WO2023063059A1

WO2023063059A1 - Gravel production management method and computer program for gravel production management

Info

Publication number: WO2023063059A1
Application number: PCT/JP2022/035631
Authority: WO
Inventors: 寛孝北爪; 史朗古屋; 友裕岸部
Original assignee: 株式会社ヤマサ
Priority date: 2021-10-13
Filing date: 2022-09-26
Publication date: 2023-04-20
Also published as: JP2023058205A; JP7012402B1

Abstract

An edge device (31) in a gravel production management system (1) for confirmation/determination/notification of the quality of ore in accordance with this gravel production management method acquires a captured image (11) of input ore (9) in a gravel plant (2), extracts individual ores from the captured image (11) using an ore detection model (34), and encloses each of the extracted ores in a bounding box (12). A management server (51) in a management office (4) calculates the size (area) of each bounding box (12), compares the calculated size (area) with a threshold value to classify the ores into small-size ores (9S) and large-size ores (9L), calculates the number thereof, and displays the bounding boxes (12), sizes, and a mixture ratio of large and small sizes on the captured image (11) displayed on a display screen. Quality management of the input ore (9) can be performed precisely in real time.

Description

Gravel production control method and computer program for gravel production control

The present invention provides an unmanned gravel production management method for improving gravel productivity by grasping the quality of raw stones (size, mixing ratio of large and small sizes, etc.) in a gravel factory, and the method comprising a computer. The present invention relates to a computer program for gravel production management used to execute a management system.

In general, rough stones mined at a quarry are crushed, sorted, and washed with various machines at a gravel factory, conformed to the prescribed particle size standards, shipped, and used as aggregates for concrete, asphalt, etc. If the gravel used for concrete and asphalt contains non-standard size or foreign matter, it will not pass the acceptance inspection at the delivery destination. Assuming limestone mining, concrete manufacturing, etc., methods for removing non-standard large-sized stones, aggregates, foreign substances, etc. without relying on human labor are proposed in Patent Documents 1 and 2, for example.

In the aggregate sorting apparatus described in Patent Document 1, aggregates after sieving are photographed with a CCD camera, and the photographed images are image-analyzed to detect large-sized aggregates greatly deviating from the standard. aggregates are removed from other aggregates. This removes elongated aggregates that have passed through the sieve and are larger than the standard size. In the method for detecting foreign matter described in Patent Document 2, the image of the aggregate being conveyed is processed so that not only large stone blocks that deviate greatly from the standard size, but also foreign matter can be detected and removed. ing.

On the other hand, in recent years, with the development of AI technology, various machine learning models have been proposed and widely used as software for object detection. For example, the YOLO series is known. Patent Document 3 proposes a fragment detection method that identifies fragments from captured images using YOLO (v3), which is a machine learning model for object detection by deep learning using a convolutional neural network. .

Japanese Patent Application Laid-Open No. 2008-212778 JP 2016-194505 A JP 2021-8754 A

When manufacturing standard-sized gravel from rough stones collected from a quarry, the quality of the rough stones, especially the size and the mixing ratio of large and small sizes, greatly affects the productivity and quality of the gravel product. comes out. In the past, people visually checked the condition of the ore to be processed, judged the decrease in productivity and quality based on intuition and experience, and issued instructions to the ore extraction site, factory, shipping department, etc. , to maintain product productivity and product quality. However, due to labor shortages, the aging and decreasing number of skilled workers, etc., it is becoming difficult to rely on human labor to determine the quality of raw stones and maintain the quality and efficiency of gravel production. In addition, there are limits to visual confirmation and judgments that rely on intuition and experience. Under different working environments, even the same quality of ore may look different. You cannot be expected to make good judgments all the time due to factors.

For example, in the gravel production process at a gravel factory, a sorting process is performed in which rough stones are sieved and sorted into large and small sizes. The sorting process of sorting into large and small sizes is repeated. Through this process, rough stones are sorted into standard size gravel and shipped as products.

Here, as shown in FIG. 10(A), when the raw stone 100 to be put into the gravel production process is in a predetermined state of size, ratio of large and small sizes, etc. (when the size of the raw stone is well balanced), can maintain productivity by reducing the number of processing cycles consisting of the crushing step 110 and the sorting step 120 performed on the rough stone 100L sorted as a large size.

