JP2016142601A - Construction quality evaluation program, construction quality evaluation method, and construction quality evaluation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect and identify a construction failure on the basis of the appearance of concrete after a form removal.SOLUTION: The construction quality evaluation program causes a computer to execute the following steps: an imaging step of imaging a concrete structure after a form removal formed in a formwork construction method, and generating image data; and an evaluation step of evaluating the imaged construction quality of the concrete structure, using the reference information that is previously machine-learned while the image of the concrete structure after the form removal is associated with the information indicating the evaluation related to the construction quality of the concrete structure, and the image data generated in the imaging step, and outputting the information related to the construction quality.SELECTED DRAWING: Figure 6

Description

  The present technology relates to a construction quality evaluation program, a construction quality evaluation method, and a construction quality evaluation apparatus.

  On the demolding surface of concrete, bubbles, flutter, bean boards, flow stripes, cold joints, subsidence cracks, sanding due to bleeding, etc. are generated. Conventionally, a concrete engineer visually confirms the concrete after demolding to judge the quality and take measures to improve the quality.

  In addition, when the variation coefficient of the brightness | luminance of the photographic data of a concrete structure is small, the technique which judges that quality is favorable since there are few external appearance nonuniformities on the concrete surface is proposed (patent document 1).

  In addition, a technique has been proposed in which a neural network is used to learn image feature amounts and an image of an inspection object is recognized (Patent Document 2).

JP 2013-127415 A JP-A-5-28245

  Conventionally, knowledge and experience of engineers have been required to judge the construction quality of concrete from the appearance. In other words, the accuracy of judgment depends on the knowledge and experience of engineers, and it has been difficult to quantify under a uniform standard.

  An object of the present invention is to make it possible to objectively evaluate construction quality based on the appearance of concrete after demolding and the tacit knowledge of a concrete engineer.

  The construction quality evaluation program according to the present invention images a concrete structure after demolding formed by a formwork method, generates an image data, an image of the concrete structure after demolding, and the concrete Assessing the construction quality of the imaged concrete structure using the reference information that has been pre-machine-learned in association with information indicating the construction quality of the structure and the image data generated in the imaging step, the construction quality And causing the computer to execute an evaluation step for outputting information on the information.

In this way, the construction quality can be objectively evaluated based on the image of the concrete structure after demolding. In addition, since it is possible to make a judgment based on the reference information by machine learning, if a user of the construction quality evaluation program according to the present invention has no knowledge or experience if the evaluation by a concrete engineer with sufficient experience is learned beforehand. Evaluation can be performed based on the tacit knowledge of the concrete engineer, and variations in judgment by the engineer are reduced. In addition, “implicit knowledge” of a concrete engineer means an evaluation standard that is difficult for verbalization, acquired by an experienced concrete engineer. Such an evaluation standard can be obtained as a function (model) used for evaluation by performing machine learning such as deep learning using a neural network. And the score which shows the evaluation result regarding construction quality and construction quality is obtained by inputting the image of the concrete structure of evaluation object into the function used for evaluation. Moreover, since the image of the concrete structure after demolding, that is, before finishing is used, the construction quality can be evaluated from the surface state, and an objective record of the construction process can be kept.

  Moreover, you may make it the information regarding construction quality contain the result which identified the multiple types of construction defect. In this way, even users with little knowledge or experience can objectively understand.

  In addition, poor construction includes subsidence cracks, temperature cracks, bubbles / flapping, beans, flow stripes / work stripes, cold joints, uneven color, formwork joint differences / steps, leakage from formwork joints, or surface peeling. You may make it. Specifically, the appearance characteristics of such poor construction may be learned and identified.

  In the imaging step, the entire concrete structure may be imaged, and in the evaluation step, the image generated in the imaging step may be divided into grid-like regions, and the construction quality may be evaluated for each region. In this way, it is possible to eliminate arbitraryness due to the selection of the shooting location of the concrete structure, and to detect a bias in construction failure in the entire concrete structure.

