memory 26.

FIG. 3 is a schematic configuration diagram of the server 16 as a construction site image acquisition device. The server 16 is composed of a device such as a server computer. Each function of the server 16 described below may be realized by one server computer, or may be realized by the cooperation of a plurality of server computers.

The communication interface 40 includes, for example, a network adapter and the like. The communication interface 40 exhibits a function of communicating with the management terminal 12 and the field terminal 14 via the communication line 18.

The memory 42 as a learning data storage unit includes, for example, an HDD (Hard Disc Drive), an SSD (Solid State Drive), an eMMC, a ROM, a RAM, and the like. The memory 42 stores a construction site image acquisition program that operates each part of the server 16. Further, the memory 42 stores construction content information regarding a plurality of construction sites managed by the server 16. The construction content information may be created by the manager of each construction site and uploaded from the management terminal 12 to the server 16. Further, as shown in FIG. 3, the memory 42 stores the learning device 44, the learning data 46, the guide information 48, and the NG information 50.

The learner 44 is configured by a model such as a CNN (Convolutional Neural Network). The learning device 44 determines the quality of the construction situation shown in the construction site image based on the input construction site image. For example, if the construction site image shows the correct construction status, the (learned) learner 44 determines that the construction status is excellent, and the construction site image has a bad construction status (for example, the correct parts are not used, or If it indicates that the construction is incorrect), the learner 44 determines that the construction status is defective. In particular, as described above, in the present embodiment, the construction site image is acquired for each construction process, so the learner 44 determines the quality of the construction status of the selected process based on the input construction site image. do.

The learning data 46 is data for training the learner 44. The learning data 46 is configured by associating a construction site image acquired by the site terminal 14 with information indicating the quality of the construction status indicated by the construction site image. In the present embodiment, since the construction site image and the information indicating the process are associated and transmitted from the site terminal 14, the learning data 46 is distinguished for each process and stored in the memory 42. In the present embodiment, the learning data 46 is configured to include an appropriate appropriate image as the learning data of the learning device 44 among the construction site images transmitted from the site terminal 14. It is desirable that the learning data 46 is composed of only proper images, but this is not always the case, and the learning data 46 may include not only proper images but also inappropriate images. The details of the learning data 46 will be described later.

The guide information 48 is information indicating an appropriate image which is an appropriate construction site image as learning data of the learning device 44. Since the conditions that can be an appropriate image differ for each construction process, guide information 48 is prepared for each construction process. The guide information 48 includes, for example, a construction object to be photographed to show the construction status of the target process, the size of the construction object in the construction site image, or a portion of the construction object to be photographed in particular. Information indicating (for example, the model number of the member) is included.

In this embodiment, the guide information 48 is prepared in the form of an image. Hereinafter, the image as the guide information 48 is referred to as a guide image. The guide image may be a still image or a moving image. Further, in the present embodiment, the guide image is a photograph, but it may be a picture such as a computer graphic as long as the field worker who sees the guide image can suitably grasp the appropriate image. Since the guide image only shows an appropriate image as learning data, it does not necessarily have to show an excellent construction situation. Further, the guide information 48 may be character information in which the conditions for the construction site image to be an appropriate image are expressed in characters.

In the guide information 48, it is preferable that the main points for the construction site image to be an appropriate image are shown. In particular, when the guide information 48 is a guide image, an inexperienced field worker may not be able to grasp which part of the guide image is the main point (point) for obtaining an appropriate image. Therefore, it is preferable that information indicating the main points for obtaining an appropriate image is added to the guide image. Such information includes, for example, a pixel value (luminance or color) different from other pixels attached to the pixel corresponding to the main point, or an annotation attached to the main point. Further, even when the guide information 48 is character information, for example, when the number of characters is large, an inexperienced field worker may not be able to grasp what is the main point (point). Therefore, it is preferable that information indicating the main points for obtaining an appropriate image is added to the character information as well. Such information includes, for example, a bold setting of characters corresponding to the main points, or a font (color, etc.) different from other characters.

Further, the guide information 48 preferably indicates an appropriate composition which is an appropriate image composition (in other words, a camera angle). The composition means the arrangement of each object to be photographed (including the object to be photographed) in the image of the construction site, the distance to each object to be photographed, or the position of the light source (for example, the sun) in the image of the construction site. It is a thing. When the guide information 48 is a guide image, the composition of the proper image is shown as it is in the guide image. When the guide information 48 is character information, character information indicating an appropriate composition is included.

