Progress management system and progress management method

Abstract

To provide technical means for automatically grasping a "quantitative construction progress degree" in real time and with high accuracy even from images captured by a general-purpose monitor camera at a construction site of a building.SOLUTION: A progress management system 1 includes an imaging unit 10, a coordinate setting unit 21, a building component recognition unit 22, a BIM data storage unit 23, and a progress degree calculation unit 24. The building component recognition unit 22 includes a machine learning type image recognition device having a trained multi-layer neural network that has been trained by deep learning with an input of training data, which is image data of a finite number of types of building components that are set in advance as objects for progress management. The progress degree calculation unit 24 calculates a progress degree of a building work by comparing 3D real-time construction state information generated by the building component recognition unit with at-the-time-of-completion construction state information included in BIM data.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a progress management system for managing the progress of construction work of a building, and a progress management method. More specifically, the present invention relates to a progress management system and a progress management method capable of accurately grasping the progress of construction in real time from surveillance images taken by general-purpose surveillance cameras or the like at a construction site of a building.

Background Art

[0002] In the conventional three-dimensional model in the design of a building, it was created by first creating a two-dimensional design drawing and then starting a three-dimensional model based on the two-dimensional design drawing. On the other hand, "BIM (Building Information Modeling)" is a method of creating a design drawing in three dimensions from the beginning and giving the three-dimensional model the three-dimensional shape and position information (color, shape, size, and position in the three-dimensional space, etc.) and the attribute information of what kind of building member it is (whether it is a wall, a column, etc.) (hereinafter, these are collectively referred to as "BIM information"). Also, in "BIM", by making the above three-dimensional model a four-dimensional model with time-axis information added, "BIM information" at each construction stage can also be output. According to such "BIM", it can be used as a 3D-CG image in a design presentation, and it is also possible to output images of pipes inside the wall that are not visible during the construction stage and members to be installed in the next process (refer to the website of the Japan Construction Information Center, Inc. (https: / / www.jcitc.or.jp / bimcim / bim / )).

[0003] At construction sites, in order to manage the progress of the work against the pre-planned work schedule, work supervisors and others keep track of the progress of the work by recording the daily work status in a work log or inputting it into an information processing device such as a computer. As a design data management device for managing the progress of such work, a management device has been proposed that, for example, uses a scanner to perform a laser scan to acquire three-dimensional data of the work area, and identifies the difference between the three-dimensional model of the construction site at the time the laser scan was performed and the three-dimensional model based on the design data updated the previous day, thereby allowing for visual determination of the progress of the work (see Patent Document 1).

[0004] However, the design data management device disclosed in Patent Document 1 requires a procedure in which the scan range is set by identifying the parts where work was scheduled for that day using construction management data. Furthermore, while a laser scanner can grasp the three-dimensional shape of the object to be scanned, it cannot recognize other physical properties or attributes of the object to be scanned. Therefore, the scanning range setting and the installation of the laser scanner must be performed by a worker each time a scan is performed. Consequently, it is clear that the degree to which the workload of progress management workers is reduced is limited.

[0005] In response to this, a construction management system has been proposed that uses "BIM" to manage the progress of construction sites, comprising: a differential data calculation means that calculates differential data including the difference between the progress data for the day and the design data, and the difference between the progress data for the day and the previous day; a 3D model data creation means that creates 3D model data based on the calculated differential data; and a video display means that visualizes and displays the created data (see Patent Document 2). According to the construction management system disclosed in Patent Document 2, unlike the design data management device disclosed in Patent Document 1, there is no need to install a laser scanner or set the scan range, and it is said that "work content can be determined" from images that can be captured with a general surveillance camera or the like.

[0006] Patent Document 2 states that "by determining the content of the work, it is possible to calculate 'progress data' such as the content of the work, the work status, and the number of workers," and also states that "the content of the work can be determined using deep learning technology." However, it does not disclose any specific means of automatically and accurately determining the progress of the construction against a predetermined work schedule (for example, an objective numerical value indicating the construction completion rate, which is what percentage of the required amount of reinforcing steel to be installed at the stage of the planned completion date of the foundation work has been correctly installed at this point. Hereinafter, such a numerical value will be referred to as the "quantitative construction progress rate") in real time from images that can be captured by general surveillance cameras or the like.

[0007] In construction site progress management, various methods have been attempted, including the method disclosed in Patent Document 2, to utilize images captured by general-purpose surveillance cameras for progress management. However, to date, none of these attempts have reached the point of automatically and accurately grasping the aforementioned "quantitative progress of construction" in real time. There has been a need for the development of a new progress management method that can be constructed using general-purpose surveillance cameras and that can perform automatic progress management with high accuracy, thereby significantly reducing the workload of management personnel. [Prior art documents] [Patent Documents]

[0008] [Patent Document 1] Japanese Patent Publication No. 2017-204222 [Patent Document 2] Japanese Patent Publication No. 2019-148946 [Overview of the project] [Problems that the invention aims to solve]

[0009] The present invention aims to provide a technical means that enables the automatic, real-time, and highly accurate determination of "quantitative construction progress" at a construction site, even from images captured by a general-purpose surveillance camera. [Means for solving the problem]

[0010] The inventors have conceived of a system or method using such a system that includes a coordinate setting unit that sets three-dimensional coordinates within a monitoring image to create a three-dimensional monitoring image, and a building component recognition unit equipped with a trained multi-layer neural network that has been trained by deep learning using image data of a finite number of building components predetermined as targets for progress management as training data, and have come to the conclusion that the above problems can be solved by utilizing BIM data, which has been gradually becoming more widespread in recent years, and have completed the present invention. Specifically, the present invention solves the above problems by the following means.

