JP2023034621A - Work discrimination system - Google Patents

Abstract

To provide a work discrimination system configured to acquire cumulative points of work results and reflect the points in useful information utilization or salary.SOLUTION: A work discrimination system 100 includes: a work analysis and learning unit 106 which determines a standard work pattern 114 for each attribute information, on the basis of a combination of an estimated work pattern composed of combinations of work elements constituting a work, based on processed video information 112 and position information 111, and attribute information 113 extracted by an attribute information extraction unit 104 from video information 110; and a point calculation unit 108 which calculates a matching degree 115 by comparing the work pattern estimated based on the processed video information 112 and the position information 111 with the standard work pattern 114 having the same attribute information as the attribute information 113, and calculates a point value 116 corresponding to the matching degree 115, the point value being transmitted using an alarm 117 or e-mail 118 via a notification unit 109.SELECTED DRAWING: Figure 1

Description

The present invention relates to a system that analyzes and discriminates the work of people, machines, etc. at a work site or the like.

Conventionally, at construction sites, production sites, etc., in order to understand work omissions and work time for each worker, work is organized using video of workers doing work and worker position information. Techniques for analyzing work elements are known (for example, the technique described in Patent Document 1). In this prior art, the work element of the worker is inferred from the positional information of the body part of the worker and the object at a certain time.

International Publication No. 2018/087844

However, just by inferring work elements from only the position information of a body or an object acquired at a certain time and displaying the results, as in the above-described prior art, it is possible to improve the work process, but it is There was a problem that the analysis results of the elements could not be objectively reflected in work results, wages, etc., and the range of utilization was limited.

In addition, in the above-described prior art, it was not possible to infer work elements for machines such as heavy machinery at construction sites, for example.

Therefore, the object of the present invention is to visualize the work movements of specific workers and machines using images and position information as points (digitization), so that it is possible to acquire accumulated points of work results and make them useful. It is possible to use information and reflect it in wages, etc., thereby promoting the motivation of workers, and conversely, by warning against dangerous or erroneous movements of workers and machines, safety It is to make it possible to ensure quality, and ultimately to improve labor productivity and quality.

An example of a work determination system includes: a work information acquisition unit that acquires work information related to work performed by a worker or a work machine; an attribute information extraction unit that extracts attribute information representing an attribute of work from the work information; Based on the work information acquired by the acquisition unit, a work pattern consisting of a combination of a plurality of work elements constituting work is estimated, and the attribute information extracted by the attribute information extraction unit for the estimated work pattern and the work information. A work analysis learning unit that determines a standard work pattern for each attribute information based on the combination and stores it in a standard work pattern storage unit; A work pattern consisting of a combination is estimated, and the estimated work pattern is compared with the standard work pattern stored in the standard work pattern storage unit having the same attribute information as the attribute information extracted by the attribute information extraction unit for the work information. and a point calculation unit that calculates a degree of matching by the comparison and calculates a point value corresponding to the degree of matching.

According to the present invention, by visualizing the work movements of a specific worker or machine using video and positional information as points (digitization), it is possible to obtain accumulated points of work results and obtain useful information. It is possible to reflect it in utilization and wages, etc., thereby promoting the motivation of workers, and conversely, by warning against dangerous or incorrect movements of workers and machines, safety is improved. It will be possible to secure and ultimately improve labor productivity and quality.

1 is a block diagram of an embodiment of a work discrimination system; FIG. FIG. 3 is a configuration diagram showing a specific example of an image information acquisition unit; FIG. 4 is a configuration diagram showing a specific example of a position information acquisition unit; 9 is a flowchart showing an example of work analysis learning processing; 9 is a flowchart showing an example of point calculation processing;

EMBODIMENT OF THE INVENTION Hereinafter, it demonstrates in detail, referring drawings for the form for implementing this invention. FIG. 1 is a block diagram of an embodiment of a work discrimination system. The work determination system 100 is both a part of the work information acquisition unit, and includes a video information acquisition unit 101, a position information acquisition unit 102, a video information processing unit 103, an attribute information extraction unit 104, and an attribute information database 105 ( hereinafter referred to as “attribute information DB 105 ”), a work analysis learning unit 106 , a standard work pattern storage unit 107 , a point calculation unit 108 and a notification unit 109 .

