JP2021060199A - Reinforcing bar estimation system, reinforcing bar estimation method and reinforcing bar estimation program - Google Patents

Abstract

To provide a reinforcing bar estimation system, a reinforcing bar estimation method, and a reinforcing bar estimation program for accurately grasping information on arranged reinforcing bars.SOLUTION: A bar arrangement confirmation device 20 includes a control unit 21 connected to an estimation model storage unit 24 storing a center position estimation model that estimates a center position of a reinforcing bar from a point cloud arrangement image and a reinforcement diameter estimation model that estimates a radius from the point cloud arrangement image whose origin is the center position. The control unit 21 acquires data on the point cloud from the surface of the reinforcing bar to be estimated, generates a first image showing the arrangement of the acquired point cloud, and estimates the center position of the reinforcing bar to be estimated using this first image and the center position estimation model. The control unit 21 generates a second image in which the point cloud in the first image is moved so that the estimated center position becomes the origin, estimates the radius of the reinforcing bar to be estimated by using this second image and the reinforcing bar diameter estimation model, and identifies a type of reinforcing bar according to this radius.SELECTED DRAWING: Figure 1

Description

The present invention relates to a reinforcing bar estimation system, a reinforcing bar estimation method, and a reinforcing bar estimation program for estimating information on arranged reinforcing bars.

Conventionally, in the construction of a building, a bar arrangement inspection is performed before concrete is placed. In this bar arrangement inspection, the number, diameter, pitch, etc. of the reinforcing bars are confirmed by comparing the design with the actually placed reinforcing bars. In this case, a technique of acquiring a three-dimensional point cloud by laser measurement to acquire bar arrangement information and a technique of acquiring bar arrangement information using a captured image are being studied (see, for example, Patent Document 1). In the bar arrangement information acquisition system described in this document, the mobile terminal detects a marker from the entire image corresponding to the background bar portion and cuts out an image between the markers. In the bar arrangement information acquisition system, the edge that becomes the boundary line of the area is extracted by detecting the part where the brightness of the pixel is discontinuous in the direction perpendicular to the axial direction of the reinforcing bar, and the unnecessary edge is removed. To do. Next, a frequency distribution is created for the brightness distribution of the image, representative brightness values in the three regions of reinforcing bar, shadow, and background are calculated based on the frequency distribution, and each representative brightness value is used for each edge. When the sandwiched area is specified and the adjacent areas across the edge are the same, the two areas are integrated as one area.

Japanese Unexamined Patent Publication No. 2013-15452

However, when the reinforcing bar arrangement information is acquired from the image analysis, it is difficult to accurately extract the reinforcing bar image due to the influence of the background and the light environment. Further, when a three-dimensional point cloud by laser measurement is used, it is difficult to accurately measure the diameter of the reinforcing bar because many point clouds cannot be acquired for one reinforcing bar or noise is mixed. In particular, with different types of reinforcing bar standards, the difference in reinforcing bar diameter may be about several mm, making it difficult to identify.

The reinforcing bar estimation system that solves the above problems has an input unit that acquires data on a point group indicating a plurality of point positions on the surface of the reinforcing bar arranged in space, and a center that estimates the center position of the reinforcing bar from the point group arrangement image. A model storage unit that stores a position estimation model and a diameter estimation model that estimates the diameter from the origin from a point group arrangement image in which the center position is arranged at the origin, and an output unit that outputs information about the estimation result. A reinforcing bar estimation system including a connected control unit that estimates the reinforcing bar, the control unit acquires data on a point group from the surface of the reinforcing bar to be estimated, and shows the arrangement of the acquired point group. A first image is generated, the center position of the reinforcing bar to be estimated is estimated by using this first image and the center position estimation model, and a point in the first image is set so that the estimated center position becomes the origin. A second image in which the group is moved is generated, the diameter of the reinforcing bar to be estimated is estimated by using this second image and the diameter estimation model, and information on the estimated diameter is output to the output unit. ..

According to the present invention, it is possible to accurately grasp the information of the arranged reinforcing bars.

Explanatory drawing which shows the structure of the reinforcing bar estimation system in embodiment. Explanatory drawing of the hardware configuration example of an embodiment. It is a figure explaining the data structure stored in the storage part in embodiment, (a) is a teacher information storage part, (b) is an estimation model storage part. The flow chart explaining the processing procedure of the learning process in Embodiment. It is explanatory drawing of the data generation processing for learning in an embodiment, (a) is a flow diagram of a processing procedure, (b) is a layout drawing of the center of a reinforcing bar, (c) is an explanatory diagram of the definition of a point cloud setting area, (d). ) Is a distribution map including noise, and (e) is a diagram showing the relationship between the arrangement of points, the center position, and the radius. It is explanatory drawing of the learning process of the center position of a reinforcing bar in an embodiment, (a) is a flow diagram of a processing procedure, (b) is a figure which arranged learning data, (c) is a point cloud with an average value as an origin. The moved figure, (d) is a figure showing the relationship between the image to be learned and the center coordinates. It is explanatory drawing of the learning process of the reinforcing bar diameter in an embodiment, (a) is a flow diagram of a process procedure, (b) is a figure which moves a center position to the origin, (c) shows the relationship between the image to learn and a radius. Figure. It is an explanatory diagram of the bar arrangement confirmation process in the embodiment, (a) is a flow diagram of the processing procedure, (b) is a layout diagram of a point cloud in the reinforcing bar region, and (c) is moved so that the average value is the origin. The point cloud layout diagram, (d) is a diagram showing the estimated center position, (e) is a diagram showing the point cloud layout diagram to move so that the center position is the origin, and (f) is a diagram showing the estimated radius. .. Explanatory drawing which divided the point cloud for each reinforcing bar region in an embodiment. The figure which shows the estimation result of the reinforcing bar diameter in an embodiment. The figure which shows the detail of the estimation result of the reinforcing bar diameter predicted in an embodiment. An explanatory diagram illustrating a process of predicting the reinforcing bar diameter for each aggregation range in which the point cloud is divided in the modified example and specifying the largest reinforcing bar diameter. It is a process flow diagram in the modified example, (a) is a bar arrangement confirmation process, (b) is a learning process of a reinforcing bar area, and (c) is a prediction process of a reinforcing bar area.