[0011] (1) A progress management system for a construction site, comprising: a shooting unit; a coordinate setting unit that sets 3D coordinates within the monitoring image generated by the shooting unit to create a 3D monitoring image; a building component recognition unit that recognizes the 3D construction position and construction volume of individual building components within the 3D monitoring image to generate 3D real-time construction status information; a BIM data storage unit that stores BIM data for progress management; and a progress calculation unit that calculates the progress of construction work, wherein the building component recognition unit comprises a machine learning type image recognition device having a multilayer neural network, the multilayer neural network is a trained multilayer neural network that has been trained by deep learning by inputting image data of a finite number of building components set in advance as targets for progress management as training data, the BIM data includes completed construction status information which is information indicating the 3D construction position and construction volume of individual building components constituting the building at the completion stage of the building, and the progress calculation unit calculates the progress of construction work by comparing the 3D real-time construction status information generated by the building component recognition unit with the completed construction status information included in the BIM data.

[0012] In the progress management system described in (1), 3D data obtained by converting surveillance images captured by a general-purpose surveillance camera into 3D data using a coordinate setting unit is analyzed by an image recognition means equipped with a pre-trained multilayer neural network that has been improved to an extremely high level of image recognition capability for a finite number of building components predetermined as targets for progress management. The results are then compared with BIM data. This makes it possible to automatically grasp the "quantitative progress of construction" at a construction site in real time and with high accuracy, which was previously difficult to calculate automatically from surveillance images captured by general-purpose surveillance cameras.

[0013] (2) The progress management system according to (1), wherein the three-dimensional monitoring image includes date and time information, the building component recognition unit generates 3D real-time construction status information which further includes the date and time information, the BIM data includes 4D construction status information which is information indicating the management standard values ​​for the three-dimensional construction position and construction amount of each building component constituting the building at any point in the flow of work time from the start of construction work to the end of construction work, and the progress calculation unit calculates the progress by comparing the 3D real-time construction status information with the 4D construction status information which corresponds to the date and time information related to the date and time information of the date and time information of the 3D real-time construction status information.

[0014] According to the progress management system in (2), the "quantitative progress rate of construction" (the percentage of progress relative to the initial construction plan at each point in time, whether the construction is behind schedule or progressing faster than expected) can be automatically grasped in real time and with high accuracy throughout the entire period of construction work.

[0015] (3) The progress management system according to (1) or (2), further comprising an abnormality notification means for notifying of an abnormality in construction if the three-dimensional construction position of the building member in the 3D real-time construction status information does not match the three-dimensional construction position of the building member in the completed construction status information.

[0016] According to the progress management system in (3), construction abnormalities caused by incorrect placement of building materials can be detected early, and by automatically excluding building materials with incorrect placement from the progress calculation process, the accuracy of grasping the "quantitative progress of construction" can be maintained more reliably.

[0017] (4) The progress management system described in (1) or (2), wherein the number of types of building materials that are set in advance as targets for progress management is 20 or less.

[0018] In the progress management system described in (4), the focus is on the fact that the types of building materials that need to be recognized at construction sites are limited, and the target of image analysis is limited to the minimum necessary types of building materials, which are 20 or fewer. This makes it easier to train the multilayer neural network and improve the image recognition capability to an extremely high level more easily than before. As a result, in the progress management system described in (1) or (2), when the building material recognition unit "recognizes the 3D construction position and construction amount of each building material in the monitored image data and generates 3D real-time construction status information", false detection of building materials is reduced, and the "quantitative construction progress" can be automatically grasped with higher accuracy. In addition, since the training cost of the multilayer neural network can be reduced, a high-performance system can be provided at a lower cost.

[0019] (5) The progress management system according to (1) or (2), wherein the imaging unit consists of a drone or a patrol robot equipped with an autonomous driving function and a camera mounted on it.

[0020] According to the progress management system of (5), even in high-rise buildings where it is difficult to photograph with fixed cameras, etc., or in buildings with complex internal structures where it is difficult to efficiently photograph the necessary parts with fixed cameras, the progress management system of the present invention can be implemented and the effects of each of the above inventions can be enjoyed.

[0021] (6) A progress management device for a construction site, comprising: a coordinate setting unit that sets three-dimensional coordinates in a monitoring image of the construction site to create a three-dimensional monitoring image; a building component recognition unit that recognizes the three-dimensional construction position and construction amount of individual building components in the three-dimensional monitoring image and generates 3D real-time construction status information; a BIM data storage unit that stores BIM data for progress management; and a progress calculation unit that calculates the progress of construction work, wherein the building component recognition unit comprises a machine learning type image recognition device having a multilayer neural network, the multilayer neural network is a trained multilayer neural network that has been trained by deep learning by inputting image data of a finite number of building components set in advance as targets for progress management as training data, the BIM data includes completed construction status information which is information indicating the three-dimensional construction position and construction amount of individual building components constituting the building at the completion stage of the building, and the progress calculation unit calculates the progress of construction work by comparing the 3D real-time construction status information generated by the building component recognition unit with the completed construction status information included in the BIM data.