  The contents described in the means for solving the problems can be combined as much as possible without departing from the problems and technical ideas of the present invention. The contents of the means for solving the problems can be provided as a device such as a computer or a system including a plurality of devices, a method executed by the computer, or a program executed by the computer. A recording medium that holds the program may be provided.

  According to the present invention, construction quality can be objectively evaluated based on the appearance of concrete after demolding and the tacit knowledge of a concrete engineer.

It is a figure for demonstrating the outline | summary of construction quality evaluation. It is a figure for demonstrating the outline | summary of construction quality evaluation. It is a figure for demonstrating an evaluation item and the teacher value in machine learning. It is a functional block diagram which shows an example of an evaluation apparatus. It is an apparatus block diagram which shows an example of a computer. It is a processing flow figure showing an example of construction quality evaluation processing. It is a figure which shows the example which divides | segments image data into a grid | lattice form.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In addition, the following embodiment is an illustration and this invention is not limited to the following structure.

<Outline of construction quality evaluation>
1 and 2 are diagrams for explaining the outline of construction quality evaluation (also referred to as “evaluation processing”) according to the present embodiment. The concrete structure evaluation process according to the present embodiment is performed after demolding in the formwork method. The concrete structures are various concrete structures formed by a formwork method such as buildings, bridges, dams, tunnels, side walls of roads, retaining walls of coasts, and the like. The concrete structure according to the present embodiment is also referred to as “evaluation target”. As shown in FIG. 1, a user indicated by a broken line operates an evaluation device 1 including an imaging device, and images an evaluation object 2. FIG. 2 is a schematic perspective view of the evaluation object 2 viewed from the user's viewpoint via the evaluation device 1. The evaluation device 1 uses a model (“function”, “function”, and “model” generated by a predetermined machine learning method using photographic data of a concrete structure in which various construction defects appear and data in which a concrete engineer evaluates the type of construction failure. (Referred to as “reference information”) in advance. Note that machine learning performed in advance is also referred to as “learning processing”. And the evaluation apparatus 1 detects the construction defect which appears in the photographic data which imaged the evaluation object 2, using the said function, and displays the classification.

  In addition, the configuration of the system for performing the construction quality evaluation is such that the imaging device and the evaluation device are connected via a network such as the Internet, and the image data generated by the imaging device is transmitted to the evaluation device. You may make it return to arbitrary apparatuses, such as an imaging device. Further, the image data generated by the imaging device via a storage medium such as a memory card may be taken into the evaluation device.

  The above-described function is generated by performing supervised learning using, for example, a neural network using a plurality of combinations of image data of a concrete structure after demolding and evaluation values by a concrete engineer. Machine learning by a neural network uses a model having one or more intermediate layers between an input layer and an output layer, and an arbitrary number of nodes in each layer, so that the difference between the output value and the teacher value is small. Is generated by adjusting function parameters between nodes. For machine learning, for example, an existing technology such as so-called deep learning can be used.

  Further, it is assumed that the image data input in the machine learning according to the present embodiment is an image of the entire concrete structure. That is, since there is room for arbitraryness depending on the selection of a place for determining the presence or absence of construction failure, for example, it is preferable to image the entire concrete structure with as high a resolution as possible. Note that the imaged concrete structure may be divided into, for example, grid-like areas, and it may be determined whether or not construction defects have occurred in each area. In addition, since the accuracy of determination is improved when the image used for the learning process and the image used for the evaluation process are closer, for example, a reference object whose size is known in advance is imaged together with the concrete structure, and the evaluation process is performed. The scale may be corrected at.

  Evaluation items include “cracking cracks”, “temperature cracks”, “bubbles / abata”, “bean plate”, “flow stripes / construction stripes”, “cold joints”, “color unevenness”, “mold joint misalignment / Step ”,“ surface peeling ”and the like. FIG. 3 is a diagram illustrating an example of data serving as a teacher value for machine learning. For each piece of image data of a concrete structure in which a construction failure has occurred, a skilled engineer evaluates items as shown in FIG.