Hereinafter, an embodiment in which the reinforcing bar estimation system, the reinforcing bar estimation method, and the reinforcing bar estimation program are embodied will be described with reference to FIGS. 1 to 11. In the reinforcing bar estimation method of the present embodiment, the arranged reinforcing bars are irradiated with laser light, a plurality of three-dimensional point positions are measured for the reflection positions, and these plurality of point positions (point group data) are used to measure the reinforcing bars. Estimate the placement and diameter, and identify the position and type of rebar.

As shown in FIG. 1, in the present embodiment, a case where the diameter of the reinforcing bar 11 of the main bar 11 constituting the reinforcing bar cage 10 is estimated will be described. The reinforcing bar cage 10 includes a plurality of main bars 11 extending in the vertical direction and hoop bars 12 arranged in a direction orthogonal to the main bars 11.

In the present embodiment, the scanning device 15 and the reinforcing bar confirmation device 20 are used for the reinforcing bar arrangement inspection of the reinforcing bar cage 10. Then, the scanning device 15 is moved outside the reinforcing bar cage 10 to acquire the point cloud data of the main reinforcing bar 11.
The scanning device 15 transmits the generated point cloud data to the bar arrangement confirmation device 20. Then, the bar arrangement confirmation device 20 estimates the reinforcing bar diameter and confirms the bar arrangement.

(Hardware configuration example)
FIG. 2 is a hardware configuration example of the information processing device H10 that functions as the scanning device 15 and the bar arrangement confirmation device 20.

The information processing device H10 includes a communication device H11, an input device H12, a display device H13, a storage unit H14, and a processor H15. Note that this hardware configuration is an example, and other hardware may be included.

The communication device H11 is an interface that establishes a communication path with another device and executes data transmission / reception, such as a network interface card or a wireless interface.

The input device H12 is a device that receives input from an administrator or the like, and is, for example, a mouse, a keyboard, or the like. The display device H13 is a display, a touch panel, or the like that displays various information.

The storage unit H14 is a storage device (for example, the reinforcing bar master information storage unit 22 to the confirmation result storage unit 26, which will be described later) that stores data and various programs for executing various functions of the scanning device 15 and the bar arrangement confirmation device 20. is there. An example of the storage unit H14 is a ROM, a RAM, a hard disk, or the like.

The processor H15 controls each process in the scanning device 15 and the bar arrangement confirmation device 20 (for example, the process in the control unit 21 described later) by using the program or data stored in the storage unit H14. Examples of the processor H15 include a CPU, an MPU, and the like. The processor H15 expands a program stored in a ROM or the like into a RAM and executes various processes corresponding to various processes. For example, when the application program of the bar arrangement confirmation device 20 is started, the processor H15 operates a process for executing each process shown in FIGS. 4 to 8 described later.

The processor H15 is not limited to the one that performs software processing for all the processing executed by itself. For example, the processor H15 may include a dedicated hardware circuit (for example, an integrated circuit for a specific application: ASIC) that performs hardware processing for at least a part of the processing executed by the processor H15. That is, the processor H15 is (1) one or more processors that operate according to a computer program (software), (2) one or more dedicated hardware circuits that execute at least a part of various processes, or ( 3) It can be configured as a circuitry including a combination thereof. The processor includes a CPU and a memory such as a RAM and a ROM, and the memory stores a program code or an instruction configured to cause the CPU to execute a process. Memory or computer-readable media includes any available medium accessible by a general purpose or dedicated computer.

(Function of rebar estimation system)
The scanning device 15 shown in FIG. 1 has a built-in laser light source, outputs light in response to a pulse signal, and detects reflected light. In this case, the scanning device 15 scans the laser beam so as to detect the extending direction (vertical direction) of the main bar 11. Then, the scanning device 15 generates a set data (point cloud data) of three-dimensional points regarding the reflection position of the irradiation light based on the time difference between the irradiation timing and the light receiving timing. In this case, the data for each point in the point cloud includes the coordinates (Z value) of the projected height (vertical direction) and the coordinates (X value and Y value) in the horizontal plane (two-dimensional coordinate system).

The bar arrangement confirmation device 20 includes a control unit 21, a reinforcing bar master information storage unit 22, a teacher information storage unit 23, an estimation model storage unit 24, a measurement point group information storage unit 25, and a confirmation result storage unit 26.
By executing the bar arrangement confirmation program, the control unit 21 functions as a learning unit 210, an inspection management unit 211, a separation unit 212, a center position estimation unit 213, and a reinforcing bar diameter estimation unit 214.

The learning unit 210 executes a process of generating an estimation model by machine learning using teacher information. In the present embodiment, the learning unit 210 generates learning data, a center position estimation model for estimating the center position of the reinforcing bar, and a reinforcing bar diameter estimation model for estimating the radius of the reinforcing bar in the learning process.

The learning unit 210 stores a reference angle (± θ) for defining a region for arranging a point cloud, which will be described later, for generating learning data. In the present embodiment, the reference angle θ is ± 60 degrees with respect to a predetermined direction. Further, in order to generate the learning data, the number of points (set number) to be included in one learning data is stored. As the number of settings, for example, "100" can be used.

The learning unit 210 performs machine learning to generate a center position estimation model and a reinforcing bar diameter estimation model. In the generation of the center position estimation model, the point cloud arrangement image and the center position are paired and learned. In the generation of the reinforcing bar diameter estimation model, the point cloud arrangement image and the radius are paired and learned. Here, as machine learning, a deep neural network (DNN) such as deep learning is used. Further, in order to evaluate the point cloud arrangement image used for generating the estimation model as an image of the same size, information on a two-dimensional area (cutout area) for imaging the point cloud into an arrangement image is stored. .. This cutout region is a square region having a predetermined length centered on the origin of the two-dimensional coordinates. The learning unit 210 arranges a group of a plurality of points in this cutout area, counts the number of points included in each grid which is a grid in 0.1 mm increments in this area, and calculates the number of points as the brightness of the grid. Each estimation model is generated using the point cloud arrangement image generated by converting to a value.