The NG information 50 is information indicating an improper image which is an improper construction site image as learning data of the learning device 44. The NG information 50 in the present embodiment is an inappropriate image itself, but the NG information 50 may be character information. The NG information 50 may be prepared in advance by an administrator or the like, but in the present embodiment, the NG information 50 is a construction site image (that is, inappropriate) transmitted from the site terminal 14 determined to be an inappropriate image in the past. Appropriate image). Examples of inappropriate images that provide NG information 50 include, for example, an image that does not properly include the construction object to be photographed in the process, an image that is out of focus, or an image that is too dark (too bright) as a whole. However, it is not limited to this. Similar to the guide information 48, one or a plurality of NG information 50 is prepared for each construction process. The NG information 50 of a certain process means an improper image in which the image of the construction site of the process is taken and becomes inappropriate. Further, similarly to the guide information 48, it is preferable that the NG information 50 is added with information indicating the main points of the improper image. Details will be described later, but in particular, in the present embodiment, even a construction site image that can be generally judged to be an appropriate image does not conform to the appropriate composition indicated by the above-mentioned guide information 48 (typically, a photographed image). The NG information 50 also includes a construction site image) that appropriately includes the construction object to be to be constructed but has a composition different from the proper composition.

The processor 52 includes, for example, a CPU (Central Processing Unit) and at least one of a dedicated processing device (for example, GPU (Graphics Processing Unit), ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), etc.). Consists of. The processor 52 may be configured not by one processing device but by the cooperation of a plurality of processing devices that are physically separated from each other. As shown in FIG. 3, the processor 52 exerts functions as a display control unit 54, a determination processing unit 56, a notification processing unit 58, and a learning processing unit 60 by the construction site image acquisition program stored in the memory 42.

The display control unit 54 controls the display 24 of the field terminal 14 to display various screens by transmitting various information to the field terminal 14. In particular, the display control unit 54 causes the site worker to display the guide information 48 on the display 24 of the site terminal 14 prior to taking the construction site image. Hereinafter, the details of the processing of the display control unit 54 will be described with reference to FIGS. 4 to 8.

First, when the work of a certain process is completed at the construction site, the site worker logs in to the site terminal 14 and inputs a display instruction on the process selection screen. The processor 30 of the field terminal 14 causes the display 24 to display the process selection screen as shown in FIG. 4 based on the display instruction. As shown in FIG. 4, a plurality of process selection buttons 80 corresponding to a plurality of processes are displayed on the process selection screen. The type of process displayed on the process selection screen is determined in advance or is determined based on the above-mentioned construction content information transmitted from the server 16. When the site worker touches the process selection button 80 corresponding to the process for which the work has been completed at the construction site, the process corresponding to the touched process selection button 80 is selected. The process may not be selected by the field worker, but the processor 30 may be selected based on the above-mentioned construction content information (automatically selected from the viewpoint of the field worker). In the present specification, the process selected in this way is referred to as a selection process. Here, it is assumed that the selection process is "foundation work".

When the selection process is selected, the processor 30 causes the display 24 to display a shooting screen as shown in FIG. A shooting button 82, a guide button 84, and an NG button 86 are displayed on the shooting screen. When the site worker touches the shooting button 82, the camera 28 takes a picture and the construction site image is acquired. The processor 30 associates the construction site image acquired by the camera 28 with the selection process information indicating the selection process, and transmits the image to the server 16.

The field worker can touch the guide button 84 prior to touching the shooting button 82, that is, prior to the camera 28 acquiring the construction site image. When the guide button 84 is touched by the field worker, the guide button operation signal indicating that the guide button 84 has been operated and the selection process information indicating the selection process are transmitted from the field terminal 14 to the server 16.

When the display control unit 54 of the server 16 receives the guide button operation signal and the selection process information from the field terminal 14, the received selection is made from the plurality of guide information 48 prepared for each process stored in the memory 42. The guide information 48 (guide image in this embodiment) of the selection process indicated by the process information is specified. Here, a guide image corresponding to the process "foundation work" is specified. Then, the specified guide image is transmitted to the site terminal 14, and the guide image is displayed on the display 24 of the site terminal 14.

FIG. 6 shows a guide image 88 displayed on the display 24 of the field terminal 14. By displaying the guide image 88 on the display 24, it can be seen that the site worker may acquire a construction site image similar to the guide image 88 as the construction site image as learning data. Since the guide image 88 shows an appropriate image which is an appropriate construction site image as the learning data 46 of the learning device 44, by displaying the guide image 88 on the display 24, the site worker is made to acquire an appropriate image. be able to. If the acquired proper image shows a good construction situation, it is a positive sample, and if the acquired proper image shows a bad construction situation, it is a negative sample.