“Subsidence cracks” are cracks that occur along the steel bars when the concrete before solidification sinks. “Temperature cracks” are cracks generated by tensile stress generated by, for example, thermal contraction of concrete in an externally constrained state. "Bubble / Abata" is a material that hardens with air and bleeding water remaining on the concrete surface. The “bean plate” refers to a material in which coarse aggregates gather and voids are generated in the concrete. The “flow stripes / construction stripes” are traces generated when the ready-mixed concrete moves in the horizontal direction, for example, and sand bars caused by bleeding. “Cold joint” refers to a state of non-integration that occurs when concrete is piled up with an interval more than an appropriate interval. “Color unevenness” refers to a non-uniform color state that occurs on the concrete surface depending on the material and its composition, the material and state of the dam, the state of curing, and the like. The “mold joint difference / step” is a step generated at the joint of the mold. "Noro leakage from formwork joints" is uneven color and surface voids caused by the formation of gaps in the formwork joints that cause moisture, cement paste, or mortar in concrete to flow out of the formwork. . “Surface peeling” refers to a phenomenon in which the concrete surface loses its integrity with the interior and peels off. Note that not all of these items may be detected, or other items may be included.

  In the learning process, for each of the above construction failures, machine learning is performed using the photograph data of the concrete structure where the construction failure has occurred and the score indicating the degree of the construction failure evaluated by the concrete engineer. A function for calculating a likelihood indicating whether or not a construction failure appears in the image of the structure is generated. In the evaluation process of the concrete structure, the image data obtained by imaging the evaluation target concrete structure is input, the likelihood is calculated using the function generated for each construction failure, and the likelihood is lower than a predetermined threshold value. If it is larger, it is determined (detected) that the corresponding construction failure has occurred.

<Functional configuration>
FIG. 4 is a functional block diagram illustrating an example of the evaluation device 1 according to the present embodiment. The evaluation apparatus 1 includes an imaging unit 11, an image storage unit 12, an image reading unit 13, an image analysis unit (also referred to as “evaluation unit”) 14, a function storage unit 15, and a result output unit 16.

  The imaging unit 11 functions as a digital still camera, a digital video camera, or the like, and stores the generated image data in the image storage unit 12. The image storage unit 12 is formed by a main storage device that temporarily holds data or an auxiliary storage device that permanently stores data, and stores image data obtained by imaging the concrete structure 2. Note that image data may be captured from an imaging device as another user terminal to the image storage unit 12 via a portable storage medium such as a flash memory or an optical disk, or image data via a network. May be taken in.

  The image reading unit 13 reads image data to be processed from the image storage unit 12. The image analysis unit 14 analyzes the image data read by the image reading unit 13 using a predetermined function stored in advance in the function storage unit 15 and evaluates the construction quality of the captured concrete structure. The function storage unit 15 stores in advance a function in which a plurality of images of the concrete structure after demolding are input in association with the evaluation of the construction quality, and the appearance features of each construction failure are machine-learned. Shall. It should be noted that a function whose features are machine-learned can be generated using an existing technique as described above, and the generation method is not limited in this embodiment.

  The result output unit 16 outputs the evaluation of the construction quality of the concrete structure as the output of the evaluation device 1. The output destination may be on a display connected to the evaluation apparatus 1 or may be another computer connected via a network (not shown). Here, “sinking cracks”, “temperature cracks”, “bubbles / abata”, “bean plate”, “flow stripes / work stripes”, “cold joints”, “color unevenness”, “mold joint difference / step”, An evaluation value such as “surface peeling” is output.

  As described above, according to the evaluation device 1, by using a function reflecting tacit knowledge in image analysis, an evaluation result that can be objectively understood by a user who is not familiar with concrete can be output. Moreover, the presence or absence of occurrence can be determined for each of a plurality of construction defects.

