JP7643943B2 - Steel fiber dispersion ratio estimation system - Google Patents

Description

The present invention relates to a steel fiber dispersion rate estimation system.

Conventionally, when creating concrete structures, steel fiber reinforced concrete is used in which 20 kg/m3 to 80 kg/m3 of steel fibers (e.g., with a diameter of about 0.62 mm to 0.75 mm and a length of about 30 mm to 60 mm) are mixed into the concrete in order to improve the brittle fracture of the concrete.
The above-mentioned steel fibers are generally formed into a bundle (hereinafter referred to as a steel fiber bundle) by fixing multiple fibers (e.g., 36 fibers) with an adhesive in order to facilitate handling and to prevent the steel fibers from becoming entangled and forming a mass of fiber balls.

Steel fiber reinforced concrete is composed of materials such as water, cement, aggregates (coarse aggregate, fine aggregate), and steel fibers.
For this reason, by pouring the above-mentioned steel fiber bundles into concrete and stirring it in a cylindrical container mounted on a concrete mixer or an agitator truck, the water-soluble adhesive dissolves and the steel fibers in the steel fiber bundles are separated by impact from aggregate, etc., and the steel fibers are dispersed in the concrete (see, for example, Patent Document 1).

However, depending on the quality of the steel fiber bundles (e.g., the quality of the adhesive) or the length of time of stirring, the adhesive may not dissolve sufficiently, leaving unseparated steel fiber bundles, resulting in incomplete dispersion of the steel fibers.
However, since there is no standard method for quantitatively evaluating the dispersion state of steel fibers, the dispersion rate of steel fibers is quantitatively determined using the wash analysis test method stipulated in JSCE (The Japan Society of Civil Engineers)-F554 (Draft test method for steel fiber content in transition reinforced concrete), which is a method for evaluating the content rate of steel fibers.

That is, steel fiber reinforced concrete containing steel fibers is placed in a metallic cylindrical container having a capacity of 6 L (liters) or more, and the steel fibers are separated from the concrete while washing the steel fiber reinforced concrete with water.
The steel fibres separated from the concrete are collected from the cylindrical vessel and dried, and the total mass V _A of the entire steel fibres is measured.

In addition, separated steel fibers, which are individually separated (single) steel fibers, and non-separated steel fiber groups, which are bundles of steel fibers in which two or more fibers are fixed together, are sorted, and the total mass _VD of the separated steel fibers is measured.
Then, the total mass V _A and the total mass V _D are substituted into the following formula (1) to determine the dispersion rate F _D (%) of the steel fibers.
F _D = (V _D /V _A )×100...(1)

JP 2011-46175 A

As described above, the quality of steel fiber reinforced concrete can be controlled by calculating the dispersion rate from the ratio of the total mass V _A of the steel fibers separated from the concrete to the total mass V _D of the separated steel fibers in the steel fibers.
However, workers must perform tasks such as drying the steel fibers separated from the concrete, measuring the total mass V _A of the steel fibers, separating them into separated steel fibers, which are individual steel fibers, and non-separated steel fibers, which remain bonded to other steel fibers, and measuring the total weight V _D of the separated steel fibers.
In particular, the process of sorting the steel fibers separated from the concrete into separated steel fibers and non-separated steel fibers takes a lot of time, making it impossible to quickly determine the dispersion rate and quickly evaluate the dispersion state of the steel fibers, and reducing the work efficiency of workers.

The present invention was made in consideration of the above circumstances, and aims to provide a steel fiber dispersion rate estimation system that can easily measure the dispersion rate of steel fiber bundles mixed into concrete when producing steel fiber reinforced concrete, thereby reducing the burden on the worker who measures the dispersion rate compared to conventional methods, and determining the dispersion rate in real time (in a shorter time compared to conventional methods), thereby enabling faster evaluation of the state of dispersion of steel fibers.