A back button 90 is provided on the screen on which the guide image 88 is displayed, and when the back button 90 is touched, the screen can be returned to the shooting screen (see FIG. 5). It should be noted that shooting may be possible on the screen on which the guide image 88 is displayed. In other words, the guide image 88 may be displayed on the shooting screen.

Further, as described above, the guide information 48 may be character information, and in that case, when the on-site worker touches the guide button 84, the character information as the guide information 48 is displayed on the display 24. ..

In the present embodiment, the guide information 48 is displayed on the display 24 by operating the guide button 84 by the field worker, but when the field worker or the processor 30 selects the selection process, a picture is taken. Guide information 48 may be displayed on the display 24 prior to the screen. In that case, when the selection process is selected, the selection process information is transmitted from the site terminal 14 to the server 16, and the display control unit 54 provides the guide information 48 corresponding to the selection process to the site terminal based on the selection process information. The information is transmitted to 14, and the processor 30 of the field terminal 14 immediately displays the received guide information 48 on the display 24.

Further, when information indicating a point for obtaining a proper image is added to the guide information 48, the display control unit 54 sets the guide information 48 displayed on the display 24 as a proper image based on the information. It is preferable to highlight the points to be achieved. FIG. 7 shows a guide image 88a in which the main points are highlighted. For example, when the main point that the construction site image of the process "foundation work" becomes an appropriate image is that the foundation, which is the construction status (here, the construction result of the foundation work), is appropriately photographed, as shown in FIG. , In the guide image 88a, the pixels corresponding to the foundation are highlighted. As the highlighting mode, for example, the pixel value (luminance value or color) of the pixel may be different from the pixel value of the other pixels, but other modes may be used.

Further, when the guide information 48 indicates the proper composition of the proper image, it is preferable that the display control unit 54 displays the proper composition on the display 24 based on the information. When the guide information 48 is the guide image 88, since the proper composition is shown in the guide image 88 as a matter of course, it can be said that the proper composition is shown only by displaying the guide image 88. Further, in addition to the guide image 88, the main points of the proper composition may be additionally displayed by character information or the like. Even when the guide information 48 is character information, the display control unit 54 may further display character information indicating an appropriate composition on the display 24.

The field worker can touch the NG button 86 (see FIG. 5) prior to touching the shooting button 82, that is, prior to the camera 28 acquiring the construction site image. When the NG button 86 is touched by the field worker, the NG button operation signal indicating that the NG button 86 has been operated and the selection process information indicating the selection process are transmitted from the field terminal 14 to the server 16.

When the display control unit 54 receives the NG button operation signal and the selection process information from the field terminal 14, the received selection process information is selected from the plurality of NG information 50 prepared for each process stored in the memory 42. The NG information 50 of the selection process shown is specified. Here, the NG information 50 corresponding to the process "foundation work" is specified. Then, the specified NG information 50 is transmitted to the site terminal 14, and the NG information 50 is displayed on the display 24 of the site terminal 14.

FIG. 8 shows an improper image 92 as NG information 50 displayed on the display 24 of the field terminal 14. When the NG information 50 is character information, the characters are displayed as the NG information 50. By displaying the NG information 50 on the display 24, it can be seen that the site worker must not acquire an image as shown by the NG information 50 as a construction site image as learning data. Further, when the NG information 50 is added with information indicating the main points for obtaining an inappropriate image, the display control unit 54 is inappropriate in the NG information 50 displayed on the display 24 based on the information. The main points for becoming an image may be highlighted.

Returning to FIG. 3, the determination processing unit 56 determines whether the construction site image acquired by the site terminal 14 and transmitted from the site terminal 14 to the server 16 is an appropriate image or an improper image. A known image processing method can be used for the determination processing by the determination processing unit 56, but in the present embodiment, the determination processing unit 56 includes a template image which is a template of an appropriate image prepared in advance in the memory 42 and a site. By performing matching processing (template matching) with the construction site image received from the terminal 14, it is determined whether or not the construction site image is an appropriate image. As described above, since the characteristics of the appropriate image differ depending on the construction process, it is preferable that the template image is prepared for each process in the memory 42. The determination processing unit 56 specifies the template image of the process indicated by the selection process information received from the site terminal 14 together with the construction site image, and performs template matching between the specified template image and the construction site image. When the guide information 48 is a guide image, the guide image can be used as the template image.