[0022] According to the progress management device of (6), by communicating with existing surveillance cameras installed at the construction site, it can perform functions equivalent to the progress management system of the present invention, and can automatically grasp the "quantitative progress of construction" in real time and with high accuracy from images captured by general-purpose surveillance cameras.

[0023] (7) A progress management method for a construction site using BIM data, wherein the BIM data includes completion-time construction status information indicating the three-dimensional construction positions and construction quantities of individual building components constituting the building at the completion stage of the building, comprising: a monitoring shooting step in which a shooting unit shoots the building to be managed for progress to obtain a monitoring image; a coordinate setting step in which a coordinate setting unit generates a three-dimensional monitoring image related to the building from the monitoring image; a construction confirmation step in which a building component recognition unit recognizes the three-dimensional construction positions and construction quantities of individual building components in the three-dimensional monitoring image to generate 3D real-time construction status information; and a progress confirmation step in which a progress calculation unit calculates the progress of the construction work by comparing the 3D real-time construction status information with the completion-time construction status information. The building component recognition unit is equipped with a machine learning-type image recognition device having a multi-layer neural network, and the multi-layer neural network is a learned multi-layer neural network that has been learned by deep learning by inputting image data of a finite number of types of building components preset as objects of progress management as teacher data.

[0024] In the progress management method of (7), 3D data obtained by converting a monitoring image taken by a general-purpose monitoring camera into three-dimensional data by a coordinate setting unit is analyzed by image recognition means equipped with a learned multi-layer neural network that has extremely improved the image recognition ability for a finite number of types of building components preset as objects of progress management, and the result is compared with the BIM data. As a result, it is possible to automatically grasp, in real time and with high accuracy, the "quantitative construction progress" at the construction site of a building, which has conventionally been difficult to calculate automatically from a monitoring image taken by a general-purpose monitoring camera.

[0025] (8) The BIM data includes 4D construction status information indicating the management reference values of the 3D construction positions and construction quantities of the individual building members constituting the building at any point in time during the flow of working hours from the start to the end of the construction work. In the monitoring and photographing step, a 3D monitoring image including photographing date and time information is generated. In the construction confirmation step, 3D real-time construction status information further including the photographing date and time information is generated. In the progress confirmation step, the progress is calculated by comparing the 3D real-time construction status information with the 4D construction status information corresponding to the date and time related to the photographing date and time information included in the 3D real-time construction status information. The progress management method according to (7).

[0026] According to the progress management method of (8), during the entire period when the construction work is carried out, for each time point, the "quantitative construction progress", which is the progress ratio (delay in construction at each time point, or progress faster than expected) with respect to the initial construction plan at that time point, can be automatically grasped in real time and with high accuracy.

[0027] (9) When the 3D construction position of the building member in the 3D real-time construction status information generated in the construction confirmation step does not match the 3D construction position of the building member in the completed construction status information, an abnormal notification step for notifying construction abnormalities is performed without executing the progress confirmation step. The progress management method according to (7) or (8).

[0028] According to the progress management method of (9), construction abnormalities due to errors in the installation positions of building members can be quickly detected, and by automatically excluding such building members with incorrect installation positions from the progress calculation process, the accuracy of grasping the "quantitative construction progress" can be more reliably maintained.

[0029] (10) The number of types of the building members preset as the object of progress management is 20 or less. The progress management method according to (7) or (8).

[0030] In the progress management method of (10), the focus is on the fact that the types of building materials that need to be recognized at construction sites are limited, and the target of image analysis is limited to the minimum necessary types of building materials, which are 20 or fewer. This makes it easier to train the multilayer neural network and makes it easier than before to improve the image recognition capability to an extremely high level. As a result, in the progress management system described in (7) or (8), when the building material recognition unit "recognizes the 3D construction position and construction amount of each building material in the monitored image data and generates 3D real-time construction status information", false detection of building materials is reduced, and the image analysis is reduced in real time and quantitatively, and the system is automatically grasped with higher accuracy. In addition, since the training cost of the multilayer neural network can be reduced, a high-performance system can be provided at a lower cost.

[0031] (11) A program for a progress management system that causes the monitoring and photography step, the coordinate setting step, the construction confirmation step, and the progress confirmation step to be executed in the progress management method described in (7) or (8).