In order to solve the above-mentioned problems, the steel fiber dispersion rate estimation system of the present invention is a system that estimates a dispersion rate indicating the proportion of steel fibers mixed in when producing concrete that are separated from the steel fiber bundles to which each of the steel fibers is bonded, and is characterized in that it comprises: a shape feature extraction unit that extracts shape feature values calculated from at least the contour length, contour area, circularity, compactness ratio, and unevenness coefficient of the steel fibers as feature values of each of the steel fibers captured from a steel fiber captured image, which is an image obtained by mixing the steel fiber bundles into concrete, performing a predetermined mixing process, and capturing the steel fibers after washing the concrete; a color feature extraction unit that extracts color feature values calculated from at least the gradient, hue, saturation, and brightness of the color components of the steel fibers as feature values of each of the captured steel fibers from the steel fiber captured image; and a dispersion rate estimation unit that estimates the dispersion rate of the steel fibers using a predetermined machine learning model by inputting each of the shape feature values and the color feature values.

According to the present invention, when producing steel fiber reinforced concrete, it is possible to provide a steel fiber dispersion rate estimation system that can easily measure the dispersion rate of steel fiber bundles mixed into concrete, thereby reducing the burden on the worker who measures the dispersion rate compared to conventional methods and enabling faster evaluation of the state of dispersion of steel fibers.

1 is a block diagram showing an example of the configuration of a steel fiber dispersion ratio estimation system according to an embodiment of the present invention. FIG. FIG. 2 is a diagram illustrating the capture of a steel fiber image in this embodiment. 4 is a diagram showing an example of the configuration of a feature table written and stored in a feature storage unit 19 in the present embodiment. FIG. FIG. 13 is a diagram showing an example of the configuration of a table showing the correspondence between captured image identification information and the dispersion rate of steel fibers. 10 is a flowchart showing an example of the operation of a process in which the steel fiber dispersion ratio estimation system according to this embodiment classifies the dispersion ratio of steel fibers whose dispersion ratio is unknown into each of known dispersion ratios (labels).

Figure 1 is a block diagram showing an example of the configuration of a steel fiber dispersion rate estimation system according to one embodiment of the present invention. In Figure 1, the steel fiber dispersion rate estimation system 10 includes a data input/output unit 11, an image preprocessing unit 12, a shape feature extraction unit 13, a color feature extraction unit 14, a machine learning model generation unit 15, a dispersion rate estimation unit 16, a feature importance evaluation unit 17, an image data storage unit 18, a feature storage unit 19, a machine learning model storage unit 20, and a teacher data storage unit 21.

The data input/output unit 11 transmits and receives data to and from external devices. The data input/output unit 11 also writes and stores the steel fiber image data, which is an image of steel fiber provided from an external device, in the image data storage unit 18 after adding image identification information. The steel fiber image is captured in an environment where the light quantity and light color (light wavelength characteristics) are constant, such as a dark box that blocks external light.

FIG. 2 is a diagram for explaining the capturing of a steel fiber image in this embodiment.
FIG. 2(a) is an image showing an example of a steel fiber bundle in which a plurality of steel fibers are connected to each other with an adhesive to form a bundle.
The steel fiber bundles are added to concrete and mixed for a predetermined period of time to produce steel fiber reinforced concrete for constructing a concrete structure.
The steel fiber reinforced concrete containing the steel fibers is placed in a metallic cylindrical container, and the steel fiber reinforced concrete is washed with water to wash away the concrete and leave only the steel fibers. The steel fibers from which the concrete has been washed away are collected from the cylindrical container and spread out on a plate of a specified area to dry, and an image of the spread steel fibers is taken.

Each of Figures 2(b) and 2(c) is an example of a steel fiber image of steel fibers (separated steel fibers and non-separated steel fiber groups) spread across the plate. For example, the dispersion rate in Figure 2(b) is 60%, and in Figure 2(c) the dispersion rate is 100%. The captured image is then trimmed so that each steel fiber is captured in an image area, and the image area obtained by this trimming is used as the steel fiber image to be used as test data or training data.