The inspection management unit 211 executes the management process in the bar arrangement inspection.
The separation unit 212 executes a process of dividing the point cloud acquired from the reinforcing bar cage 10 into a range (reinforcing bar region) indicating each of the reinforcing bars (main reinforcing bar 11) of the reinforcing bar cage 10.

The center position estimation unit 213 executes a process of estimating the center position of the reinforcing bar by using the point cloud data and the center position estimation model.
The reinforcing bar diameter estimation unit 214 executes a process of estimating the radius of the reinforcing bar by using the point cloud data and the reinforcing bar diameter estimation model.

The reinforcing bar master information storage unit 22 stores the reinforcing bar master information related to each type of reinforcing bar standardized by JIS standards and the like. The reinforcing bar master information of the present embodiment includes information on the type (name) of the reinforcing bar, the nominal diameter (diameter) of the reinforcing bar, and the like.

As shown in FIG. 3A, the teacher information storage unit 23 stores the learning data 230. The learning data 230 is generated and recorded in the learning data generation process described later. The learning data 230 is data used for the learning process of the center position of the reinforcing bar and the learning process of the diameter of the reinforcing bar, which will be described later, and includes data related to the teacher information identifier, the learning point group, the center position, and the radius.

In the teacher information identifier data area, data relating to an identifier for identifying each learning data is recorded.
In the learning point cloud data area, data relating to the position (two-dimensional coordinates) of each point (constituent point) used for learning is recorded.
In the center position data area, data regarding the center position of the point group used for learning is recorded. In the present embodiment, the center position of this point group corresponds to the center position of the reinforcing bar.
In the radius data area, data regarding the radius used for learning is recorded.

As shown in FIG. 3B, the estimation model storage unit 24 stores the center position estimation model 241 and the reinforcing bar diameter estimation model 242.

The center position estimation model 241 is an estimation model for estimating the center position of the point cloud from the point cloud arrangement image. This center position estimation model 241 is stored when the learning process of the center position of the reinforcing bar is executed using the learning data 230.

The reinforcing bar diameter estimation model 242 is an estimation model for calculating the radius (the length from the center position to the outer surface of the reinforcing bar) of the reinforcing bar obtained from the point cloud arrangement image. This reinforcing bar diameter estimation model 242 is stored when the learning process of the radius of the reinforcing bar is executed using the learning data 230.

The measurement point cloud information storage unit 25 stores the point cloud data acquired in the reinforcing bar to be processed for confirming the bar arrangement information. This point cloud data is recorded in association with the point cloud identification information that identifies each point cloud data. This point cloud data is three-dimensional coordinate data of the point cloud of the reinforcing bar cage 10 measured by using the scanning device 15.

The confirmation result storage unit 26 records the reinforcement arrangement confirmation information about the reinforcing bar to be confirmed. In this bar arrangement confirmation information, the center position of each reinforcing bar identified from the acquired point cloud and each reinforcing bar type (name of the reinforcing bar) are recorded in association with the point cloud identification information.

(Learning process)
Next, the learning process will be described with reference to FIGS. 4 to 7. This learning process is executed before the bar arrangement confirmation work is performed in the bar arrangement inspection.

As shown in FIG. 4, the control unit 21 of the bar arrangement confirmation device 20 executes the learning data generation process (step S1-1). Details of this process will be described later with reference to FIG.
Next, the control unit 21 of the bar arrangement confirmation device 20 executes the learning process of the center position of the reinforcing bar (step S1-2). Details of this process will be described later with reference to FIG.
Then, the control unit 21 of the bar arrangement confirmation device 20 executes the learning process of the reinforcing bar diameter (step S1-3). Details of this process will be described later with reference to FIG. 7.

<Data generation process for learning>
Next, the learning data generation process (step S1-1) will be described with reference to FIG. This learning data generation process is repeatedly executed according to the number of learning data for each type of reinforcing bar defined by the JIS standard. In this embodiment, "3000" data are used as the number of learning data.

First, as shown in FIG. 5A, the control unit 21 of the bar arrangement confirmation device 20 executes the arrangement process of the center of the reinforcing bar (step S2-1). Specifically, the learning unit 210 of the control unit 21 randomly sets the point coordinates (center position) at the center of the reinforcing bar in the predetermined two-dimensional display area.
For example, as shown in FIG. 5B, the center position C1 is set.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the definition process of the point cloud setting area (step S2-2). Specifically, the learning unit 210 of the control unit 21 randomly sets the reference direction at the center position C1. Then, the range of the reference angle (± θ) is specified as the point cloud setting area with respect to this reference direction.

For example, as shown in FIG. 5C, it is assumed that the reference direction D1 is set from the center position C1. In this case, the reference angle (± 60 degrees) range is set as the point cloud setting area with respect to the reference direction D1.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes a point cloud generation process to which noise is added (step S2-3). Specifically, the learning unit 210 of the control unit 21 sets the radius of the point cloud including the noise generated under the condition stored in advance.

For example, as shown in FIG. 5D, noise is generated by multiplying a normal distribution with an average value of "0" and a standard deviation of "1.5" by a random number of "-1.5 to 1.5" to generate noise. Generate a distribution containing. Then, the number of constituent points to be arranged is assigned according to this distribution, and the radius to be assigned to each constituent point is specified.

Next, the learning unit 210 sets a random angle with respect to each constituent point to which the radius is assigned. Here, a random angle within the reference angle range is set with respect to the reference direction D1.
Then, from the center position C1, the two-dimensional coordinates of each constituent point are calculated according to the assigned radius and the set angle for each constituent point.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the learning data recording process (step S2-4). Specifically, the learning unit 210 of the control unit 21 assigns the teacher information identifier, and the two-dimensional coordinate data (learning point group) of each constituent point generated in step S2-3, the center position C1, and the radius. The learning data 230 including r1 is generated and recorded in the teacher information storage unit 23.

For example, as shown in FIG. 5E, the learning data 230 is associated with point cloud data consisting of a center position C1, a radius r1, and coordinates of each constituent point.
The above processing is repeated for each type of reinforcing bar specified by JIS, and the learning data 230 for the number of generated learning data is recorded in the teacher information storage unit 23.