[0032] According to the program in (11), it is possible to automatically grasp the "quantitative progress of construction" at a construction site in real time and with high accuracy, which was previously difficult to calculate automatically from surveillance images taken with general-purpose surveillance cameras. [Effects of the Invention]

[0033] According to the present invention, it is possible to provide a technical means that enables the "quantitative progress of construction" to be automatically grasped in real time and with high accuracy, even from images captured by a general-purpose surveillance camera at a construction site. [Brief explanation of the drawing]

[0034] [Figure 1] This is a block diagram showing the configuration of the progress management system of the present invention. [Figure 2]This figure schematically shows an example of an embodiment of the progress management system of the present invention. [Figure 3] This is a drawing illustrating a specific example of a building component recognized by the building component recognition unit in the progress management system of the present invention. [Figure 4] This flowchart shows the flow of the progress management method of the present invention, which can be performed using the progress management system of the present invention. [Modes for carrying out the invention]

[0035] <Progress Management System> The best mode for carrying out the present invention will be described below. Figure 1 is a block diagram showing the configuration of a progress management system 1, which is an example of an embodiment of the progress management system of the present invention. By having the configurations shown in Figure 1, the progress management system 1 can automatically grasp the "quantitative progress of construction" in real time and with high accuracy from images captured by a general-purpose surveillance camera at a construction site.

[0036] [Overall structure] The progress management system 1 consists of an imaging unit 10 and a calculation processing unit 20. The calculation processing unit 20 includes at least a coordinate setting unit 21, a building component recognition unit 22, a BIM data storage unit 23, and a progress calculation unit 24. The progress management system 1, having the above configuration, can be integrated into a single progress management system (or progress management device). Alternatively, the progress management system 1 can have other configurations in which only some of the functions of the calculation processing unit 20 are mounted on the imaging unit 10. One example of such other configurations is one in which the coordinate setting unit 21 is mounted on the camera that constitutes the imaging unit 10.

[0037] Here, the progress management system 1 is also an information processing system in which each of the above-mentioned parts, namely the imaging unit 10 and the calculation processing unit 20, are connected to each other in a manner that enables information communication (see Figure 2). The connections between each part can be made by wired connection using a dedicated communication cable, or by wired LAN connection. Furthermore, the progress management system 1 can also be configured using various wireless communication methods, such as wireless LAN, short-range wireless communication, or mobile phone lines, in addition to wired connections.

[0038] As described above, in the progress management system 1, which is configured by connecting each part to enable information communication, at least the imaging unit 10 is installed within the construction site of the building whose progress is to be managed. However, it is not necessarily required that the calculation processing unit 20 be installed within the site. The calculation processing unit 20 can be located in another facility at any location physically distant from the site, or on the internet (so-called cloud). Such a distributed information system embodiment is naturally included within the technical scope of the progress management system of the present invention.

[0039] [Photography Department] The imaging unit 10 is comprised of a surveillance camera that performs a procedure (monitoring imaging step S10) to capture a 3D space constituting the area subject to progress management at a construction site as a 2D image. Any existing digital camera can be used without particular restriction as the surveillance camera, as long as it has the function of processing the captured surveillance image 11 into digital image data so that it can be processed by the arithmetic processing unit 20, and outputting the image data to the arithmetic processing unit 20. However, it is preferable that the camera be capable of recording the date and time of capture information in the image data for matching with the time axis information of the BIM data. Regarding other imaging functions, there are no particular limitations; a general-purpose monocular camera that captures the 3D space of the monitored area as a 2D visible light image can be used. Therefore, the effects of the present invention can be enjoyed without introducing expensive 3D cameras or laser scanners such as infrared cameras.

[0040] Figure 2 is a schematic diagram showing an example of how the surveillance camera constituting the imaging unit 10 is mounted. The surveillance camera may be a fixed camera installed in a fixed position capable of capturing a predetermined three-dimensional space within the construction site set as the area to be monitored, but it is more preferable to use a mobile camera with pan, tilt, and zoom functions that can change the mounting position and shooting area as needed. Figure 2 shows an example in which two cameras are installed to capture information of the three-dimensional space of the area to be monitored, but the mounting location can be changed as appropriate depending on the conditions of the three-dimensional space of the area to be monitored. Note that all drawings shown below, including Figure 2 (except for the flowchart in Figure 5), are schematic diagrams, and the size, shape, etc. of each part have been modified as needed to facilitate understanding of the configuration and operation of the present invention.

[0041] Furthermore, the camera unit 10 can also use an existing surveillance camera installed at the facility as the surveillance camera constituting the camera unit 10 of the progress management system 1, as long as the camera is installed in a position that allows it to continuously photograph the building subject to the progress management control described above. In this case, by installing a "progress management device" consisting of the calculation processing unit 20, which will be explained in detail below, at the construction site and connecting it to an existing surveillance camera at the construction site so as to enable communication, it is possible to achieve the same functionality as the progress management system 1 and automatically grasp the "quantitative progress of construction" in real time and with high accuracy from images captured by a general-purpose surveillance camera.

[0042] Furthermore, the camera unit 10 can be configured not with the general surveillance cameras described above, but with cameras mounted on drones or patrol robots equipped with autonomous driving capabilities, or with portable information processing terminals (smartphones, tablet devices, etc.) with communication and camera functions carried by workers.

[0043] [Processing Unit] The calculation processing unit 20 takes the image data transmitted from the imaging unit 10 as input values ​​and performs the calculations necessary to automatically perform progress management within the monitored area. The calculation processing unit 20 consists of at least a coordinate setting unit 21, a building component recognition unit 22, a BIM data storage unit 23, and a progress calculation unit 24. The calculation processing unit 20 can also be configured as a dedicated information processing device specifically for controlling the operation of each part of the progress management system 1, or it can be configured using a general-purpose information processing device (including a personal computer), for example.