Figure 2 (d) shows an example of a method for capturing steel fiber images. A plate 201 with steel fibers spread out (preferably with a white surface) is placed in a container 200, such as a dark box that blocks external light. The entire spread steel fiber is then captured by the imaging device 210 at the timing when the flash 211 is fired. This allows the imaging environment for capturing the steel fiber images to be constant, and multiple steel fiber images can be captured under the same conditions. Furthermore, in this embodiment, the camera parameters of the imaging device when capturing each steel fiber image are also the same.

Returning to FIG. 1, the captured image pre-processing unit 12 performs the trimming process described above to obtain a steel fiber captured image, and also performs processes such as removing the shadows of other steel fibers in each image area of the steel fibers in the steel fiber captured image. The captured image pre-processing unit 12 then writes and stores the processed steel fiber captured image in the image data storage unit 18 in association with the captured image identification information.

The shape feature extraction unit 13 performs processing to extract the contours of each of the steel fibers (separated steel fibers and non-separated steel fiber groups) in the steel fiber captured image.
Then, shape feature extraction unit 13 calculates the outer periphery of each of the extracted contours as the contour length, calculates the average value of the contour lengths of each of the extracted contours as the contour length average value, calculates the standard deviation of the contour lengths of each of the contours as the contour length standard deviation, and writes and stores them in feature memory unit 19.
Similarly, the shape feature extraction unit 13 calculates the internal area of the extracted contour as the contour area, calculates the average value of the contour areas of each of the extracted contours as the contour area average value, calculates the standard deviation of the contour areas of each of the contours as the contour area standard deviation, and writes and stores them in the feature memory unit 19.

The shape feature extraction unit 13 calculates a contour circularity, which is the circularity of the shape of the contour line of the extracted contour, calculates the average value of the contour circularity of each of the extracted contours as the contour circularity average value, calculates the standard deviation of the contour circularity of each of the contours as the contour circularity standard deviation, and writes and stores them in the feature memory unit 19.
Here, the shape feature extraction unit 13 calculates the circularity circ of each contour line of the steel fiber by the following formula (2): In this formula (2), area is the contour area of the steel fiber, and length is the contour length of the steel fiber.
circ=4π×area/(length) ² …(2)

The shape feature extraction unit 13 calculates a compactness ratio as a degree of unity in the shape of the contour line of the extracted contour (degree of aspect ratio of the shape), calculates the average value of the above compactness ratios for each extracted contour as the compactness ratio average value, calculates the standard deviation of the compactness ratios for each contour as the compactness ratio standard deviation, and writes and stores them in the feature memory unit 19.
Here, the shape feature extraction unit 13 calculates the compact ratio compc of each contour line of the steel fiber by the following formula (3): In this formula (3), area is the contour area of the steel fiber, and length is the contour length of the steel fiber.
compc=length/(2π×area) ^1/2 …(3)

The shape feature extraction unit 13 calculates a roughness coefficient as a degree of smoothness in the shape of the contour line of the extracted contour, calculates the average value of the roughness coefficients of each of the extracted contours to obtain the roughness coefficient average value, calculates the standard deviation of the roughness coefficients of each of the contours to obtain the roughness coefficient standard deviation, and writes and stores them in the feature memory unit 19.
Here, the shape feature extraction unit 13 calculates the unevenness coefficient unev of each contour line of the steel fiber by the following formula (4): In this formula (4), area is the contour area of the steel fiber, and length is the contour length of the steel fiber.
unev=length ² /(4π×area)…(4)

The color feature extraction unit 14 extracts color information for each pixel in the steel fiber image and obtains the color feature values of the steel fiber image. For example, the color information is calculated by calculating the average gradients of the R, G, and B color components of each pixel in the steel fiber image, and obtaining the R average value, G average value, and B average value. At this time, the color feature extraction unit 14 calculates the grayscale from the R, G, and B color components for each pixel in the steel fiber image and obtains the grayscale average value, which is the average value of the grayscale. The color feature extraction unit 14 then writes and stores the obtained R average value, G average value, and B average value in the feature storage unit 19.