The determination processing unit 56 may determine that the construction site image conforming to the appropriate composition shown in the guide information 48 is an appropriate image. In other words, the determination processing unit 56 may determine that the construction site image that does not conform to the proper composition shown in the guide information 48 is an inappropriate image. Here, conforming to the proper composition is not only when the composition of the construction site image completely matches the proper composition, but also to some extent (above the predetermined similarity threshold value), the composition of the construction site image and the proper composition are similar. It is also included when it is. For example, if the above template image has the proper composition shown in the guide information 48, the construction site image matching the proper composition is determined to be the proper image by the above template matching.

FIG. 9 shows a construction site image 94 determined by the determination processing unit 56 to be an appropriate image when the appropriate image is shown in the guide image 88 of FIG. As can be seen by comparing the guide image 88 of FIG. 6 and the construction site image 94 of FIG. 9, the foundation is properly photographed in the construction site image 94, and the composition conforms to the proper composition shown in the guide image 88. have. On the other hand, when the proper image is shown in the guide image 88 of FIG. 6, the construction site image 96 determined by the determination processing unit 56 to be an inappropriate image is shown in FIG. As can be seen by comparing the guide image 88 of FIG. 6 and the construction site image 96 of FIG. 10, the foundation is properly photographed in the construction site image 96, but the composition does not conform to the proper composition shown in the guide image 88. have.

In the present embodiment, the determination processing unit 56 determines whether the construction site image acquired by the site terminal 14 is an appropriate image or an inappropriate image, but the determination is made by, for example, an administrator or the like. It may be done visually by a human being.

The construction site image determined to be an appropriate image is stored in the memory 42 as learning data 46. The construction site image determined to be an inappropriate image is not stored in the memory 42 as learning data 46, but is stored in the memory 42 as NG information 50. As described above, in the learning data 46, where the construction site image (appropriate image) and the information indicating the quality of the construction status indicated by the construction site image are associated, the quality of the construction status indicated by the construction site image is indicated. Information is attached by an administrator or the like. Further, in addition to the construction site image (appropriate image) and the information indicating the quality of the construction status indicated by the construction site image, the selection process information transmitted from the site terminal 14 is further associated. As described above, the learning data 46 may include not only proper images but also inappropriate images (that is, NG information 50).

The notification processing unit 58 notifies the field worker of the determination result by the determination processing unit 56. In the present embodiment, the notification processing unit 58 transmits information indicating a determination result of whether the construction site image is a proper image or an inappropriate image to the site terminal 14 which is the transmission source of the construction site image. When the determination processing unit 56 determines that the construction site image is an appropriate image, the notification processing by the notification processing unit 58 causes, for example, "the captured construction site image is appropriate as learning data" on the display 24 of the site terminal 14. Characters such as "This is an image" are displayed.

On the other hand, when the determination processing unit 56 determines that the construction site image is an inappropriate image, the notification processing unit 58 transmits the NG information 50 already stored in the memory 42 to the site terminal 14 before the determination processing. Then, the NG information 50 is displayed on the display 24 of the field terminal 14. That is, the notification processing unit 58 causes the display 24 to display NG information 50 different from the construction site image (which itself is an inappropriate image) acquired by the site terminal 14 this time. Specifically, since the NG information 50 is prepared for each process as described above, the notification processing unit 58 identifies and identifies the NG information 50 of the process indicated by the selection process information received from the site terminal 14. The NG information 50 is transmitted to the field terminal 14, and the NG information 50 is displayed on the display 24. The display screen of the NG information 50 may be the same as the screen shown in FIG. Further, a plurality of NG information 50s may be displayed on the display 24.

By displaying the above-mentioned NG information 50 on the display 24 of the site terminal 14, the site worker can determine what kind of image is an inappropriate image in addition to the construction site image that he / she has just taken. Can be grasped. As a result, when the site worker tries to acquire the construction site image of the process until the appropriate image is acquired, it is suppressed that the inappropriate image is repeatedly acquired.

The learning processing unit 60 trains the learning device 44 using the learning data 46. Specifically, the construction site image included in the learning data 46 is input to the learning device 44, and the prediction of the quality of the construction situation shown in the construction site image, which is the output of the learning device 44, and the learning data 46. The learner 44 is trained based on the difference between the included teacher data and the quality (actual result) of the construction situation shown in the construction site image. As described above, the learning data 46 includes the appropriate image determined to be appropriate as the learning data of the learning device 44 among the construction site images transmitted from the site terminal 14. Therefore, by learning the learning device 44 using the learning data 46, the learning device 44 can be learned with high efficiency.