<Learning process of the center position of the reinforcing bar>
Next, the learning process (step S1-2) of the center position of the reinforcing bar will be described with reference to FIG.
First, as shown in FIG. 6A, the learning data 230 to be processed is sequentially specified, and the following processing is repeatedly executed for each learning data 230.

The control unit 21 of the bar arrangement confirmation device 20 executes a call processing of learning data (step S3-1). Specifically, the learning unit 210 of the control unit 21 acquires the learning data 230 to be processed from the teacher information storage unit 23.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the origin arrangement process of the average value of the X value and the Y value of the point cloud (step S3-2). Specifically, the learning unit 210 of the control unit 21 adds all the X values of the learning point cloud included in each acquired learning data 230, divides by the number of points, and obtains the average value on the X axis. calculate. Further, the learning unit 210 adds all the Y values of the learning point cloud included in each learning data 230 and divides by the number of points of the learning point cloud to calculate the average value on the Y axis. Then, the learning unit 210 calculates a conversion vector whose origin is the average value of the calculated X-axis and Y-axis. The learning unit 210 calculates the coordinates of each point of the learning point cloud after movement by using the calculated conversion vector and the coordinates of each point of the learning point cloud of the learning data 230.
FIG. 6 (b) shows a state before moving the learning point cloud, and FIG. 6 (c) shows a state after moving the learning point cloud so that the average value of the X coordinate and the Y coordinate becomes the origin. Is shown.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the imaging process (step S3-3). Specifically, the learning unit 210 of the control unit 21 arranges each constituent point of the coordinates calculated in step S3-2 in the two-dimensional coordinates, cuts out in the cutout area A1, and sets the number of points included in each grid. A point cloud arrangement image in which each grid is displayed with a corresponding brightness value is generated.

Then, the control unit 21 of the bar arrangement confirmation device 20 executes the first teacher data generation process (step S3-4). Specifically, the learning unit 210 of the control unit 21 uses the generated point cloud arrangement image as the input layer, and the first teacher data using the center position associated with the point cloud that generated the point cloud arrangement image as the output layer. Generate.
For example, first teacher data is generated in which a point cloud arrangement image using a point cloud obtained by removing the center position C2 from the image shown in FIG. 6D is used as an input layer and the center position C2 is used as an output layer.
The above processing is executed for each learning data.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the generation process of the center position estimation model (step S3-5). Specifically, the learning unit 210 of the control unit 21 performs machine learning using a plurality of first teacher data using the point cloud arrangement image as the input layer and the center position as the output layer, and performs machine learning from the point cloud arrangement image. A center position estimation model 241 of the intermediate layer for estimating the center position is generated. Then, the learning unit 210 stores the generated center position estimation model 241 in the estimation model storage unit 24.

<Reinforcing bar diameter learning process>
Next, the learning process of the reinforcing bar diameter (step S1-3) will be described with reference to FIG. 7.
First, as shown in FIG. 7A, the learning data 230 to be processed is sequentially specified, and the following processing is repeatedly executed for each learning data 230.

The control unit 21 of the bar arrangement confirmation device 20 executes the learning data calling process in the same manner as in step S3-1 (step S4-1).
Next, the control unit 21 of the bar arrangement confirmation device 20 executes the origin arrangement process at the center position (step S4-2). Specifically, the learning unit 210 of the control unit 21 calculates a conversion vector whose origin is the center position C2, as shown in FIG. 7B. Then, the learning unit 210 calculates the coordinates of each point of the learning point cloud after the movement by using the calculated conversion vector and the coordinates of each point of the learning point cloud of the learning data 230.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the imaging process (step S4-3). Specifically, the learning unit 210 of the control unit 21 arranges each constituent point of the coordinates calculated in step S4-2 in the two-dimensional coordinates, cuts out in the cutout area A1, and sets the number of points included in each grid. A point cloud arrangement image in which each grid is displayed with a corresponding brightness value is generated.

Then, the control unit 21 of the bar arrangement confirmation device 20 executes the second teacher data generation process (step S4-4). Specifically, the learning unit 210 of the control unit 21 generates second teacher data using the generated point cloud arrangement image as the input layer and the radius associated with the point cloud that generated the point cloud arrangement image as the output layer. To do.
For example, a second teacher data is generated in which a point cloud arrangement image using a point cloud obtained by removing the center position C2 and the radius r3 from the image shown in FIG. 7C is used as an input layer and the radius r3 is used as an output layer.
The above processing is executed for each learning data.

Then, the control unit 21 of the bar arrangement confirmation device 20 executes the generation process of the reinforcing bar diameter estimation model (step S4-5). Specifically, the learning unit 210 of the control unit 21 performs machine learning using a plurality of second teacher data in which the point cloud arrangement image is set as the input layer and the radius is set as the output layer, and the radius is obtained from the point cloud arrangement image. Generates a reinforcing bar diameter estimation model 242 for the intermediate layer to estimate. Then, the learning unit 210 stores the generated reinforcing bar diameter estimation model 242 in the estimation model storage unit 24.

(Reinforcement confirmation processing)
Next, the bar arrangement confirmation process executed in the bar arrangement inspection will be described with reference to FIGS. 8 to 11. Here, the scanning device 15 is used to move around the reinforcing bar cage 10 and acquire point cloud data from the outer circumference of the reinforcing bar cage 10. Then, the bar arrangement confirmation device 20 acquires the point cloud data from the scanning device 15 and stores it in the measurement point cloud information storage unit 25.

First, as shown in FIG. 8A, the control unit 21 of the bar arrangement confirmation device 20 executes the point cloud data acquisition process (step S5-1). Specifically, the inspection management unit 211 of the control unit 21 acquires point cloud data from the measurement point cloud information storage unit 25.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the projection process of the point cloud on the two-dimensional coordinate system (step S5-2). Specifically, the inspection management unit 211 of the control unit 21 generates point cloud data including X values and Y values in which the Z values of the point cloud data are all the same value.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes a point cloud separation process for each reinforcing bar region (step S5-3). Specifically, the separation unit 212 of the control unit 21 outputs a confirmation screen to the display device H13. On this confirmation screen, a point cloud projected on a two-dimensional plane is displayed.