<Device configuration>
FIG. 5 is an apparatus configuration diagram illustrating an example of a computer. The evaluation device 1 is a computer as shown in FIG. 5, for example. A computer 1000 illustrated in FIG. 5 includes a CPU (Central Processing Unit) 1001, a main storage device 1002, and an auxiliary storage device (external storage device) 100.
3, a communication IF (Interface) 1004, an input / output IF (Interface) 1005, a drive device 1006, and a communication bus 1007. The CPU 1001 performs processing and the like according to this embodiment by executing a program. The main storage device 1002 caches programs and data read by the CPU 1001 and develops a work area of the CPU. Specifically, the main storage device is a RAM (Random Access Memory), a ROM (Read Only Memory), or the like. The auxiliary storage device 1003 stores programs executed by the CPU 1001, setting information used in the present embodiment, and the like. Specifically, the auxiliary storage device 1003 is an HDD (Hard-disk Drive), an SSD (Solid State Drive), an eMMC (embedded Multi-Media Card), a flash memory, or the like. Main storage device 1002 and auxiliary storage device 10
03 functions as the image storage unit 12, the structure DB 104, the deterioration factor DB 106, and the feature DB 108 of the evaluation apparatus 1. The communication IF 1004 transmits / receives data to / from other computers. The communication IF 1004 is specifically a wired or wireless network card or the like. As described above, the evaluation device 1 may be configured to receive image data from the imaging device via a network. The input / output IF 1005 is connected to the input / output device and accepts input from the user or outputs information to the user. Specifically, the input / output device is a keyboard, a mouse, a display, a touch panel, or the like. The drive device 1006 reads data recorded on a storage medium such as a magnetic disk, a magneto-optical disk, and an optical disk, and writes data to the storage medium. The above components are connected by a communication bus 1007. A plurality of these components may be provided, or some of the components (for example, the communication IF 1004 and the drive device 1006) may not be provided. Further, the input / output device may be integrated with the computer. In addition, the program executed in this embodiment is provided via a portable storage medium readable by the drive device 1006, a portable auxiliary storage device 1003 such as a flash memory, a communication IF 1004, and the like. It may be. Then, when the CPU 1001 executes the program, the above-described computer is operated as the evaluation apparatus 1 shown in FIG. 4, for example. Specifically, the evaluation apparatus 1 can be configured by a smartphone, a slate PC, a laptop PC, or the like.

<Construction quality evaluation process>
FIG. 6 is a process flowchart showing an example of the construction quality evaluation process. The function storage unit 15 stores in advance a function in which a plurality of images of a concrete structure are input in association with the type and degree of construction failure and the features of each appearance are machine-learned. .

  First, the imaging unit 11 of the evaluation device 1 receives an operation of a user, for example, and images a concrete structure after demolding to be evaluated (FIG. 8: S1). In this step, as shown in FIGS. 1 and 2, for example, an entire image of the evaluation object 2 is captured.

  Next, the image reading unit 13 of the evaluation apparatus 1 reads image data from the image storage unit 12 (S2). In this step, the image data stored in the auxiliary storage device may be read out, or when the image data stored in the main storage device is analyzed in real time and the evaluation object 2 is detected, You may make it utilize in a process.

  Thereafter, the image analysis unit 14 of the evaluation apparatus 1 analyzes the read image data and evaluates the construction quality of the concrete structure (S3). In this step, a function generated in advance for each evaluation item is read from the function storage unit 15 and “sinking crack”, “temperature crack”, “bubble / abata”, “bean plate”, The likelihoods indicating the presence / absence of “flow stripes / construction stripes”, “cold joints”, “color unevenness”, “formula joint differences / steps”, “surface peeling”, etc. are obtained in order. These evaluation values are examples of information relating to the construction quality of the concrete structure according to the present embodiment, and are calculated by inputting the image data read out in S2 to a predetermined function. Moreover, when the calculated likelihood exceeds a predetermined threshold value, it is determined that a construction failure corresponding to the captured concrete structure has occurred.

  Then, the output part 16 of the evaluation apparatus 1 outputs the information regarding the construction quality of a concrete structure (S4). Specifically, the output unit 16 determines the “sinking crack”, “temperature crack”, “bubble / abata”, “bean plate”, “flow stripe / work stripe”, “cold joint”, “color” obtained in S3. Of the "unevenness", "mold joint misalignment / step difference", "surface peeling", etc., the detected construction failure is displayed. Further, in this step, instead of the presence or absence of construction failure or in combination with the presence or absence of construction failure, likelihood or a score corresponding to this may be output.