[0044] In any of the above configurations, the arithmetic processing unit 20 is equipped with hardware such as a CPU, memory, and communication unit. The arithmetic processing unit 20, having such a configuration, can execute the various operations and progress management methods described below by executing the progress management system program, which is a computer program according to the present invention.

[0045] (Coordinate setting section) The coordinate setting unit 21 performs a procedure (coordinate setting step S20) to set three-dimensional coordinates in the monitoring image 11 of the monitored area captured by the shooting unit 10, thereby creating a "three-dimensional monitoring image". The three-dimensional coordinates set here are a set of three-dimensional position vectors that correspond the position of each point in the plane of the background image of the two-dimensional monitoring image 11 to its position in the three-dimensional space that constitutes the actual monitored area. In this specification, an image or image data in which such three-dimensional position information is embedded in the two-dimensional monitoring image 11 in a state that can be mechanically read by an information processing device is referred to as a "three-dimensional monitoring image".

[0046] Furthermore, the 3D coordinates that the coordinate setting unit 21 sets in the monitoring image 11 are the same coordinates as those used to identify the installation position of each building component in the BIM data stored in the BIM data storage unit 23, or coordinates that allow for a one-to-one correspondence between the coordinates and the individual 3D position vectors. Details of the operation of the coordinate setting procedure (coordinate setting step S20) will be described later.

[0047] (Building component recognition unit) The building component recognition unit 22 recognizes the 3D construction position and construction amount of individual building components within the "3D monitoring image" whose 3D coordinates have been set by the coordinate setting unit 21, through image recognition processing, and performs a procedure (construction confirmation step S30) to generate "3D real-time construction status information" based on this recognition. To perform this procedure, the building component recognition unit 22 is equipped with a machine learning type image recognition device having a multilayer neural network as an image recognition device.

[0048] The extraction and identification of the type of individual building components from the "3D monitoring images" for generating "3D real-time construction status information" in the building component recognition unit 22, i.e., image recognition processing, can be performed using any of the various conventionally known methods, or a combination thereof. The identification of the type of object to be processed, i.e., the type of building component, is performed by a machine learning type image analysis device having the aforementioned neural network (a so-called "image recognition device using deep learning technology").

[0049] The image data relating to the building components whose type has been identified by the image recognition processing described above (for example, the monitoring image 11 in Figure 3) has three-dimensional position vector information set by the coordinate setting unit 21. Therefore, it is possible to detect the location of each identified building component (for example, the formwork 31 and reinforcing bars 32 that make up building 3 in Figure 2) within the three-dimensional space that constitutes the actual monitoring target area, along with the attributes of each building component. In this way, the building component recognition unit 22 can automatically and quantitatively detect where, what type, and to what extent (how many, how many, etc.) building components are installed, and generate "3D real-time construction status information," which is data relating to the three-dimensional construction position and construction amount of the building components.

[0050] Furthermore, since the BIM data stores design information regarding the location and quantity of each building component (for example, formwork 31 and reinforcing bars 32), by comparing the location information of the "current location" with the "desired location," it is possible to prevent misidentification, such as including reinforcing bars that are not yet installed and are located far from the construction site in the calculation of the installed reinforcing bars.

[0051] As described above, the image recognition device provided in the building component recognition unit 22 is a machine learning type image recognition device having a multilayer neural network. The multilayer neural network in the image recognition device according to the present invention is a trained multilayer neural network that has been trained by deep learning using image data of a finite number of building components predetermined as targets for progress management as training data.

[0052] The finite number of building components that are pre-set to calculate the "quantitative construction progress" are, as an example, shown in Table 1 below. For example, by pre-setting the seven types of building components shown in Table 1 as targets for progress management, it is possible to generate "3D real-time construction status information," and based on that, the "quantitative construction progress" can be calculated. Depending on the type and scale of the building being managed, the basic building components that have a certain degree of influence on the progress of the construction are often all standardized, and the number of types of building components that need to be set as targets to be recognized in advance for the purpose of progress management is often around the seven types shown below, or at most 20 types or less. By configuring the system assuming that the number of types of building components to be pre-set as targets for progress management is 20 or less, a general-purpose progress management system that can be used at most construction sites can be constructed.

[0053] [Table 1]

[0054] In order to sufficiently improve the recognition accuracy of the machine learning-type image recognition device having the multilayer neural network described above, it is necessary to train the multilayer neural network using deep learning by inputting a large number of images of the object to be recognized as training data. Even when recognizing a single type of object, if there are many variations in shape and size of the actual objects that make up the type, the recognition accuracy will not improve sufficiently unless a very large amount of training data covering all of them is trained. However, building materials used at construction sites are mostly relatively standardized, and as mentioned above, the types can be narrowed down to about 20 types or less. Therefore, in the progress management system 1, a preferred embodiment is one in which the target of image analysis is limited to the minimum necessary types of building materials, which are 20 or fewer. This makes it easier to train the multilayer neural network and improve the image recognition capability to an extremely high level more easily than before. In addition, this reduces the training cost of the multilayer neural network, so a high-performance system can be provided at a lower cost.