FIG. 9 is a diagram in which the point cloud data acquired from the scanning device 15 is projected onto a two-dimensional plane when moving around the reinforcing bar cage 10. Then, it is divided into a plurality of reinforcing bar regions A2 including a group of point clouds generated by each reinforcing bar. In this case, the administrator sets the range of the point cloud indicating each reinforcing bar as the reinforcing bar region A2 on the confirmation screen.

Then, as shown in FIG. 8A, the control unit 21 of the bar arrangement confirmation device 20 sequentially identifies the reinforcing bar region as the processing target region, and executes the following processing. In this case, the separation unit 212 of the control unit 21 specifies the X value and the Y value of the constituent points of the point group included in each reinforcing bar region.

First, the control unit 21 of the bar arrangement confirmation device 20 executes the point cloud movement process so that the average value of the X value and the Y value becomes the origin (step S5-4). Specifically, the center position estimation unit 213 of the control unit 21 calculates the average value of the X values and the average value of the Y values of all the points in the reinforcing bar region to be processed.

Next, the center position estimation unit 213 calculates the conversion vector so that the calculated average value of each axis becomes the origin. The center position estimation unit 213 calculates the coordinates of each point after movement by using the calculated conversion vector and the coordinates of each point in the reinforcing bar region to be processed.
For example, as shown in FIG. 8C, the point cloud shown in FIG. 8B is moved so that the average value of the X value and the Y value becomes the origin.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the imaging process (step S5-5). Specifically, the center position estimation unit 213 of the control unit 21 arranges each point of the coordinates calculated in step S5-4 in two-dimensional coordinates, cuts out in the cutout area A1, and counts the number of points included in each grid. A point cloud arrangement image in which each grid is displayed with a brightness value corresponding to is generated.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the center position estimation process using the center position estimation model (step S5-6). Specifically, the center position estimation unit 213 of the control unit 21 uses the center position estimation model 241 recorded in the estimation model storage unit 24 and the point cloud arrangement image generated in step S5-5 to obtain the center position. Estimate the two-dimensional coordinates of.
For example, as shown in FIG. 8D, the center position C4 is estimated.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the point cloud movement process so that the estimated center position is the origin (step S5-7). Specifically, the reinforcing bar diameter estimation unit 214 of the control unit 21 calculates a conversion vector whose origin is the estimated center position. Then, the reinforcing bar diameter estimation unit 214 calculates the coordinates of each point after movement by using the calculated conversion vector and the coordinates of each point in the reinforcing bar region to be processed.
For example, as shown in FIG. 8E, each point is moved so that the estimated center position C4 is the origin.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the imaging process (step S5-8). Specifically, the reinforcing bar diameter estimation unit 214 of the control unit 21 arranges each point of the coordinates calculated in step S5-7 in two-dimensional coordinates, cuts out in the cutout area A1, and counts the number of points included in each grid. A point cloud arrangement image in which each grid is displayed with a brightness value corresponding to is generated.

Then, the control unit 21 of the bar arrangement confirmation device 20 executes the estimation process of the radius of the reinforcing bar using the reinforcing bar diameter estimation model (step S5-9). Specifically, the reinforcing bar diameter estimation unit 214 of the control unit 21 inputs a point cloud arrangement image into the reinforcing bar diameter estimation model 242 recorded in the estimation model storage unit 24, and estimates the radius of the reinforcing bar.
For example, as shown in FIG. 8 (f), the radius r4 is estimated.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes a recording process of the position and type of the reinforcing bar (step S5-10). Specifically, the inspection management unit 211 of the control unit 21 associates the center position (coordinates) estimated in step S5-6 with the point cloud identification information as the position of the reinforcing bar in this reinforcing bar region, and stores the confirmation result. Record at 26. Further, the inspection management unit 211 uses the reinforcing bar master information storage unit 22 to specify the reinforcing bar type corresponding to the diameter of the reinforcing bar estimated in step S5-9. The inspection management unit 211 associates the specified reinforcing bar type with the position of the reinforcing bar in this reinforcing bar region and records it in the confirmation result storage unit 26.

Then, the above processing is repeatedly executed for each reinforcing bar region.
After that, the control unit 21 of the bar arrangement confirmation device 20 displays the position and type of the reinforcing bar recorded in the confirmation result storage unit 26 on the display device H13.

(Estimation result of reinforcing bar diameter)
In FIGS. 10 and 11, training data is generated by the above-described processing, a center position estimation model 241 and a reinforcing bar diameter estimation model 242 are generated based on the training data, and the reinforcing bars are used by using these estimation models. The simulation result of estimating the diameter is shown.

The upper part of FIG. 10 shows the simulation results for the “D10” reinforcing bar, the middle part shows the “D13” reinforcing bar, and the lower part shows the simulation results for the “D16” reinforcing bar. In each of the figures of FIG. 11, the pentagon indicates the center position arranged in the learning data generation process, and the point cloud of the learning data is arranged in an arc shape around the pentagon. Further, in each figure, the star shape is the origin, and the circumference surrounding the star shape is the circumference corresponding to the estimated radius of the reinforcing bar. In each of the simulation result diagrams, the point cloud of the learning data arranged at the estimated center position is almost arranged on the circumference of each reinforcing bar diameter of the star shape, indicating that the simulation result is accurate. There is.

FIG. 11 shows the distribution of radii estimated for each reinforcing bar (D10, D13, D16) by a plurality of simulation results. Centering on the nominal diameter of each reinforcing bar shown by the alternate long and short dash line in FIG. 12, the average error was 0.4 mm or less, and the standard deviation was around 0.3 mm. Therefore, since the error and standard deviation are small for the diameters of different reinforcing bars, the type of reinforcing bar can be clearly specified.

According to this embodiment, the following effects can be obtained.
(1) In the present embodiment, the control unit 21 of the bar arrangement confirmation device 20 uses the center position estimation model to estimate the center coordinates (step S5-6) and the reinforcing bar diameter estimation model to estimate the radius of the reinforcing bar. The process (step S5-9) is executed. As a result, the control unit 21 specifies the center position and the radius of the reinforcing bar by using an estimation model in which the center position and the radius of the reinforcing bar are individually machine-learned. Compared with the case where the radius of the reinforcing bar is estimated without the center position as the origin, the radius can be specified more accurately by estimating the radius of the reinforcing bar with the center position as the origin. Therefore, even when the difference in reinforcing bar diameter is only about several mm in the standards of different types of reinforcing bars, the type of reinforcing bar can be accurately specified.