For color information, the gradients of the color components R, G, B and grayscale are each divided into multiple ranges, and for example, in the case of 256 gradations, they are classified into eight classes: 0-31, 32-63, 64-95, 96-127, 128-159, 160-191, 192-223, and 224-255. The color feature extraction unit 14 then divides the number of pixels contained in each class by the total number of pixels in the steel fiber image, and the division results for each class are regarded as the occurrence ratio of each class.

For example, the color feature extraction unit 14 divides the number of pixels in the eight classes of the color component R by the total number of pixels in the steel fiber image, and obtains the division values of the classes of the gradient from 0-31 to 224-255, and obtains the occurrence ratio of each class of the gradient of the color component R. The color feature extraction unit 14 also performs a process of standardizing the occurrence ratio so that the maximum and minimum occurrence ratio values in the eight classes of the color component R are 1 and 0, respectively, and writes and stores each of the obtained occurrence ratios of the color component R (the occurrence ratio of the color component R in each of the eight classes) in the feature storage unit 19. The color feature extraction unit 14 also performs the same process as for the color component R for each of the color components G, B, and grayscale, and writes and stores each of the obtained occurrence ratios of the color component G, the color component B, and the grayscale occurrence ratio in the feature storage unit 19.

The color feature extraction unit 14 also calculates the average values of hue H, saturation S, and brightness V in the HSV display for each pixel of the steel fiber image as other color feature quantities, and obtains the average hue, average saturation, and average brightness values.The color feature extraction unit 14 then writes and stores the obtained average hue, average saturation, and average brightness values in the feature storage unit 19.The color feature extraction unit 14 divides the range of the values of hue H, saturation S, and brightness V into eight classes, similar to the gradient, and classifies each pixel into each class.

For example, the color feature extraction unit 14 converts the values of the hue scale 0 to 359 indicating the hue H into 256 gradations from 0 to 255 and classifies them into 8 classes, and as in the case of gradations, classifies each pixel according to the hue H, and divides the number of pixels classified into each class by the total number of pixels in the steel fiber image. The color feature extraction unit 14 then performs a standardization process so that the maximum and minimum occurrence rates in the hue H class are 1 and 0, respectively, to obtain a standardized hue occurrence rate. The color feature extraction unit 14 writes each of the acquired hue occurrence rates (the hue occurrence rate in each of the 8 classes) into the feature storage unit 19 and stores them.

Similarly, the color feature extraction unit 14 converts the values from 0% to 100% indicating the saturation S into 256 gradations from 0 to 255 and classifies them into eight classes, and as in the case of gradations, classifies each pixel according to its saturation S, and divides the number of pixels classified into each class by the total number of pixels in the steel fiber image. The color feature extraction unit 14 then performs a standardization process so that the maximum and minimum occurrence rates in the saturation S class are 1 and 0, respectively, to obtain a standardized saturation occurrence rate. The color feature extraction unit 14 writes each of the acquired saturation occurrence rates (the saturation occurrence rate for each of the eight classes) into the feature storage unit 19 and stores them.

The color feature extraction unit 14 also converts the values from 0% to 100% indicating the brightness V into 256 gradations from 0 to 255 and classifies them into eight classes, and as in the case of gradations, classifies each pixel according to its brightness V, and divides the number of pixels classified into each class by the total number of pixels in the steel fiber image. The color feature extraction unit 14 then performs a standardization process so that the maximum and minimum occurrence rates in the brightness V class are 1 and 0, respectively, to obtain a standardized brightness occurrence rate. The color feature extraction unit 14 writes each of the acquired brightness occurrence rates (the brightness occurrence rate for each of the eight classes) into the feature storage unit 19 and stores them.