[0055] (BIM data storage unit) The BIM data storage unit 23 is comprised of various information processing devices that store BIM data for progress management and can extract necessary information from the stored data as needed.

[0056] Traditionally, building design has been carried out by first designing on 2D blueprints and then creating a 3D model from them. In contrast, BIM (Building Information Modeling) is a method of directly designing a 3D model of a building without using 2D blueprints, which includes 3D "shape information" and "attribute information" such as whether the shape is a "wall" or a "window," and what kind of structure the wall has. Furthermore, BIM can also be a 4D model by adding time axis information to the above 3D model. The "BIM data" used in this invention is design data that includes at least information indicating the correct construction position and correct construction amount of each building component that constitutes the building in the actual 3D space at the completion stage of the building. Furthermore, the "BIM data" used in the present invention is preferably a four-dimensional model with added time axis information, and more specifically, data that includes 4D construction status information, which is information indicating the three-dimensional construction position and construction volume management standard values ​​of individual building members constituting a building at any point in the flow of work time from the start of construction work to the end of construction work.

[0057] If the "BIM data" is a 4D model with the time axis information described above added, the building component recognition unit 22 generates 3D real-time construction status information that further includes the date and time of shooting information. This allows for not only understanding the progress based on the completion stage, but also automatically and accurately in real time the "quantitative construction progress" for each workday, such as "At A time on A month A, the progress of process A against the initial plan has only reached 70%." Details of the method for calculating the progress will be described later.

[0058] (Progress calculation section) The progress calculation unit 24 performs a procedure (progress confirmation step S40) to calculate the progress of the construction work by comparing the 3D real-time construction status information generated by the building component recognition unit 22 with the completed construction status information stored in the BIM data storage unit 23. The progress calculated here is the "quantitative construction progress." The "quantitative construction progress" refers to the progress of the construction work relative to the pre-planned work schedule (for example, an objective numerical value indicating the construction completion rate, which is what percentage of the required amount of reinforcing steel to be installed at the stage of the planned completion date of the foundation work has been correctly installed at this point). Details of the operation of the procedure (progress confirmation step S40) for calculating the progress of the construction work will be described later.

[0059] (Abnormality notification means) Furthermore, it is preferable that the calculation processing unit 20 includes an anomaly notification means (not shown) that notifies of a construction anomaly if the 3D construction position of a building component in the 3D real-time construction status information generated by the building component recognition unit 22 does not match the 3D construction position of the building component in the completed construction status information. This makes it possible to quickly detect construction anomalies caused by incorrect placement of building components, and by automatically excluding building components with incorrect placement from the progress calculation process, the accuracy of grasping the "quantitative construction progress" can be maintained more reliably.

[0060] [Progress Management Method (Operation of the Progress Management System)] The progress management method for construction sites of the present invention is a progress management method that effectively utilizes BIM data. The BIM data used in the progress management method of the present invention is data that includes completed construction status information, which is information indicating the three-dimensional construction position and construction amount of individual building members constituting the building at the completion stage of the building. Furthermore, as described above, this BIM data is preferably a four-dimensional model (4D construction status information) with time axis information added.

[0061] Figure 5 is a flowchart showing the flow of the progress management method of the present invention, which can be suitably executed using the progress management system 1. The progress management method of the present invention is a management method that can automatically grasp the "quantitative progress of construction" in real time and with high accuracy, even from images taken with a general-purpose surveillance camera at a construction site of a building. This progress management method is a process in which the following steps are performed sequentially: surveillance shooting step S10, coordinate setting step S20, construction confirmation step S30, and progress confirmation step S40. Furthermore, as shown in Figure 5, if the 3D construction position of the building member in the 3D real-time construction status information generated in the construction confirmation step S30 does not match the 3D construction position of the building member in the completed construction status information, it is more preferable that the progress confirmation step S40 is not executed and an abnormality notification step S50 is performed to notify of the construction abnormality.

[0062] (Surveillance and filming step) In the monitoring and photography step S10, the photography unit 10 performs monitoring photography to obtain a monitoring image 11 by photographing the building that is subject to progress management. Here, monitoring photography involves continuously taking still images at predetermined intervals and performing the monitoring operation described later using the sequence of captured images. However, by making the shooting interval very short, it can be considered that the monitoring operation is essentially being performed as video recording.

[0063] The surveillance photography in step S10 can be carried out by arranging the necessary number of various surveillance cameras as illustrated in Figure 2, but as mentioned above, it can also be carried out using a portable information processing terminal (such as a smartphone or tablet) with communication and photography functions carried by a worker, or by using a camera mounted on a drone or a patrol robot equipped with autonomous driving capabilities.

[0064] (Coordinate setting step) In the coordinate setting step S20, the coordinate setting unit 21 performs a coordinate setting process to generate a 3D monitoring image of the building from the monitoring image 11 captured by the imaging unit 10. This "coordinate setting process" is a process of setting 3D coordinates, which are a set of 3D position vectors that correspond the position of each point in the plane of the background image to the position in the actual monitoring area space, within the background image of the 2D monitoring image in the 3D space constituting the monitoring target area. The "background image" mentioned above is a 2D image taken by a monitoring camera or the like of the monitoring target area which includes the building whose progress is to be managed, and is an image composed of the bottom surface of the 3D space constituting the monitoring target area and permanent structures.