In particular, in the present embodiment, the learning processing unit 60 trains the learning device 44 using the learning data 46 including the appropriate image having the composition conforming to the appropriate composition shown in the guide information 48. As a result, a plurality of construction site images having a substantially unified composition will be used as the learning data 46. As a result, the learner 44 can be further trained with high efficiency.

The learned learner 44 can output the quality of the construction situation shown in the input construction site image with high accuracy. In the present embodiment, as described above, the construction site image contains the construction site name, the construction process, the shooting date and time, the construction person name, and the construction person name in the form of an electronic construction signboard or metadata. The name of the contractor to which it belongs is given. Therefore, the learner 44 can determine the quality of the construction status by distinguishing each of these attributes. For example, the manager or the like can grasp the accuracy (politeness) of the construction of each construction staff based on the result of the learning device 44 distinguishing each construction staff and determining the quality of the construction status. In addition, the accuracy (politeness) of the construction work of each contractor can be grasped based on the result of the learning device 44 distinguishing each contractor and determining the quality of the construction status.

Further, it can be said that the process of the construction is completed when the learning device 44 determines that the construction condition shown in the construction site image is excellent. Therefore, based on the output of the learner 44, it is possible to create an actual process chart showing the completion date and time of each process.

Hereinafter, the flow of the learning data acquisition process in the construction site image acquisition system 10 will be described with reference to the flowchart shown in FIG.

In step S10, the field worker logs in to the field terminal 14, and the processor 30 of the field terminal 14 causes the display 24 to display the process selection screen (see FIG. 4). The field worker selects a selection process on the process selection screen.

In step S12, the processor 30 causes the display 24 to display a shooting screen (see FIG. 5). When the field worker operates the guide button 84 included in the shooting screen, the communication interface 20 transmits the guide button operation signal and the selection process information indicating the selection process selected by the field worker in step S10 to the server 16. ..

In step S14, the display control unit 54 of the server 16 identifies the guide information 48 corresponding to the process indicated by the selection process information from the memory 42 based on the selection process information received in step S12, and the guide information 48 Is transmitted to the field terminal 14.

In step S16, the processor 30 of the field terminal 14 displays the guide information 48 (for example, the guide image 88) received in step S14 on the display 24 (see FIG. 6). Here, the main points may be highlighted in the guide information 48 (see FIG. 7), and the guide information 48 indicates an appropriate composition of an appropriate image.

In step S18, the site worker operates the back button 90 to return to the shooting screen, and while referring to the guide information 48 displayed in step S16, shoots a construction object indicating the construction status at the construction site. The camera 28 acquires the construction site image. The communication interface 20 transmits the construction site image acquired by the camera 28 to the server 16.

In step S20, the determination processing unit 56 of the server 16 determines whether the construction site image received in step S18 is an appropriate image or an improper image. When the determination processing unit 56 determines that the construction site image is an inappropriate image, the process proceeds from step S22 to step S24.

In step S24, the notification processing unit 58 notifies the site terminal 14 that the construction site image received in step S18 is an inappropriate image, and also provides the NG information 50 stored in the memory 42 corresponding to the process. It is transmitted to the field terminal 14.

On the other hand, when the determination processing unit 56 determines that the construction site image received in step S18 is an appropriate image, the process proceeds from step S22 to step S26.

In step S26, the notification processing unit 58 notifies the site terminal 14 that the construction site image received in step S18 is an appropriate image. After that, in step S28, the processor 52 stores the construction site image as learning data 46 in the memory 42. At this time, the selection process information received in step S12 and the quality of the construction status shown in the construction site image are associated with the construction site image and stored as learning data 46.

In step S30, the processor 30 of the field terminal 14 determines the content of the notification received from the server 16. When the content of the notification indicates that the construction site image transmitted in step S18 is an inappropriate image, in step S32, the processor 30 causes the display 24 to display the NG information 50 received from the server 16. After that, based on the operation of the site worker, the process returns to step S18, the processor 30 displays the shooting screen on the display 24 again, and causes the field worker to shoot the construction site image again.

On the other hand, when the notification content indicates that the construction site image transmitted in step S18 is an appropriate image, the processor 30 displays the notification content on the display 34.

By the above processing, the acquisition process of one learning data is completed. By repeating the above process, the training data group is acquired.