Figure 3 is a diagram showing an example of the configuration of a feature table written and stored in the feature storage unit 19 in this embodiment. Figure 3(a) shows an example of the configuration of a shape feature table, which is one of the feature tables. For each record, the shape feature table has columns for captured image identification information, and a feature group of shape features, including the above-mentioned contour length average value, contour length standard deviation, contour area average value, contour area standard deviation, contour circularity total value, contour circularity average value, contour circularity standard deviation, compact ratio average value, compact ratio standard deviation, unevenness coefficient average value, and unevenness coefficient standard deviation.

Figure 3 (b) shows an example of the configuration of a color feature table, which is one of the feature tables. The color feature table has columns for each record, including the captured image identification information and a feature group of color features, namely, the above-mentioned R average value, color component R appearance ratio, G average value, color component G appearance ratio, B average value, color component B appearance ratio, grayscale average value, grayscale appearance ratio, hue average value, hue appearance ratio, saturation average value, saturation appearance ratio, lightness average value, and lightness appearance ratio. In addition, each of the color component R appearance ratio, color component G appearance ratio, color component B appearance ratio, grayscale appearance ratio, hue appearance ratio, saturation appearance ratio, and lightness appearance ratio is written in one column as a set of eight appearance ratio values for each class.

Returning to FIG. 1, the machine learning model generation unit 15 generates a machine learning model that outputs an estimate of the classification result of the dispersion rate (hereinafter simply referred to as the dispersion rate) indicating the proportion of steel fibers separated from the steel fiber bundle when each of the above-mentioned shape feature and color feature is input. That is, the machine learning model in this embodiment is a classifier that classifies multiple dispersion rates, and by inputting a group of feature data of each of the shape feature and color feature, outputs a confidence level (an estimate value indicating the likelihood, an estimate value p described later) when the dispersion rate of the steel fibers captured in the steel fiber image is classified into each label. For example, in this embodiment, three types of particle size distributions, particle size distributions #1, #2, and #3, are used as labels. Therefore, the machine learning model outputs a confidence level for each label of particle size distributions #1, #2, and #3, and it is estimated that the particle size distribution with the highest confidence level output is the most likely to be the dispersion rate of the steel fibers captured in the steel fiber image.

The machine learning model generation unit 15 captures steel fiber images of steel fibers with known dispersion rates and generates a data group for training the machine learning model. In this embodiment, for example, a multilayer perceptron classifier is used as the machine learning model, but any other classifier using a neural network or the like may be used. In this embodiment, the data group is divided at a predetermined ratio, and for example, 80% of the data group is used as a teacher data group for learning (training) the machine learning model. The remaining 20% of the data group is then used as test data for determining the accuracy rate of the created machine learning model.

The machine learning model generation unit 15 also divides the teacher data group into a training data group and a validation data group at a predetermined ratio, and for example, each of 80% of the teacher data group is used as a training data group to be used for actual learning (training) of the machine learning model. Then, the machine learning model generation unit 15 sets each of the remaining 20% of the teacher data group as a validation data group to be used for verifying the accuracy rate of the machine learning model trained with the data of the training data group. Here, the machine learning model generation unit 15 extracts a portion of data at a predetermined ratio (20%) from the training data group, replaces it with the data of the validation data group, trains the machine learning model with the data of the new training data group after the replacement, and uses the data of the new validation data group to verify the accuracy rate of the machine learning model. Then, the machine learning model generation unit 15 performs the data replacement operation in each of the above-mentioned training data group and validation data group a predetermined number of times to train the machine learning model.

The features used as the teacher data are written and stored as a teacher data table in the teacher data storage unit 21. This teacher data table is a table with the same configuration as the shape feature table and color feature table shown in FIG. 3. The teacher data table is written and stored with each of the shape feature and color feature determined by analyzing the teacher data image, which is a steel fiber image of steel fibers with a known dispersion rate.