[0065] The "coordinate setting process," which sets a set of three-dimensional position vectors, or three-dimensional coordinates, within a two-dimensional surveillance image 11, can be specifically performed by, for example, the method disclosed in "Patent No. 6581280," or the method disclosed in "Patent No. 5899506," both inventions of the inventors of the present invention. According to these methods, the set coordinates are associated with the actual position (actual dimensions, actual distance) within the three-dimensional space of the surveillance target area. This also means that each point on the coordinates includes distance information from the imaging unit 10. As a result, by determining the location of the surveillance target within a predetermined area, it becomes possible to obtain three-dimensional information relating to the size and three-dimensional shape of the surveillance target.

[0066] Furthermore, when the camera is mounted on a mobile device such as a drone, the coordinate setting can be done, for example, by using the technology disclosed in Japanese Patent Publication No. 2018-151696, which "arranges multiple CV images with added CV values ​​in a virtual space according to their three-dimensional coordinates."

[0067] (Construction verification step) In the construction confirmation step S30, the building component recognition unit 22 performs construction confirmation processing to generate 3D real-time construction status information by recognizing the 3D construction position and construction amount of each building component in the 3D monitoring image that was verified in the coordinate setting step S20.

[0068] The recognition of the "3D construction position and construction amount of individual building components" can be performed using any of the various conventionally known methods, or a combination thereof. The identification of the type of individual building component is performed by a machine learning-type image recognition device having a multilayer neural network. The multilayer neural network of the image recognition device equipped in the building component recognition unit 22 is a trained multilayer neural network that has been trained by deep learning using image data of a finite number of building components (for example, the 7 types of building components shown in Table 1) that are set in advance as targets for progress management as input as training data. Image recognition technology using deep learning is publicly available, for example, at the following. "Deep Learning and Image Recognition, Operations Research" (http: / / www.orsj.o.jp / archive2 / or60-4 / or60_4_198.pdf)

[0069] Furthermore, if the 3D construction position of a building component in the 3D real-time construction status information generated in the construction confirmation step S30 does not match the 3D construction position of the building component in the 4D construction status information, it is preferable to configure the process flow so that the next step, the progress confirmation step S40, is not executed, and instead the abnormality notification step S50, which notifies of the construction abnormality, is performed.

[0070] (Progress check step) In the progress confirmation step S40, the progress calculation unit 24 calculates the progress of the construction work for each building component by comparing the 3D real-time construction status information generated in the construction confirmation step S30 with the completed construction status information in the BIM data stored in the BIM data storage unit 23.

[0071] The comparison between 3D real-time construction status information and completed construction status information is performed, for example, in the manner shown in Table 2. In this example, first, for the foundation section, the amount of reinforcing bars that were recognized in construction confirmation step S30 and confirmed to be installed in the correct position relative to the BIM data is 75, and this is generated as "3D real-time construction status information". Then, by comparing this information with the amount of steel installed at completion, which is 100 bars, as shown in the "completed construction status information" for the building component (steel frame), the "quantitative construction progress" of the steel frame installation work in the foundation section is calculated to be 75%.

[0072] [Table 2]

[0073] If BIM data is defined as data that includes 4D construction status information, which is information indicating the 3D construction position and construction volume management standard values ​​of individual building components constituting the building at any point in time within the flow of work time from the start to the end of construction work, then by similarly comparing the 3D real-time construction status information with the 4D construction status information corresponding to the date and time of the date and time information of the 3D real-time construction status information, the progress rate can be calculated, and the "quantitative construction progress rate" as the progress rate against the initial construction plan at each point in time (construction delays or faster-than-expected progress at each point in time) can be automatically grasped in real time and with high accuracy for the entire period during which construction work is carried out.

[0074] In the above example, "steel frame" was used as the building component for which the construction status is recognized. However, the same method can be used to automatically and accurately grasp the "quantitative construction progress" in real time for various other building components, including those shown in Table 1. Furthermore, by integrating the "quantitative construction progress" for each of these building components as needed, the overall "quantitative construction progress" for each construction day can be freely calculated.

[0075] (Anomaly notification step) In the abnormality notification step S50, if the installation position of the recognized building component is incorrect, an abnormality in construction is notified. If this notification is issued, the construction status is corrected, and the monitoring and photography step S10, coordinate setting step S20, construction confirmation step S30, and progress confirmation step S40 are repeated for the corrected construction location. [Explanation of Symbols]

[0076] 1. Progress Management System 10. Photography Department (Camera) 11 Surveillance Images 20 Arithmetic Processing Unit 21 Coordinate setting section 22 Building component recognition unit 23 BIM Data Storage Unit 24 Progress Calculation Unit 3 Buildings 31 Formwork 32 Reinforcement bars S10 Surveillance and filming step S20 Coordinate setting step S30 Construction Verification Steps S40 Progress Confirmation Step S50 Anomaly Notification Step