On the other hand, as shown in FIG. 10(B), when the size of the rough stone 100 is large as a whole and the mixing ratio of the large and small sizes is biased, especially when the mixing ratio of the large size is large, the large size The number of processing cycles for the raw stone 100L sorted out as 100L increases, and the amount of small-sized raw stone 100S smaller than the standard size obtained in one cycle decreases, resulting in a decrease in gravel productivity. In addition, if the amount of re-crushed 100L of raw stone is large, the load on the crushing machine will be high, and it will be necessary to reduce the amount of newly input raw stone 100 in each processing cycle, resulting in a further decrease in productivity. put away. If this situation continues, there is a risk that raw material inventory will stagnate, leading to a vicious circle of tightness in the stock field and temporary suspension of ore extraction.

In view of these points, the object of the present invention is to accurately grasp the quality of the raw stone (size, mixing ratio of large and small sizes, etc.) from the photographed image of the raw stone in the gravel factory without relying on human labor. To propose a gravel production control method capable of maintaining productivity and quality. It is another object of the present invention to provide a computer program for gravel production management used to cause a computer-based system to execute the method.

In order to solve the above problems, the present invention
A gravel production process in which the rough stones are processed into standard size gravel by repeating the sorting process of sifting rough stones into large and small sizes and the crushing process of crushing the rough stones that have been selected as large in a crushing machine. is controlled by a management system equipped with a computer,
an image acquisition step of acquiring a photographed image of the raw ore input to the gravel production process;
Using a rough stone detection model equipped with a machine learning function, the raw stone image parts in which the individual raw stones appear are extracted from the photographed image, and a bounding box having a size surrounding each of the extracted rough stone image parts is generated. a detection step;
The size (area) of the bounding box representing the raw stone image portion is calculated, and the calculated size (area) is compared with a reference value for size selection, and the raw stone reflected in the raw stone image portion is reduced to the standard size. A raw stone quality determination step of classifying into the following small-sized raw stones and large-sized raw stones exceeding the standard size, and calculating the number of these small-sized raw stones and large-sized raw stones;
The size (area) of each bounding box, and the number of each of the small-sized ore and the number of the large-sized ore are displayed on the display screen of the display device together with the photographed image in which each ore image portion is surrounded by the bounding box. a display step of displaying
It is characterized by having

In the present invention, using an AI-based ore detection model, the photographed image of the input ore to be processed is analyzed, and individual ore images are separated and extracted from the background image to detect each ore, and the detection result is Based on this, the quality (size, mixture ratio of large and small sizes) of the input ore is discriminated, and the discriminated result is visualized by displaying it on the display screen in real time along with the photographed image. When constructing the rough stone detection model, various images (images containing rough stones of different sizes, rough stone images with different mixing ratios of large and small sizes, images with different shooting conditions such as illumination and shooting angle) are prepared and trained as training data. , the detection accuracy can be easily improved. In addition, by calculating the size (area) of the rough stone on the image and the number of large and small sizes and visualizing it as a numerical value, the worker can check the quality of the raw stone (size, mixing ratio of large and small) from the screen in real time. You can check it intuitively. Based on the displayed content, the worker can determine whether or not the quality of the input ore is appropriate, and can issue appropriate instructions to the mining site, etc., thereby preventing a decrease in productivity.

In the present invention, in the ore quality determination step, the input ore is processed into standard size gravel based on at least one of the number ratio and area ratio of small size ore and large size ore contained in the input ore. It is possible to judge the quality of productivity in each case. By storing the correlation between these ratios and productivity in the management system in advance, it is possible to accurately determine the quality of productivity. By displaying the pass/fail judgment on the screen, workers can check the suitability of the input ore.

In the present invention, there may be a registration process for registering history information of gravel production management in the gravel production process in the database. In this case, the history information includes at least raw ore information specifying the input ore, the quality determination result in the ore quality determination process of the input ore, and the processing date and time of the input ore in a form associated with each other. It is desirable to be

A management system for managing gravel production according to the method of the present invention is wired or wirelessly connected to the factory side system installed in the gravel factory that performs the sorting process and the crushing process, and the factory side system installed in the management office that manages the gravel factory. It can be configured to include a management side system that performs communication. In this case, the management side system may be configured to include a management server, a management database, and a management side monitor, and the factory side system may be configured to include a camera, an edge device, and a factory side monitor. can.