The feature importance evaluation unit 17 selects either a shape feature or a color feature in the training data used to train the trained model, and evaluates the impact on the confidence level when the data of that feature is swapped between steel fiber image images.
For example, data on the contour length in the shape feature is randomly swapped (shuffled) between each piece of captured image identification information, and the swapped feature set is input to the trained model, and the degree to which the contour length is important as a feature in training the machine learning model is determined based on the level of confidence reduction. By swapping the feature data for all features, the importance of the feature in training the machine learning model can be confirmed. For example, using this result, it is possible to compact the machine learning model by not using features with confidence reduction values of a predetermined percentage or less relative to the largest confidence reduction value for training the machine learning model.

Figure 4 is a diagram showing an example of the configuration of a table showing the correspondence between the captured image identification information and the dispersion rate of steel fibers. Figure 4 (a) shows an example of the configuration of the teacher data information table in the teacher data storage unit 21. The teacher data information table has columns for the captured image identification information, the particle size distribution type, and the use type for each record. This captured image identification information is identification information that individually identifies each steel fiber captured image of steel fibers with a known dispersion rate used as teacher data. The particle size distribution indicates the type of dispersion rate of the steel fiber that captured the steel fiber captured image (the type of dispersion rate that serves as a label when learning). In this embodiment, for example, three types of dispersion rates are used, which are shown as dispersion rate #1, dispersion rate #2, and dispersion rate #3. The use type indicates whether each feature amount of the steel fiber captured image, which is the teacher data, is used as data of the learning data group or the verification data group in the teacher data group, or as test data.

4B shows an example of the configuration of the estimation result table in the feature storage unit 19. The estimation result table has columns for image identification information, dispersion rate #1, dispersion rate #2, and dispersion rate #3 for each record. This image identification information is identification information that individually identifies each steel fiber image in which the dispersion rate of the steel fiber is unknown. The dispersion rate #1 column shows an estimate value p (e.g., 0<p<1) indicating the confidence (likelihood) that the dispersion rate of the steel fiber is classified into the label dispersion rate #1, which is estimated by the machine learning model after the feature value of the steel fiber image in which the dispersion rate is unknown is input. Similarly, the dispersion rate #2 column shows an estimate value p indicating the confidence that the dispersion rate of the steel fiber is classified into the label dispersion rate #2, which is estimated by the machine learning model after the feature value of the steel fiber in which the dispersion rate is unknown is input. The dispersion rate #3 column shows an estimate value p indicating the confidence that the dispersion rate of the steel fiber is classified into the label dispersion rate #3, which is estimated by the machine learning model after the feature value of the steel fiber in which the dispersion rate is unknown is input.

Figure 4 is a flowchart showing an example of the operation of the process of classifying the dispersion rate of steel fibers with unknown dispersion rates into each of the known dispersion rates (labels) by the steel fiber dispersion rate estimation system according to this embodiment. In this embodiment, as an example, a process of generating a machine learning model (classifier) that classifies steel fibers with unknown dispersion rates into these three types of dispersion rates (labels) when there are three types of known particle size distributions, and a classification process of classifying steel fibers with unknown dispersion rates into three types of labels by the generated machine learning model will be described.

The user arranges (spreads and piles) steel fibers with a known dispersion rate on a white surface of a plate of a specified area, and in a closed space that blocks external light, uses an imaging device to capture a steel fiber image (a captured image of the fibers) of the arranged steel fibers to be used as training data (step S1). The captured image preprocessing unit 12 then performs preprocessing such as trimming the captured steel fiber image and removing shadows.