  Further, the image analysis unit 14 receives an evaluation value related to the construction quality of the concrete structure determined by the engineer, performs further machine learning, and updates the function stored in the function storage unit 15. Good (S5). The accuracy of construction quality evaluation according to this embodiment can be improved by learning a lot of image data. On the other hand, if the teacher value in machine learning is inappropriate, the accuracy of construction quality evaluation is lowered. Therefore, the process of feeding back the construction quality in this step is not necessarily performed.

  As described above, if an evaluation by a knowledgeable and experienced concrete engineer is learned in advance, it will occur in a concrete structure that has been imaged among multiple construction defects based on the image of the concrete structure after demolding. It can detect defective construction. That is, construction quality can be evaluated by outputting the result of identifying a plurality of types of construction failures. Moreover, since the presence or absence of construction failure is output, even a user with little knowledge and experience can understand objectively. And since mechanical judgment based on the unified standard can be performed, evaluation can be performed based on tacit knowledge of the concrete engineer, and variation in judgment by the engineer is reduced. The “implicit knowledge” of a concrete engineer is an evaluation standard that is difficult for verbalization, acquired by an experienced concrete engineer. Such an evaluation standard can be obtained as a function (model) used for evaluation by performing machine learning such as deep learning using a neural network. And the score which shows the evaluation result regarding construction quality and construction quality is obtained by inputting the image of the concrete structure of evaluation object into the function used for evaluation. Moreover, since the image of the concrete structure after demolding, that is, before finishing is used, the construction quality can be evaluated from the surface state, and an objective record of the construction process can be kept.

<Modification>
In S3 of FIG. 6, for example, a part where the concrete structure is imaged may be extracted from the image data based on the hue or the like in the image data, and the construction quality may be evaluated for the extracted part. Moreover, in S3 of FIG. 6, the imaged concrete structure may be divided into grid-like regions, and the presence or absence of construction defects may be detected for each region. FIG. 7 is a diagram illustrating an example in which image data is divided into grid-like regions. In the example of FIG. 7, an example in which the entire image data is divided is shown, but an area where the concrete structure 2 a is imaged may be extracted, and the extracted portion may be further divided. In this way, it is possible to know the approximate ratio and the bias of the portion with poor construction relative to the entire concrete structure.

  In addition, the above-described function may be generated using image data captured at a predetermined scale in the learning process, and a reference object whose size is determined may be captured together in S1 of FIG. In this case, in S3, the image to be evaluated is enlarged or reduced so that the scale of the image data matches the learning process, and the presence or absence of construction failure is detected. If it does in this way, the precision of construction quality evaluation will improve.

Further, it is also possible to learn features of the appearance using an image of a pattern or the like appearing on the concrete surface that is not defective in construction, and use it for determining whether or not it is defective in construction. For example, in learning processing, machine learning by a neural network is performed using images of normal formwork joints, induction joints, joints, P-con traces, etc., and form joints, induction joints, joints, P-con traces, etc. Generate a function to identify. And in an evaluation process, when a formwork joint, a induction joint, a joint, P contour etc. are identified from an image, it judges that it is not a construction defect. In this way, the accuracy of determining whether or not the construction is defective is improved. Note that the evaluation device 1 may output information indicating the identified formwork joints, induction joints, joints, P contour marks, and the like as information on construction quality.

  Moreover, you may make it the evaluation apparatus 1 output the grade of construction failure. For example, in the learning process, machine learning is performed using image data in which the degree of construction failure is classified into a plurality of stages such as minor, medium, and serious. On the other hand, in the evaluation process, based on not only the presence / absence of construction failure but also the magnitude of the likelihood calculated by the function, it is determined which of the plurality of stages corresponds.