In the management system having this configuration, in the gravel factory, the camera executes the photographing process of photographing the input rough stone that is input to the raw stone input port of the sieve, and the edge device uses the photographed image acquisition process and the rough stone detection model. and a transmission step of transmitting the photographed image and the detection result to the management system.
Further, in the management office, the management server performs a receiving step of receiving the photographed image and the detection result, a quality judgment step based on the received detection result, a registration step, and a management monitor for the photographed image together with the detection result and the judgment result. and a transmission step of transmitting the quality determination result to the edge device.
Furthermore, in the gravel factory, the edge device receives the quality judgment result, and the display process of displaying the photographed image of the camera on the screen of the factory monitor along with the raw stone detection result and the received quality judgment result. do.

1 is a schematic configuration diagram showing the overall configuration of a gravel production management system to which the present invention is applied; FIG. 4 is a schematic flow chart showing a procedure for creating a raw stone detection model; 4 is a schematic flow chart showing an inference procedure for detecting a rough stone from a photographed image using a raw stone detection model; (A) is a schematic flow chart showing a procedure for judging the quality of input ore from the result of ore detection, and (B) is an explanatory view thereof. FIG. 4 is an explanatory diagram showing an example of a display screen displaying a photographed image and a quality determination result of a raw stone; It is an explanatory view showing the main part in a gravel production control system. FIG. 4 is an explanatory diagram showing an example of utilization of history information such as gravel production management collected in a management database in a gravel production management system; FIG. 10 is an explanatory diagram showing an example of analysis results of the size (area) of raw stone and the production amount of gravel; FIG. 4 is an explanatory diagram showing an analysis example of raw stone size and history information related to physical distribution; (A) and (B) are explanatory diagrams showing problems in the gravel production process.

An embodiment of a gravel production management system to which the method of the present invention is applied will be described below with reference to the drawings. In addition, embodiment shows an example of this invention, and does not intend to limit this invention to embodiment.

(overall structure)
FIG. 1 is a schematic configuration diagram showing the overall configuration of a gravel production management system. The gravel production management system 1 includes a factory-side system 3 installed in a gravel factory 2, a management-side system 5 installed in a management office 4 that manages the gravel factory 2, and a raw stone quarry for collecting raw stones for the gravel. It is equipped with a collection site side system 7 installed in 6. Each of these

systems

3, 5, and 7 is mainly composed of a computer, and executes the steps and processes described below by executing pre-installed software. These

systems

3, 5 and 7 are connected via a wireless or wired communication line 8. FIG.

The factory-side system 3 includes an edge device 31, an input/output device including a factory-side monitor 32, and a camera 33 for photographing input rough stones. In the gravel factory 2, a raw stone 9 carried from a raw stone quarry 6 is put into a sieve 22 from a raw stone input port 21 and sieved into a small-sized raw stone 9S smaller than the standard size and a large-sized raw stone 9L larger than the standard size. The small size ore 9S is transferred to the next step. The large size ore 9L is repeatedly subjected to a processing cycle consisting of a crushing step 23 by a crushing machine and a screening step 24 by a sieve to be processed into a small size ore 9S smaller than the standard size. The sorting/crushing process (or crushing/sorting process) is performed in multiple steps as necessary. The small-sized rough stone 9S obtained through such a gravel production process undergoes a predetermined post-treatment process 10 to obtain gravel as a product, which is shipped from the shipping department 25. FIG.

The camera 33 for photographing the raw stone is installed at a position capable of photographing the raw stone 9 (hereinafter sometimes referred to as the raw stone 9) that is input from the raw stone input port 21. The input ore 9 is photographed in real time by the camera 33 (photographing step), and the obtained captured image 11 of the input ore 9 is captured by the edge device 31 . In the edge device 31, the ore detection process is performed using the installed learned ore detection model 34 (ore detection process, see FIGS. 2 and 3 described later). The ore detection result 35 including the photographed image 11 is transmitted from the transmission/reception unit 36 of the edge device 31 to the management side system 5 of the management office 4 via the communication line 8 (transmission step).