The shape feature extraction unit 13 and the color feature extraction unit 14 extract shape features and color features (feature data groups) from each of a plurality of steel fiber images, e.g., three types of steel fibers with known dispersion rates, and write and store them in the teacher data table in the teacher data storage unit 21. Here, for example, several hundred steel fiber images are taken for each of the three types of steel fibers with known dispersion rates, and are obtained as feature data groups of teacher data (step S2).
In addition, the machine learning model generation unit 15 sets 80% of the feature data group in the teacher data table in the teacher data storage unit 21 as training data and 20% as test data, and enters them in the purpose type column of the teacher data table (steps S3 and S4).

The machine learning model generation unit 15 regards 80% (i.e., 64% of the data group) of the feature data group (i.e., the teacher data group) of the teacher data table in the teacher data storage unit 21 as a learning data group, and 20% (i.e., 16% of the data group) of the data as a validation data group, and enters these in the purpose type column of the teacher data table (steps S5 and S6).
Then, the machine learning model generation unit 15 performs learning by adjusting the parameters of the machine learning model (classifier) using data whose use type is the learning data group in the teacher data table in the teacher data storage unit 21 (steps S7, S8, and S9). At this time, the machine learning model generation unit 15 uses data whose use type is the verification data group in the teacher data table in the teacher data storage unit 21 to calculate the confidence of the classification of the machine learning model to be learned as a correct answer rate (model verification, correct/incorrect judgment). In addition, the machine learning model generation unit 15 exchanges the data of the verification data group with a part of the data of the learning data group (for example, 20% of the learning data group), and causes the generated machine learning model to be re-trained using the data of the new learning data group after the exchange. The re-training (re-adjustment of parameters) of the machine learning model in which the data of the verification data group and the data of the part of the learning data group are exchanged is performed a preset number of times, and a trained machine learning model (trained model) for estimating the dispersion rate of steel fibers (fiber dispersion rate) is generated, and written and stored in the machine learning model storage unit 20.

The machine learning model generation unit 15 also uses data with a usage type of test data in the teacher data table in the teacher data storage unit 21 to determine the confidence (correct/incorrect judgment) of the classification of the machine learning model to be trained for the trained model of variance rate estimation (trained machine learning model) as the accuracy rate (steps S10 and S11). Here, if the accuracy rate of the trained model is low, the machine learning model generation unit 15 may be configured to return to step S3 and train the machine learning model again.

The user places (spreads and piles) steel fibers with an unknown dispersion rate on a white-surfaced plate, and in a closed space that blocks external light, an image of the placed steel fibers is taken with an imaging device (step S12). Then, the image preprocessing unit 12 performs preprocessing such as trimming the captured steel fiber image and removing shadows.

The shape feature extraction unit 13 and the color feature extraction unit 14 extract shape features and color features from each of the steel fiber images of steel fibers with unknown dispersion rates, and write and store the feature data group in a feature table in the feature memory unit 19 in correspondence with the assigned captured image identification information (step S13).
The variance rate estimation unit 160 reads the trained model from the machine learning model storage unit 20, and also reads a feature data group from the feature table of the feature storage unit 19. Then, the variance rate estimation unit 16 inputs the feature data group to the trained model, and writes and stores the confidence of each output label in the label column in the result table of the feature storage unit 19 (steps S14 and S15).

According to this embodiment, by inputting each of the shape features and color features extracted from a steel fiber image, which is an image of deposited steel fibers, into a trained model that has been trained using shape features and color features extracted from a steel fiber image of steel fibers with a known dispersion rate, it is possible to create concrete without the need for a large-scale dispersion rate measuring device as in the past, reduce the burden on workers who measure the dispersion rate of the mixed steel fibers, and easily determine the dispersion rate of the steel fibers.
Furthermore, according to this embodiment, it is possible to easily estimate the particle size distribution of steel fibers at construction sites or at businesses that manufacture ready-mixed concrete, as described above, so that it is possible to easily secure (provide) particle size distribution data for quality control in the construction of concrete structures and for traceability in the construction of concrete structures.