The image information acquisition unit 101 acquires image information 110 obtained by imaging work performed by a worker or a working machine at a construction site, construction site, production site, or the like, as work information. FIG. 2 is a configuration diagram showing a specific example of the video information acquisition unit 101. As shown in FIG.
For example, video information captured by one or more cameras 201 installed at a construction site, a construction site, or a production site is transmitted from a WiFi router device 202 installed in a school via the Internet 203 to the image information 110 shown in FIG. It is aggregated in a server device (not shown) connected to the Internet 203, which is one work determination system 100. FIG.
Further, video information captured by the wearable camera 204 attached to the worker's helmet is aggregated in the server device as video information 110 via the Internet 203 from the WiFi router device 202 similar to that described above.

The position information acquisition unit 102 acquires position information 111 indicating the work position of the worker or work machine as work information. FIG. 3 is a configuration diagram showing a specific example of the position information acquisition unit 102. As shown in FIG.
For example, the work machine 300 is equipped with a multi-GNSS terminal that receives highly accurate positioning information such as GPS, GLONASS, and QZSS from the quasi-zenith satellites 301, 302, and 303 every second as the position information acquisition unit 102. . The terminal also utilizes QZSSL1-S (submeter class) and L6 (centimeter class) augmentation signals to correct for positioning errors.
Also, an IMES terminal is installed indoors such as a building as the position information acquisition unit 102 .
Location information 111 obtained from these terminals is aggregated in the aforementioned server device via the Internet or the like.

Returning to the description of FIG. 1, the attribute information DB 105 stores a data set of an appearance image of a worker or a working machine and attributes to which the worker or the working machine belongs.

The attribute information extraction unit 104 extracts from the video information 110 attribute information 113 representing the attribute of work. More specifically, the attribute information extraction unit 104 extracts an appearance image obtained by performing an appearance recognition process on the image information 110 to obtain an appearance image of each data set among the data sets stored in the attribute information DB 105 . extracting the attribute information corresponding to the data set that matches the image as attribute information 113;
The appearance image described above is, for example, a face image obtained by performing face recognition processing, which is appearance recognition processing, on the video information 110 corresponding to the worker.
Alternatively, the appearance image described above is, for example, a machine contour image obtained by performing machine contour recognition processing, which is appearance recognition processing, on the video information 110 corresponding to the working machine.

The video information processing unit 103 outputs processed video information 112 by performing predetermined processing on the video information 110 .
The processed video information 112 may be, for example, time-lapse video data obtained by performing time-lapse processing as the predetermined processing on the video information 110, which is a moving image.
Alternatively, the processed image information 112 may be, for example, image feature amount data obtained by extracting image feature amounts from each frame image of the image information 110, which is a moving image.

The work analysis learning unit 106 executes work analysis learning processing. FIG. 4 is a flowchart illustrating an example of work analysis learning processing. For example, although not shown, this processing is an operation in which the processor of the server device executes a work analysis learning processing program stored in the memory.

In the processing of the flowchart of FIG. 4, first, the processing waits until the processed video information 112 is input from the video information processing unit 103 of FIG. 1 (repeated determination of NO in step S401).

If the determination in step S401 is YES, the position information 111 is obtained from the position information obtaining unit 102 (step S402).

Next, based on the processed video information 112 acquired in step S401 and the position information 111 acquired in step S402, a plurality of work elements constituting a series of work are extracted (step S403). The work elements are, for example, the individual tasks of a worker, such as carrying materials, assembling materials, tightening screws, and the like.

Next, the time for each work element extracted in step S403 is extracted from the processed video information 112 (step S404).

Based on the information obtained in steps S401 to S404, work patterns are estimated (step S405).

Further, the attribute information 113 obtained by the attribute information extracting unit 104 is obtained for the original video information 110 of the processed video information 112 in step S401 (step S406).

Based on the combination of the work pattern estimated in step S405 and the attribute information 113 acquired in step S406, the standard work pattern 114 for each attribute information is determined. 113 are stored in the standard work pattern storage unit 107 in FIG. 1 (step S406). After that, it returns to the standby state of step S401 again.

Here, the work analysis learning unit 106 of FIG. 1 may determine the standard work pattern 114 for each attribute information 113 by machine learning, for example. More specifically, the work analysis learning unit 106 may determine the standard work pattern 114 for each attribute information 113 by machine learning using a deep neural network.

The point calculation unit 108 in FIG. 1 executes point calculation processing. FIG. 5 is a flowchart illustrating an example of point calculation processing. This processing is, for example, similar to the case of FIG. 4, although not shown, the processor of the server device described above executes the point calculation processing program stored in the memory.

In the processing of the flowchart of FIG. 5, first, the processing waits until the processed video information 112 is input from the video information processing unit 103 of FIG. 1 (repeated determination of NO in step S501).

If the determination in step S501 is YES, the position information 111 is obtained from the position information obtaining unit 102 (step S502).