The management-side system 5 includes a management server 51, a management database 52 built into or external to the management server 51, and an input/output device including a management-side monitor 53. The management server 51 receives the raw stone detection result 35 via the transmission/reception unit 54 (receiving step). The management server 51 analyzes the received ore detection result 35 in the ore quality judging unit 55 to judge the quality of the ore (quality judging step, see FIG. 4 described later).

The quality determination result 56 includes the size (area) of the raw stone shown in the photographed image 11 of the input raw stone, the number of large and small size raw stones, and the like. The raw stone detection result 35 and the quality determination result 56 are registered in the management database 52 (registration process) and displayed on the display screen of the management side monitor 53 via the display control unit 57 (display process, see FIG. 5 described later). ). Also, the quality judgment result 56 is transmitted from the transmitting/receiving section 54 to the factory side system 3 and the collection site side system 7 via the communication line 8 . The quality judgment result 56 can also be displayed on a terminal device 59 installed in each department via a communication network 58 such as a LAN.

When the edge device 31 of the factory-side system 3 receives the quality determination result 56 via the transmission/reception unit 36 (receiving step), the received quality determination result 56 is transmitted via the display control unit 37 to the captured image 11 of the camera 33. are displayed on the display screen of the monitor 32 on the factory side (display step). Similarly, in the collection site system 7, the image 11 taken by the camera 33 and the quality judgment result 56 of the raw stone are displayed on the monitor 71, the portable communication terminal 72 carried by the field worker, and the like.

As described above, in the gravel production management system 1, the raw stones are detected from the photographed image 11 of the input raw stones 9, the quality of the detected raw stones (size, the number of large and small sizes, etc.) is determined, and the photographed image 11 and the raw stone detection result are determined. 35 and the quality judgment result 56 are displayed in real time on the terminal screens installed in the gravel factory 2, the management office 4 and the ore quarry 6. The quality of the input ore 9 can be confirmed, judged, and notified unmanned, and the person in charge of the management office 4 who visually confirms the screen can grasp the quality of the ore from the display screen and control the product quality. Appropriate instructions can be issued to the ore quarry 6, shipping department 25, and the like. In addition, workers in the ore quarry 6 and workers in the shipping department 25 can visually check the quality results on the display screen of the terminal and share the quality results. Therefore, it is possible to maintain the productivity and quality of gravel by efficiently collaborating between the management department, the gravel production department, and the rough stone extraction department and the shipping department that are pre- and post-processes.

(Rough stone detection model, raw stone detection process)
The trained ore detection model 34 installed in the edge device 31 installed in the gravel factory 2 is an object detection model consisting of a machine learning machine algorithm such as a convolutional neural network (CNN) utilizing AI. It is possible to use the YOLO series, which has been developed.

FIG. 2 is a schematic flow chart showing the procedure for creating the ore detection model 34. FIG. Although the preparation procedure is a general one, the training data of the object detection model includes a photographed image of the raw stone thrown in from the raw stone inlet 21 . As photographed images, input ore images for learning were photographed by changing the position, angle, distance, brightness, state of ore, etc. of the camera 33 (ST21), and a data set for learning was created (ST22). These were input as teaching data, and deep learning was performed so as to obtain desired detection results (ST23), and a raw stone detection model 34, which is a learned model capable of accurately extracting raw stones, was obtained (ST24).

FIG. 3 is a schematic flow chart showing an inference procedure for detecting ores from the photographed image 11 using the ore detection model 34 . In the gravel factory 2, the camera 33 photographs the input rough stone 9 passing through the raw stone passage area (the raw stone input port 21) (ST31), and the obtained captured image 11 of the input rough stone 9 is sent to the edge device 31 (ST32). : image acquisition step). In the edge device 31, the photographed image 11 is analyzed using the ore detection model 34 (ST33: analysis step), and the feature amount of the image portion where each of the input ore 9 included in the photographed image 11 is shown is extracted. (ST34), based on the feature amount, each part of the ore image in which each input ore 9 is captured is extracted, and a bounding box (rectangular frame) having a minimum size capable of containing the extracted ore image part is generated. (ST35: raw stone object detection). On the display screen of the monitor, each detected input raw stone 9 is represented by each bounding box 12 displayed on the photographed image 11 .