Claims

1. A progress management system for construction sites, The photography department, A coordinate setting unit sets three-dimensional coordinates within the monitoring image generated by the aforementioned imaging unit to create a three-dimensional monitoring image, A building component recognition unit recognizes the three-dimensional construction position and construction amount of individual building components in the aforementioned three-dimensional monitoring image and generates 3D real-time construction status information. A BIM data storage unit that stores BIM data for progress management, A progress calculation unit for calculating the progress of construction work, The building component recognition unit comprises a machine learning type image recognition device having a multilayer neural network, and the multilayer neural network is a trained multilayer neural network that has been trained by deep learning by inputting image data of a finite number of building components predetermined as targets for progress management as training data. The BIM data includes information indicating the three-dimensional construction location and construction volume of each building component constituting the building at the completion stage of the building, The progress calculation unit calculates the progress of the construction work by comparing the 3D real-time construction status information generated by the building component recognition unit with the completed construction status information included in the BIM data. Progress management system.

2. The aforementioned 3D surveillance image includes information on the date and time of capture. The building component recognition unit generates 3D real-time construction status information which further includes the shooting date and time information. The BIM data includes 4D construction status information, which is information indicating the 3D construction position and construction volume management standard values ​​of individual building components constituting the building at any point in time within the flow of work time from the start of construction work to the end of construction work. The progress calculation unit calculates the progress by comparing the 3D real-time construction status information with the 4D construction status information corresponding to the date and time related to the shooting date and time information contained in the 3D real-time construction status information. The progress management system according to claim 1.

3. If the three-dimensional construction position of the building component in the 3D real-time construction status information does not match the three-dimensional construction position of the building component in the completed construction status information, the system is equipped with an abnormality notification means for notifying of an abnormality in construction. The progress management system according to claim 1 or 2.

4. The number of types of building materials to be set in advance as targets for progress management is 20 or less. The progress management system according to claim 1 or 2.

5. The aforementioned imaging unit consists of a drone or a patrol robot equipped with an autonomous driving function, with a camera mounted on it. The progress management system according to claim 1 or 2.

6. A progress management device for construction sites, A coordinate setting unit that sets 3D coordinates within a construction site monitoring image to create a 3D monitoring image, A building component recognition unit recognizes the three-dimensional construction position and construction amount of individual building components in the aforementioned three-dimensional monitoring image and generates 3D real-time construction status information. A BIM data storage unit that stores BIM data for progress management, A progress calculation unit for calculating the progress of construction work, The building component recognition unit comprises a machine learning type image recognition device having a multilayer neural network, and the multilayer neural network is a trained multilayer neural network that has been trained by deep learning by inputting image data of a finite number of building components predetermined as targets for progress management as training data. The BIM data includes information indicating the three-dimensional construction location and construction volume of each building component constituting the building at the completion stage of the building, The progress calculation unit calculates the progress of the construction work by comparing the 3D real-time construction status information generated by the building component recognition unit with the completed construction status information included in the BIM data. Progress management device.

7. A construction site progress management method using BIM data, The BIM data includes information indicating the three-dimensional construction location and construction volume of individual building components constituting the building at the completion stage of the building, The camera unit performs a monitoring and photography step to obtain monitoring images by photographing the aforementioned building, which is subject to progress management, The coordinate setting unit performs a coordinate setting step of generating a three-dimensional monitoring image relating to the building from the monitoring image, A construction confirmation step involves a building component recognition unit recognizing the 3D construction position and construction amount of each building component in the 3D monitoring image and generating 3D real-time construction status information. The progress calculation unit includes a progress confirmation step in which it calculates the progress of the construction work by comparing the 3D real-time construction status information with the completed construction status information. The aforementioned building component recognition unit comprises a machine learning type image recognition device having a multilayer neural network, and the multilayer neural network is a trained multilayer neural network that has been trained by deep learning using image data of a finite number of building components predetermined as targets for progress management as training data. Progress management methods.

8. The BIM data includes 4D construction status information, which is information indicating the 3D construction position and construction volume management standard values ​​of individual building components constituting the building at any point in time within the flow of work time from the start of construction work to the end of construction work. In the aforementioned surveillance and photography step, a three-dimensional surveillance image containing the date and time of photography information is generated. In the aforementioned construction confirmation step, the 3D real-time construction status information is generated, which further includes the shooting date and time information. In the progress confirmation step, the progress is calculated by comparing the 3D real-time construction status information with the 4D construction status information corresponding to the date and time of the shooting date and time information contained in the 3D real-time construction status information. The progress management method according to claim 7.

9. If the three-dimensional construction position of the building component in the 3D real-time construction status information generated in the construction confirmation step does not match the three-dimensional construction position of the building component in the completed construction status information, the progress confirmation step is not performed, and an abnormality notification step is performed to notify of the construction abnormality. The progress management method according to claim 7 or 8.

10. The number of types of building materials to be set in advance as targets for progress management is 20 or less. The progress management method according to claim 7 or 8.

11. In the progress management method according to claim 7 or 8, the monitoring and photography step, the coordinate setting step, the construction confirmation step, and the progress confirmation step are performed. A program for a progress management system.

Images