The steel fiber dispersion rate estimation process may be performed by recording a program for implementing the functions of the steel fiber dispersion rate estimation system 10 of FIG. 1 in the present invention on a computer-readable recording medium, and reading and executing the program recorded on the recording medium into a computer system. The term "computer system" as used herein includes hardware such as an OS (Operating System) and peripheral devices. The term "computer system" also includes a WWW (World Wide Web) system equipped with a homepage providing environment (or display environment). The term "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs (Read Only Memory), and CD-ROMs (Compact Disc - Read Only Memory), as well as storage devices such as hard disks built into computer systems. The term "computer-readable recording medium" also refers to storage devices that hold a program for a certain period of time, such as volatile memory (RAM (Random Access Memory)) inside a computer system that becomes a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line.

The above program may also be transmitted from a computer system in which the program is stored in a storage device or the like to another computer system via a transmission medium, or by transmission waves in the transmission medium. Here, the "transmission medium" that transmits the program refers to a medium that has the function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The above program may also be one that realizes part of the above-mentioned functions. Furthermore, it may be a so-called difference file (difference program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

10...Steel fiber dispersion rate estimation system 11...Data input/output unit 12...Captured image preprocessing unit 13...Shape feature extraction unit 14...Color feature extraction unit 15...Machine learning model generation unit 16...Dispersion rate estimation unit 17...Feature quantity demand evaluation unit 18...Image data storage unit 19...Feature quantity storage unit 20...Machine learning model storage unit 21...Teacher data storage unit

Claims (6)

A steel fiber dispersion rate estimation system for estimating a dispersion rate indicating a rate at which steel fibers mixed in concrete are separated from the steel fiber bundles to which each of the steel fibers is bonded,
a shape feature extraction unit which extracts shape feature quantities calculated from at least a contour length, a contour area, a circularity, a compactness ratio, and a roughness coefficient of each of the steel fibers as feature quantities of the steel fibers captured from a steel fiber captured image which is an image obtained by mixing the steel fiber bundles into concrete, performing a predetermined mixing process, and capturing the steel fibers after washing the concrete;
a color feature extraction unit for extracting color feature values calculated from at least the gradation, hue, saturation, and brightness of the color components of the steel fibers as feature values of each of the captured steel fibers from the steel fiber captured image;
a dispersion rate estimation unit that estimates the dispersion rate of the steel fibers using a predetermined machine learning model by inputting each of the shape feature and the color feature. The shape feature extraction unit,
The sum, average and standard deviation of the feature amounts of the contour length, the contour area and the circularity of each of the steel fibers in the steel fiber image are calculated as the shape feature amount ,
The color feature extraction unit,
The steel fiber dispersion rate estimation system according to claim 1, characterized in that the feature amounts of the gradient, the hue, the saturation, and the brightness of the color components of each of the steel fibers in the steel fiber image are classified into a plurality of ranges of feature amount size , and the occurrence rate of each of the ranges is calculated to be the color feature amount. The steel fiber dispersion rate estimation system according to claim 2, characterized in that the color components are each a color component R, a color component G, and a color component B, and a gray scale calculated from each of the color components R, G, and B. The steel fiber dispersion ratio estimation system according to any one of claims 1 to 3, further comprising a machine learning model generation unit that trains the machine learning model using, as training data, each of the shape feature and the color feature extracted from the steel fiber image in which a steel fiber having a known dispersion ratio is imaged. The machine learning model generation unit:
The steel fiber dispersion ratio estimation system according to claim 4, characterized in that the teacher data is divided into learning data and verification data at a predetermined ratio, the machine learning model is trained with the learning data, and a learning process for determining the accuracy rate of the estimation result of the machine learning model using the verification data is performed a predetermined number of times by replacing a part of the learning data with the verification data. The steel fiber dispersion ratio estimation system according to any one of claims 1 to 5, further comprising a feature importance evaluation unit that, in training the machine learning model, evaluates the contribution of each of the shape feature and the color feature to a training result of the machine learning model using a sorting importance technique.

Images