(Rough stone quality judgment process)
FIGS. 4A and 4B are a schematic flow chart and an explanatory diagram showing the procedure for judging the quality of the input ore 9 from the ore detection result 35 (information of the bounding box 12) detected by the ore detection model 34. FIG. When receiving the raw stone detection result 35, the management server 51 of the management system 5 calculates the size (area) of the bounding box 12 representing the raw stone image portion, and the calculated size (area) is used as the reference value for size selection. Compared with the threshold value, the input ore 9 shown in the ore image portion is classified into a small size ore 9S that is equal to or smaller than the standard size and a large size ore 9L that exceeds the standard size. Calculate the number.

Describing according to the flowchart of FIG. 4(A), initial settings for determination are performed first. A predetermined standard size n is set as a threshold value t, which is a reference value for sorting out the large and small size rough stones 9L and 9S (ST41: t=n). The count number Sc of the small size ore 9S and the count number Lc of the large size ore 9L are reset to "0" (ST42: Sc=0, ST43: Lc=0). Next, the vertical (h) and horizontal (w) sizes of each bounding box 12 are obtained from the information of the bounding box 12 (rectangular frame) representing the detected ore image part (ST44), and from this information, An area (a) representing the size of each bounding box is calculated (ST45: a=h*w). The calculated area (a) is compared with the threshold (t) (ST46). If the area (a) is smaller than the threshold value (t), add 1 to the count value Sc of the small size rough stone 9S (ST47), and if it is larger, add 1 to the count value Lc of the large size rough stone 9L. (ST48).

Here, in the ore quality determination step, when processing the input ore into standard size gravel based on the ratio of the number (mixing ratio) or the area ratio of the small size ore 9S and the large size ore 9L contained in the input ore 9 It is possible to judge the quality of the productivity of

When the mixing ratio of the large-sized ore 9L contained in the input ore 9 is large, as described above (see FIG. 10), it is necessary to repeatedly perform the crushing process and the sorting process, resulting in a decrease in gravel productivity. . If the correlation between the mixing ratio of large and small stones and the productivity of gravel smaller than the standard size at each gravel mill is grasped in advance, It is possible to judge the quality of gravel productivity from the correlation.

For example, in each gravel factory 2, the mixing ratio of the large and small sizes of the input rough stones 9 and the degree of productivity when producing gravel of a standard size or less set from the input rough stones are registered in the management database 52 as a production history. do. Based on the production history in each gravel factory 2, the correlation between the mixing ratio of large and small sizes of input rough stones and the productivity of gravel of a set standard size or less is obtained using a statistical method such as regression analysis. . In the ore quality determination unit 55 (see FIG. 1) of the management server 51, based on the obtained correlation, the mixing ratio of the large and small sizes of the input ore 9 and the set standard size are used to determine the gravel processing from the input ore 9. Estimate productivity. The estimation result (productivity judgment) is displayed in real time on the display screen together with the photographed image 11 as one of the raw stone quality judgment results 56, for example. The person in charge of management can quickly and accurately issue countermeasures for improving productivity to relevant departments based on the displayed contents.

(Judgment result display process)
On the screen of the management-side monitor 53 of the management-side system 5, the raw stone photographed image is displayed and the determination result is also displayed. FIG. 5 is an explanatory diagram showing an example of a display screen. As shown in this figure, on the display screen of the monitor 53 on the management side, a photographed image 11 is displayed with each ore image portion surrounded by a bounding box 12 . Along the upper frame of each bounding box 12 displayed, the calculated size (area) is displayed as "size 119.0" or the like. At one of the upper corners of the display area of the display screen 53, a threshold value for size selection such as "Threshold: 50" is displayed. etc., and below that, the number of small-sized rough stones 9S is displayed as "Small Count: 23". Note that the display form is an example, and can be displayed in various forms. In addition, when the ratio of the size of the input ore 9 included in the photographed image 11 and the ratio of the area of the large and small sizes are calculated, they can be displayed on the screen in various display forms such as a pie chart and a bar graph. can be displayed in

(Effect)
FIG. 6 is an explanatory diagram collectively showing the main functions and actions of the gravel production management system 1 of this example. As described above, in the gravel production management system 1, the camera 33 is installed at the first process location in the gravel factory 2, for example, the place where the input ore 9 passes through, such as the ore input port, and the photographed image 11 of the input ore 9 is displayed. are getting The acquired photographed image 11 is analyzed using the ore detection model 34 to obtain the ore detection result 35 (vertical and horizontal coordinates, vertical and horizontal size of the bounding box 12, detection date and time, etc.). Based on the raw stone detection result 35, the management server 51 calculates the area of the detected raw stone, determines the size of the large and small stones, and counts the number of large and small size raw stones. In real time, the screen displays the detected ore and the digitized quality (size and number) of the ore.