Although the embodiments according to the present invention have been described above, the present invention is not limited to the above embodiments, and various modifications can be made without departing from the spirit of the present invention.

For example, in the present embodiment, the functions of the display control unit 54, the determination processing unit 56, and the notification processing unit 58 are exhibited by the processor 52 of the server 16, but these functions are exhibited by the processor 30 of the field terminal 14. You may do so. In that case, the guide information 48 and the NG information 50 may be stored in the memory 26 of the field terminal 14.

Further, in the present embodiment, learning data 46 for learning the learning device 44 for determining the quality of the construction status shown in the construction site image acquired for each process at the construction site where the construction having a plurality of processes is performed. Although the method of acquiring the above is described, the present invention is for learning the learning device 44 for determining the quality of the construction condition shown in the construction site image acquired at the construction site where the construction having only a single process is performed. It can also be suitably applied when acquiring the training data 46 of the above.

10 Construction site image acquisition system, 12 management terminal, 14 site terminal, 16 server, 18 communication line, 20,40 communication interface, 22 input interface, 24 display, 26,42 memory, 28 camera, 44 learner, 46 learning data , 48 guide information, 50 NG information, 52 processor, 54 display control unit, 56 judgment processing unit, 58 notification processing unit, 60 learning processing unit.

Claims (10)

Learning to judge the quality of the construction situation shown in the construction site image before the site worker takes the construction site image on the display part of the camera that captures the construction situation at the construction site and acquires the construction site image. A display control unit that displays guide information indicating an appropriate image, which is an appropriate construction site image, as learning data when learning the vessel.
A learning data storage unit that stores the construction site image acquired by the camera as learning data of the learning device.
A construction site image acquisition system characterized by being equipped with. The camera acquires a construction site image for each construction process.
The display control unit displays the guide information indicating the appropriate image when learning the learner for determining the quality of the construction status of the selected process, which is the selected process.
The learning data storage unit associates the selection process with the construction site image acquired by the camera and stores it as learning data of the learning device.
The construction site image acquisition system according to claim 1. In the guide information, the main points for obtaining the proper image are highlighted.
The construction site image acquisition system according to claim 1 or 2. The guide information indicates an appropriate composition which is the composition of the appropriate image.
The construction site image acquisition system according to any one of claims 1 to 3, wherein the construction site image acquisition system is characterized in that. The display control unit causes the display unit of the camera to display NG information indicating an inappropriate construction site image as learning data when learning the learning device.
The construction site image acquisition system according to any one of claims 1 to 4, wherein the construction site image acquisition system is characterized in that. Using the learning data, a learning processing unit that learns the learning device so as to output the quality of the construction situation indicated by the input construction site image, and the appropriate image having a composition conforming to the appropriate composition. A learning processing unit for learning the learner by using it as the learning data is further provided.
The construction site image acquisition system according to claim 4, wherein the construction site image acquisition system is characterized in that. A determination processing unit that determines whether or not the construction site image acquired by the camera is the proper image, and determines that the construction site image having a composition conforming to the proper composition is the proper image. Processing unit and
With more
The learning processing unit trains the learning device using the learning data including the construction site image determined by the determination processing unit to be the proper image.
The construction site image acquisition system according to claim 6, wherein the construction site image acquisition system is characterized in that. A notification processing unit that notifies the site worker of the determination result by the determination processing unit, and the construction site image acquired by the camera is an inappropriate image that is inappropriate as learning data when learning the learning device. When the determination processing unit determines that there is, NG information indicating the inappropriate image determined by the determination processing unit in the past is displayed on the display unit of the camera.
The construction site image acquisition system according to claim 7. Learning to judge the quality of the construction situation shown in the construction site image before the site worker takes the construction site image on the display part of the camera that captures the construction situation at the construction site and acquires the construction site image. A display control unit that displays guide information indicating an appropriate image, which is an appropriate construction site image, as learning data when learning the vessel.
A learning data storage unit that stores the construction site image acquired by the camera as learning data of the learning device.
A construction site image acquisition device characterized by being equipped with. Computer,
Learning to judge the quality of the construction situation shown in the construction site image before the site worker takes the construction site image on the display part of the camera that captures the construction situation at the construction site and acquires the construction site image. A display control unit that displays guide information indicating an appropriate image, which is an appropriate construction site image, as learning data when learning the vessel.
A learning data storage unit that stores the construction site image acquired by the camera as learning data of the learning device.
A construction site image acquisition program characterized by functioning as.

Images