(2) In the present embodiment, the control unit 21 of the bar arrangement confirmation device 20 performs the noise-added point cloud generation process (step S2-3) and learning data generation process (step S1-1). The data recording process (step S2-4) is executed. As a result, learning data can be generated using a point cloud that takes into account noise generated during measurement, and each estimation model (241,242) can be generated using the learning data.

(3) In the present embodiment, the control unit 21 of the bar arrangement confirmation device 20 uses the learning data 230 to perform the origin placement process (step S3-2) of the average value of the X value and the Y value of the point cloud, and the image. The conversion process (step S3-3), the first teacher data generation process (step S3-4), and the center position estimation model generation process (step S3-5) are executed. As a result, it is possible to generate a center position estimation model that predicts the center position by learning using the point cloud arrangement image with the average value of the X value and the Y value of the point cloud as the origin and the center position.

(4) In the present embodiment, the control unit 21 of the bar arrangement confirmation device 20 uses the learning data 230 to perform the origin arrangement process (step S4-2), the imaging process (step S4-3), and the image formation process (step S4-2) at the center position. The second teacher data generation process (step S4-4) and the reinforcing bar diameter estimation model generation process (step S4-5) are executed. Thereby, the reinforcing bar diameter estimation model 242 for estimating the radius of the reinforcing bar can be generated by learning using the point cloud arrangement image with the center position as the origin.

(5) In the present embodiment, the control unit 21 of the bar arrangement confirmation device 20 acquires a plurality of point cloud data arranged in the extending direction (vertical direction) of the main bar 11, and points groups to the two-dimensional coordinate system. (Step S5-2) is executed. As a result, the diameter of the reinforcing bar can be estimated accurately by using the point cloud arrangement image including many points in projecting the vertically dispersed point cloud onto the two-dimensional plane.

(6) In the present embodiment, the control unit 21 of the bar arrangement confirmation device 20 executes the point cloud separation process for each reinforcing bar region (step S5-3), and determines the center coordinates and the radius of the reinforcing bar for each reinforcing bar region. presume. As a result, it is possible to confirm the bar arrangement for each main bar 11 by using the point cloud data obtained by collectively acquiring the plurality of main bars 11.

This embodiment can be modified and implemented as follows. The present embodiment and the following modified examples can be implemented in combination with each other within a technically consistent range.
-In the above embodiment, the position of the reinforcing bar and the radius of the reinforcing bar are estimated using the point cloud data acquired by the scanning device 15 that has moved the outer circumference of the reinforcing bar cage 10. The position and diameter of the main bar 11 may be estimated using the point cloud data acquired for each predetermined distance by moving the scanning device 15. In this case, since there is a possibility that a plurality of point cloud data can be acquired in one reinforcing bar, the statistical value (average value, mode value, etc.) of each reinforcing bar diameter predicted using the multiple point cloud data is used. , Reinforcing bar diameter may be estimated. Further, for example, the three-dimensional shape (point cloud) may be specified by a method such as Structure from Motion using a plurality of camera images.

-In the above embodiment, the control unit 21 of the bar arrangement confirmation device 20 executes the projection process (step S5-2) of the point cloud on the two-dimensional coordinate system. In this process, the point cloud for one reinforcing bar was projected onto one horizontal plane to estimate the reinforcing bar, but the reinforcing bar was estimated using the images projected from the point cloud for one reinforcing bar on a plurality of horizontal planes. May be done. In this case, the point cloud is divided into a plurality of regions in the extending direction (Z direction) of the reinforcing bars, the reinforcing bars are estimated for each region, and the information on the estimated reinforcing bars in each region is integrated.

For example, a point cloud is divided for each predetermined range (aggregation range) in the extending direction of the reinforcing bar, and the divided point cloud is projected on each horizontal plane. Next, the diameter of the reinforcing bar in each horizontal plane is estimated using the point cloud projected on each horizontal plane, and the type of reinforcing bar corresponding to the diameter of the reinforcing bar is specified. Then, the type of the reinforcing bar may be specified by using the statistical value of the type of the reinforcing bar estimated in all horizontal planes.

Specifically, the control unit 21 of the bar arrangement confirmation device 20 stores the length of the aggregation range.
Then, as shown in FIG. 12, the control unit 21 of the bar arrangement confirmation device 20 executes the identification process of the point cloud included in each aggregation range (step S6-1). Specifically, the inspection management unit 211 of the control unit 21 is included for each stored length (aggregation range A3) based on the Z values of the plurality of points P1 acquired in the extending direction of the main bar 11. Each point P1 is specified.

Then, the control unit 21 of the bar arrangement confirmation device 20 executes the process of specifying the center position and the reinforcing bar type for each aggregation range (step S6-2). Specifically, the inspection management unit 211 of the control unit 21 projects the point P1 included in each aggregation range A3 onto the horizontal plane. Then, the control unit 21 uses the coordinates of the projected point P1 to move the point cloud so that the average value of the X value and the Y value becomes the origin (step S5-4) to record the position and type of the reinforcing bar. The process (step S5-10) is executed.

After that, the control unit 21 of the bar arrangement confirmation device 20 executes the process of specifying the center position and the type of the reinforcing bar using the values specified in all the aggregation ranges (step S6-3). Specifically, the inspection management unit 211 of the control unit 21 specifies the average value of the center positions calculated in each aggregation range as the position of the main bar 11. Further, the inspection management unit 211 specifies the most specified reinforcing bar type among the reinforcing bar types specified in each aggregation range as the reinforcing bar type of the main reinforcing bar 11. Thereby, the accuracy of the estimated reinforcing bar diameter can be improved.

Further, instead of using all the measured point clouds, a part of the point cloud may be used to predict the diameter of the reinforcing bar and specify the type of the reinforcing bar. In this case, the type of reinforcing bar is specified by using the statistical value of the type of reinforcing bar specified for each part. For example, a plurality of groups including randomly extracted point groups are generated, and the prediction of the reinforcing bar diameter and the identification of the reinforcing bar type corresponding to the reinforcing bar diameter are repeated for each group, and the reinforcing bar type with the most prediction is selected. Estimated to be the type of reinforcing bar.