According to the gravel production management system 1 of this example, the quality of the input ore can be accurately determined without relying on human labor, and workers, managers, etc. can confirm the determination results on the screen in real time. In addition, it is possible to check the photographed image of the input rough stone and the judgment result on the terminal installed in each department in the gravel production factory, the terminal installed in the management office, and the terminal installed in the rough stone quarry. Therefore, information such as the quality of input ore can be shared within the factory and related facilities, and notifications and instructions regarding countermeasures, treatments, etc. based on the judgment results can be promptly and accurately made. In addition, without relying on human labor, it is possible to accurately determine the productivity when producing gravel of a set standard size from input ore. It is possible to quickly and accurately deliver to

(History management process/utilization of history information)
FIG. 7 is an explanatory diagram showing an example of utilization of history information such as gravel production management collected in the management database 52 in the gravel production management system 1. As shown in FIG. As shown in this figure, in the gravel production management system 1 of this example, the history information of the gravel production management of the gravel production process performed in the gravel factory 2 is registered in the management database 52 . The history information can include raw stone information that identifies the raw stone (specific information about the raw stone quarry), gravel production status, quality judgment results of the raw stone, distribution history, and the like. Logistics history can include, for example, the history of carrying rough stones from the quarry to the gravel factory, the handling history of the rough stones brought in at the gravel factory, and the shipping history of the produced gravel.

By analyzing historical information with dates, it is possible to efficiently manage gravel production at each gravel factory or the like. For example, analysis of historical information may reveal new parameters related to improvements in productivity. Such parameters that are effective in determining productivity can be reflected in rough stone quality determination. For example, in addition to the previously mentioned correlation between productivity and the ratio of large and small sizes of input rough,
Correlation between the location (position, geology, etc.) of the ore quarry, progress (extraction depth, etc.), and the area ratio of the large and small sizes of the ore extracted,
Correlation between the rough stone quarry and the productivity of gravel of the set standard size,
etc. can be analyzed.

Fig. 8 is an explanatory diagram showing an example of the analysis results of the size of the ore (area) and the amount of gravel produced. The raw stone size DB shown in FIG. 8(A) is the area (total area of the raw stones collected) to which the date and time (time stamp) are linked, and the production DB shown in FIG. 8(B) is the date and time (time stamp). is the product weight at each linked gravel mill. As shown in FIG. 8(C), the product amount and area generally show a high correlation, and the smaller the area, the larger the product amount. Based on this correlation, it can be determined that the ore quality is good when the area is less than or equal to a predetermined value, or when the amount of product is greater than or equal to a predetermined value.

Fig. 9 is an explanatory diagram showing an analysis example of raw stone size and logistic-related history information. The contents of the raw stone size DB shown in FIG. 9(A) are the same as those in FIG. 8(A). The physical distribution DB shown in FIG. 9B includes origin information (collection site information) and destination information (gravel factory information) of rough stones linked to date and time. This information includes the geological information of the starting point (collection site). It also includes latitude and longitude information from the departure point (collection site) to the arrival point (gravel factory), and based on this information, information on the distance from the departure point to the arrival point can be generated. It can also generate information about excavation time in each section.

Based on this information, it is possible to analyze the correspondence relationship between the position of each work section and the average size of the ore in each work section, as shown in Fig. 9(C). Also, as shown in FIG. 9(D), the relationship between the distance from a certain starting point A and the average size can be analyzed. Furthermore, as shown in FIG. 9E, the relationship between excavation time (progress) and average size in each section can be analyzed.

Various analyzes can be performed from the history information accumulated in the management database 52, and are not limited to the above analysis examples.