  In the learning process, machine learning is performed using a teacher value that increases as the degree of construction failure increases. In the evaluation process, the likelihood calculated by the function is output as a score indicating the degree of construction failure. You may do it.

  Moreover, you may make it use a different function for every kind of concrete structure of evaluation object. For example, in the learning process, “sinking cracks”, “temperature cracks”, “bubbles / abata”, “bean plate”, “flow stripes / construction stripes”, “cold joints”, “color unevenness”, “mold joint misalignment / steps” ”,“ Surface peeling ”, and machine learning for each type of structure such as buildings and civil engineering structures, and generate different functions for each type of structure and each type of construction failure To do. And in an evaluation process, it is judged whether each construction defect has generate | occur | produced using the function corresponding to the structure of evaluation object. By generating a function for each type of structure, the workability evaluation system is improved. In addition, the function generated by machine learning is for each type and part of the structure, such as the wall of the building, the footing of the pier, the side wall of the dam, the lining concrete of the tunnel, the side wall of the road, the retaining wall of the coast, etc. And you may make it produce | generate for every kind of construction defect mentioned above.

  Moreover, you may make it implement each modification mentioned above combining as much as possible in the range which does not deviate from the subject and technical idea of this invention.

<Others>
The present invention includes a computer program that executes the above-described processing and a computer-readable recording medium that records the program. The recording medium on which the program is recorded can perform the above-described processing by causing the computer to execute the program.

  Here, the computer-readable recording medium refers to a recording medium in which information such as data and programs is accumulated by electrical, magnetic, optical, mechanical, or chemical action and can be read from a computer. Examples of such a recording medium that can be removed from the computer include a flexible disk, a magneto-optical disk, an optical disk, a magnetic tape, and a memory card. In addition, examples of the recording medium fixed to the computer include a hard disk drive and a ROM.

DESCRIPTION OF SYMBOLS 1 Evaluation apparatus 11 Image pick-up part 12 Image storage part 13 Image reading part 14 Image analysis part 15 Function storage part 16 Output part 2 Concrete structure (evaluation object)

Claims (7)

An imaging step of imaging the concrete structure after demolding formed by the mold method and generating image data;
Imaging using the image of the concrete structure after demolding, the reference information that has been previously learned by associating the information indicating the evaluation of the construction quality of the concrete structure, and the image data generated in the imaging step An evaluation step of evaluating the construction quality of the concrete structure that has been made and outputting information on the construction quality;
Is a construction quality evaluation program that runs a computer. The construction quality evaluation program according to claim 1, wherein the information on the construction quality includes a result of identifying a plurality of kinds of construction failures.   The construction quality evaluation according to claim 2, wherein the poor construction includes subsidence cracking, temperature cracking, air bubbles / flapping, beans board, flow stripes / work stripes, cold joints, color unevenness, form joint misalignment / step difference, or surface peeling. program. In the imaging step, the entire concrete structure is imaged,
The construction quality evaluation program according to any one of claims 1 to 3, wherein, in the evaluation step, the image generated in the imaging step is divided into grid-like regions and the construction quality is evaluated for each region. The reference information is generated in advance for each type of concrete structure,
In the evaluation step, different reference information is used depending on the type of the concrete structure imaged in the imaging step.
The construction quality evaluation program according to any one of claims 1 to 4. An imaging step of imaging the concrete structure after demolding formed by the mold method and generating image data;
Imaging using the image of the concrete structure after demolding, the reference information that has been previously learned by associating the information indicating the evaluation of the construction quality of the concrete structure, and the image data generated in the imaging step An evaluation step of evaluating the construction quality of the concrete structure that has been made and outputting information on the construction quality;
The construction quality evaluation method that the computer executes. An imaging unit that images a concrete structure after demolding formed by a mold method, and generates image data;
Imaging using the image of the concrete structure after demolding, the reference information that is pre-machine-learned by associating the information indicating the evaluation regarding the construction quality of the concrete structure, and the image data generated by the imaging unit An evaluation unit that evaluates the construction quality of the concrete structure and outputs information on the construction quality;
Construction quality evaluation device with

Images