Specifically, the bar arrangement confirmation process shown in FIG. 13A may be executed. In this case, the control unit 21 of the bar arrangement confirmation device 20 stores data on the number of points randomly extracted in one group and the number of groups to be generated.

In this process, the control unit 21 of the bar arrangement confirmation device 20 performs a point cloud data acquisition process (step 7-1) and a point cloud separation process for each reinforcing bar region, as in steps S5-1 and S5-3. Execute (step 7-2).

Then, the reinforcing bar regions to be processed are sequentially specified, and the following processing is repeatedly executed for each reinforcing bar region.
First, the control unit 21 of the bar arrangement confirmation device 20 executes the point cloud group creation process (S7-3). Specifically, the inspection management unit 211 of the control unit 21 randomly extracts as many points as the number of points in the point group included in the reinforcing bar region to create a group. The inspection management unit 211 repeats this group creation process to generate as many groups as there are groups.

Then, the control unit 21 of the bar arrangement confirmation device 20 executes the following processing for each generated point cloud group.
First, the control unit 21 of the bar arrangement confirmation device 20 executes a projection process of the point cloud of the point cloud group onto the two-dimensional coordinate system in the same manner as in step S5-2 (step S7-4).

Then, the control unit 21 of the bar arrangement confirmation device 20 performs the point cloud movement process (step S5-4) to the position of the reinforcing bar so that the average value of the X value and the Y value becomes the origin, as in the bar arrangement confirmation process. And the kind of recording process (step S5-10) is executed. Here, the position and type of the reinforcing bar are temporarily stored in the memory.

Then, the control unit 21 of the bar arrangement confirmation device 20 executes the process of determining the position and type of the reinforcing bar (step S7-5). Specifically, the inspection management unit 211 of the control unit 21 calculates the average value of the positions (center coordinates) of the reinforcing bars of the point cloud group temporarily stored in the memory, specifies this average value as the position of the reinforcing bars, and determines the average value. The confirmation result is recorded in the storage unit 26. Further, the inspection management unit 211 identifies the most common type of reinforcing bar in the recorded point cloud group and records it in the confirmation result storage unit 26.

-In the above embodiment, the control unit 21 of the bar arrangement confirmation device 20 sets the point cloud of each reinforcing bar according to the instruction of the administrator in the point cloud dividing process (step S5-3) for each reinforcing bar region. Reinforcing bar areas in the range shown were set. The control unit 21 of the bar arrangement confirmation device 20 may execute the identification of each reinforcing bar region. In this case, for example, learning processing such as deep learning or pattern matching processing can be used.

As shown in FIGS. 13 (b) and 13 (c), the point cloud may be separated for each reinforcing bar region by performing the learning processing of the reinforcing bar region and the prediction processing of the reinforcing bar region.
As shown in FIG. 13B, in the learning process of the reinforcing bar region, the control unit 21 of the reinforcing bar arrangement confirmation device 20 executes the imaging process (step S8-1). Specifically, the learning unit 210 of the control unit 21 generates an image in which the point cloud projected onto the two-dimensional coordinate system is arranged.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the acquisition process of the reinforcing bar region (step S8-2). Specifically, the learning unit 210 of the control unit 21 associates the arrangement image with the reinforcing bar region set by the instruction of the administrator.

Then, the control unit 21 of the bar arrangement confirmation device 20 executes the area estimation model generation process (step S8-3). Specifically, the learning unit 210 of the control unit 21 generates a region estimation model using the teacher data with the point cloud image as the input layer and the reinforcing bar region as the output layer, and records it in the estimation model storage unit 24. ..

After that, in the point cloud separation process (step S5-3) for each reinforcing bar region in the reinforcing bar arrangement confirmation process, the prediction processing of the reinforcing bar region shown in FIG. 13 (c) is executed.
In this process, the control unit 21 of the bar arrangement confirmation device 20 executes an imaging process of the point group to be processed (step S9-1). Specifically, the separation unit 212 of the control unit 21 generates an image (arrangement image) showing the arrangement of the point cloud projected on the horizontal plane.

Next, the control unit 21 of the bar arrangement confirmation device 20 executes the identification process of each reinforcing bar region using the region estimation model (step S9-2). The separation unit 212 of the control unit 21 identifies each reinforcing bar region in the arrangement image by using the arrangement image generated in step S9-1 and the area estimation model.

Then, the separation unit 212 of the control unit 21 executes the point cloud separation process for each of the specified reinforcing bar regions (step S9-3). Specifically, the separation unit 212 of the control unit 21 identifies each point having an X value and a Y value in each reinforcing bar region.

-In the above embodiment, the control unit 21 of the reinforcing bar arrangement confirmation device 20 has specified the center position and the reinforcing bar diameter (reinforcing bar type) of the main reinforcing bar 11 of the reinforcing bar cage 10. The reinforcing bar for which the reinforcing bar diameter is estimated to confirm the reinforcing bar arrangement is not limited to the main reinforcing bar 11. For example, it may be used for estimating the reinforcing bar diameter of the hoop bar 12 of the reinforcing bar cage 10 or estimating the reinforcing bar diameter of the slab bars arranged in a flat lattice.

-In the above embodiment, the control unit 21 of the bar arrangement confirmation device 20 executes learning data generation processing, learning processing of the center position of the reinforcing bar, and learning processing of the reinforcing bar diameter (steps S1-1 to S1-3). , Each estimation model (241,242) was generated. The learning process of the estimation model (241,242) is not limited to the case where the learning data generation process is performed immediately before. For example, the center position of the reinforcing bar and the reinforcing bar diameter may be learned by using the point cloud data acquired in the past and the measured value of the reinforcing bar diameter obtained from the point cloud. Further, the learning process of the center position and the learning process of the reinforcing bar diameter do not have to be executed as a series of learning processes. The distance from the scanning device (imaging device) to the reinforcing bar may be used during machine learning. In this case, the noise generation rate may fluctuate according to the shooting distance, but by machine learning using the distance, the center position and the reinforcing bar diameter are learned in consideration of the noise according to the shooting distance. be able to.