Claims

A gravel production process for producing gravel of a standard size from the rough stones by repeating a sorting process of sorting the raw gravel stones into large and small sizes and a crushing process of crushing the rough stones selected as large sizes, is performed by a computer. In the gravel production management method managed by the management system,
an image acquisition step of acquiring a photographed image of the raw ore input to the gravel production process;
Using a rough stone detection model equipped with a machine learning function, the raw stone image parts in which the individual raw stones appear are extracted from the photographed image, and a bounding box having a size surrounding each of the extracted rough stone image parts is generated. a detection step;
The size of the bounding box representing the rough stone image portion is calculated, and the calculated size is compared with a reference value for size selection, and the rough stone reflected in the rough stone image portion is classified as a small size rough stone of the standard size or less and A rough stone quality determination step of classifying into large-sized raw stones exceeding the standard size and calculating the number of these small-sized raw stones and large-sized raw stones;
The size of each bounding box and the number of each of the small-sized rough stones and the number of the large-sized rough stones are displayed on the display screen of the display device together with the photographed image in which each rough stone image portion is surrounded by the bounding box. a display step;
gravel production management method.
In claim 1,
In the ore quality judgment step,
Based on at least one of the number ratio and the area ratio of the small-sized ore and the large-sized ore included in the input ore, the quality of the productivity when processing the input ore into gravel of the standard size or smaller. Gravel production control methods that make judgments.
In claim 1 or 2,
a registration step of registering history information of gravel production management in the gravel production process in a database;
The history information includes at least raw ore information specifying the input ore, the judgment result of the ore quality determination process of the input ore, and the processing date and time of the input ore in a form associated with each other. Gravel production management method.
In claim 3,
A factory-side system installed in a gravel factory that performs the sorting process and the crushing process, and a management side that is installed in a management office that manages the gravel factory and communicates with the factory-side system by wire or wireless. A configuration comprising a system and
The management side system is configured to include a management server, a management database, and a management side monitor,
The factory-side system is configured to include a camera, an edge device, and a factory-side monitor,
In the gravel factory, the camera performs a photographing step of photographing the input rough stone input to the raw stone input port of the sieve, and the edge device performs the image acquisition step and the rough stone detection using the rough stone detection model. and a transmission step of transmitting the photographed image and the raw stone detection result to the management system,
at said administrative office,
a receiving step of receiving the photographed image and the detection result; a step of judging the quality of the ore based on the received detection result; and performing a display step of displaying on the management monitor and a transmission step of transmitting the determination result to the edge device,
In said gravel mill,
Gravel production in which the edge device executes a reception step of receiving the determination result and a display step of displaying the image captured by the camera on the screen of the factory-side monitor together with the detection result and the received determination result. Management method.
Gravel production management for managing a gravel production process for processing the rough stones into standard size gravel by repeating a sorting process of sorting rough stones into large and small sizes and a crushing process of crushing rough stones that have been sorted as large sizes. A computer program for
an image acquisition step of acquiring a photographed image of the raw ore input to the gravel production process;
Using a rough stone detection model equipped with a machine learning function, the raw stone image parts in which the individual raw stones appear are extracted from the photographed image, and a bounding box having a size surrounding each of the extracted rough stone image parts is generated. a detection step;
The size of the bounding box representing the rough stone image portion is calculated, and the calculated size is compared with a reference value for size selection, and the rough stone reflected in the rough stone image portion is classified into a small size rough stone of a standard size or less and the above A raw stone quality judgment step of classifying into large-sized raw stones exceeding the standard size and calculating the number of these small-sized raw stones and large-sized raw stones;
The size of each bounding box and the number of each of the small-sized rough stones and the number of the large-sized rough stones are displayed on the display screen of the display device together with the photographed image in which each rough stone image portion is surrounded by the bounding box. a display step;
A computer program for gravel production management, characterized by causing a computer to execute
In claim 5,
The raw stone quality judging step determines the quality of gravel productivity in the gravel production step based on at least one of a number ratio and an area ratio of small-sized rough stones and large-sized rough stones contained in the input raw stones. computer program for gravel production control.
In claim 5 or 6, further
cause a computer to execute a registration step of registering the gravel production management history in the gravel production process in a database;
In the registration step, the gravel is stored in the database in a form in which at least raw stone information specifying the input raw stone, the determination result of the raw stone quality determination step of the input raw stone, and the processing date and time of the input raw stone are associated with each other. A computer program for production control.