-In the above embodiment, the control unit 21 of the bar arrangement confirmation device 20 estimates the radius of the reinforcing bar using the reinforcing bar diameter estimation model 242, identifies the type of the reinforcing bar according to this radius, and records the type of the reinforcing bar. To display. The information regarding the diameter of the output reinforcing bar is not limited to the type of reinforcing bar. For example, the estimated reinforcing bar diameter, the statistical value of the variation of the estimated diameter, and the like may be output.

θ ... region range angle, A1 ... cutout region, A2 ... reinforcing bar region, A3 ... aggregation range, C1, C2, C4 ... center position, P1 ... point, r1, r3, r4 ... radius, H10 ... information processing device, H11 ... communication device, H12 ... input device, H13 ... display device, H14 ... storage unit, H15 ... processor, 10 ... reinforcing bar cage, 11 ... main bar, 12 ... hoop bar, 15 ... scanning device, 20 ... bar arrangement confirmation device, 21 ... Control unit, 22 ... Reinforcing bar master information storage unit, 23 ... Teacher information storage unit, 24 ... Estimated model storage unit, 25 ... Measurement point group information storage unit, 26 ... Confirmation result storage unit, 210 ... Learning unit, 211 ... Inspection Management unit, 212 ... Separation unit, 213 ... Center position estimation unit, 214 ... Reinforcing bar diameter estimation unit, 230 ... Learning data, 241 ... Center position estimation model, 242 ... Reinforcing bar diameter estimation model.

Claims (7)

An input unit that acquires data on a point cloud indicating multiple point positions on the surface of reinforcing bars arranged in space, and an input unit.
Model memory that stores a center position estimation model that estimates the center position of the reinforcing bar from the point group arrangement image and a diameter estimation model that estimates the diameter from the origin from the point group arrangement image in which the center position is arranged at the origin. Department and
A reinforcing bar estimation system that has a control unit connected to an output unit that outputs information about the estimation result and estimates the reinforcing bars.
The control unit
Obtain data on the point cloud from the surface of the rebar to be estimated,
A first image showing the arrangement of the acquired point cloud is generated, and the center position of the reinforcing bar to be estimated is estimated by using this first image and the center position estimation model.
A second image in which the point cloud in the first image is moved so that the estimated center position becomes the origin is generated, and the diameter of the reinforcing bar to be estimated is determined by using this second image and the diameter estimation model. A reinforcing bar estimation system characterized in that it estimates and outputs information on the estimated diameter to the output unit. The control unit
Determine the center position for learning and the radius of the reinforcing bar,
Based on the set center position and radius, the position of the constituent points including noise is determined.
A first image for learning a point cloud including the constituent points is generated, and the first image is generated.
A first teacher data in which the first image for learning and the center position for learning are associated with each other is generated.
Machine learning using the first teacher data is performed to generate the center position estimation model, which is recorded in the model storage unit.
A second teacher data in which the second learning image arranged by moving each of the constituent points so that the center position for learning becomes the origin and the set radius is generated is generated.
The reinforcing bar estimation system according to claim 1, wherein machine learning using the second teacher data is performed to generate the diameter estimation model and record the diameter estimation model in the model storage unit. The control unit
Data on the point cloud of the reinforcing bar is acquired from the input unit, and the data is obtained.
The reinforcing bar estimation system according to claim 1 or 2, wherein the point cloud is projected onto a plane orthogonal to the extending direction of the reinforcing bar to generate a first image. The control unit
For each of a plurality of aggregation ranges divided by a reference length in the extending direction of each reinforcing bar, a point cloud included in each of the aggregation ranges is projected onto the plane to generate the first image.
Using the first image for each aggregation range, the diameter of the reinforcing bar to be estimated is estimated.
The reinforcing bar estimation system according to claim 3, wherein the reinforcing bar to be estimated is specified based on the statistical value regarding the diameter of the reinforcing bar estimated for each aggregation range. The control unit
Data on point clouds from a plurality of reinforcing bars is acquired from the input unit, and data is obtained.
The point cloud is divided into reinforcing bar regions including the point cloud indicating each reinforcing bar.
The reinforcing bar estimation system according to any one of claims 1 to 4, wherein the diameter of the reinforcing bar is estimated for each of the reinforcing bar regions. An input unit that acquires data on a point cloud indicating multiple point positions on the surface of reinforcing bars arranged in space, and an input unit.
Model memory that stores a center position estimation model that estimates the center position of the reinforcing bar from the point group arrangement image and a diameter estimation model that estimates the diameter from the origin from the point group arrangement image in which the center position is arranged at the origin. Department and
This is a method of estimating reinforcing bars using a reinforcing bar estimation system equipped with a control unit connected to an output unit that outputs information about the estimation result.
The control unit
Obtain data on the point cloud from the surface of the rebar to be estimated,
A first image showing the arrangement of the acquired point cloud is generated, and the center position of the reinforcing bar to be estimated is estimated by using this first image and the center position estimation model.
A second image in which the point cloud in the first image is moved so that the estimated center position becomes the origin is generated, and the diameter of the reinforcing bar to be estimated is determined by using this second image and the diameter estimation model. A reinforcing bar estimation method characterized by estimating and outputting information on the estimated diameter to the output unit. An input unit that acquires data on a point cloud indicating multiple point positions on the surface of reinforcing bars arranged in space, and an input unit.
Model memory that stores a center position estimation model that estimates the center position of the reinforcing bar from the point group arrangement image and a diameter estimation model that estimates the diameter from the origin from the point group arrangement image in which the center position is arranged at the origin. Department and
A program that estimates the diameter of a reinforcing bar using a reinforcing bar estimation system equipped with a control unit connected to an output unit that outputs information about the estimation result.
The control unit
Obtain data on the point cloud from the surface of the rebar to be estimated,
A first image showing the arrangement of the acquired point cloud is generated, and the center position of the reinforcing bar to be estimated is estimated by using this first image and the center position estimation model.
A second image in which the point cloud in the first image is moved so that the estimated center position becomes the origin is generated, and the diameter of the reinforcing bar to be estimated is determined by using this second image and the diameter estimation model. A reinforcing bar estimation program characterized in that it functions as a means for estimating and outputting information on the estimated diameter to the output unit.

Images