Next, based on the processed video information 112 acquired in step S501 and the position information 111 acquired in step S502, a plurality of work elements constituting a series of work are extracted (step S503).

Next, the time for each work element extracted in step S503 is extracted from the processed video information 112 (step S504).

Based on the information obtained in steps S501 to S504, work patterns are estimated (step S505).

After that, the attribute information 113 obtained by the attribute information extracting unit 104 is obtained for the original video information 110 of the processed video information 112 in step S501 (step S506).

Then, the standard work pattern storage unit 107 in FIG. 1 is accessed, and the work pattern estimated in step S505 is stored in the standard work pattern storage unit 107 having the same attribute information 113 as the attribute information 113 acquired in step S506. It is compared with the standard work pattern 114 to calculate the matching degree 115 between them (step S507).

Finally, the point value 116 corresponding to the degree of matching calculated in step S507 is output from the point calculation unit 108 in FIG. 1 to the notification unit 109 (step S508). After that, it returns to the standby state of step S501 again.

Returning to the description of FIG. 1 , the notification unit 109 performs either or both of the alarm notification 117 and the email notification 118 according to the point value calculated by the point calculation unit 108 .

According to the above-described embodiment, for example, by visualizing specific actions and movements such as work using video and position information, points (digitization) are obtained, and accumulated points of action results can be acquired and useful. Reflecting on information utilization and wages, promoting motivation, conversely, by warning dangerous behavior as accumulation points, ensuring safety, ultimately leading to improvement of labor productivity and quality. becomes possible.

For example, the work determination system 100 starts operating by designating a work start point in a factory work process. Next, in FIG. 1, the motions of workers, working machines, etc. are automatically measured using video information 110 and position information 111, and the motions are analyzed as standard work patterns 114 using AI etc. by the work analysis learning unit 106. be done. Furthermore, the point calculation unit 108 converts the work of each worker and work machine into points, outputs the point data judged as good or bad, and stores the point data in a hard disk (not shown).
In this way, by scoring the actions of employees and work machines engaged in each task based on the video information 110 and the position information 111, it is possible to manage the performance of the task (whether or not the time and process are strictly observed, It is possible to connect to the

In addition, by giving points to and accumulating various individual attribute information 113 such as personnel data for vendors, full-time employees, subcontractors, and part-time workers, in the future, for example, for each specific customer, big data can be generated and data mining can be performed. , etc., can be used as basic data for behavior analysis and marketing data for all industries, and analysis data for work machines such as heavy machinery and automatic cleaning vehicles.

As described above, in this embodiment, it is possible to determine whether the person is a contractor, a full-time employee, a contractor, a part-time worker, or another dangerous person as the attribute information 113, and use the work pattern as evidence. . If it is a legitimate person, it is possible to perform personnel evaluation, attendance record, etc. such as working seriously. There are various people such as contractors, full-time employees, outsourced workers (for example, masters and craftsmen), and part-time workers at construction and civil engineering sites. Basic data can be obtained for behavioral analysis, such as whether a person is following a line, or whether a person making normal movements tends to cross a certain line of movement, and the like. In this embodiment, these can be obtained as numerical data and point data.

In this embodiment, for the image information 110 (various image information 110) and the position information 111, machine learning such as deep learning is used to determine whether or not each of them is moving correctly and whether or not they are moving safely. It is possible to determine by the work analysis learning process by .

In the case of workers, by attaching a device such as an RFID to an ID card that can grasp positional information, it is possible to determine which room the worker should move in, including the room where they should not move much, and the room that they cannot enter in the first place. It becomes possible to determine whether or not

Furthermore, in this embodiment, the function of the notification unit 109 in FIG. 1 can issue a warning 117 for extremely dangerous work patterns.
Alternatively, the notification unit 109 can perform processing such as skipping by mail 118 together with detailed data for work patterns that are not so dangerous.

According to the present embodiment, whether or not the wrong heavy machinery is in the wrong area (room), and whether or not an alert is issued when it enters, is determined and detected based on the video information 110 and the position information 111. becomes possible.

According to this embodiment, it is also possible to confirm that the heavy machinery performs repetitive movements without leaving the area from 9:00 in the morning to 17:00 in the evening.

Alternatively, according to this embodiment, for example, in road paving work, it is possible to determine whether or not the contractor's heavy machinery has performed the rollers for a predetermined number of repetitions, and to determine and detect an alert if not performed. . Conventionally, it was left up to the contractor, so it was difficult for subcontractors to pursue the legitimacy of construction, but the introduction of this embodiment makes it possible.

For example, in the case of river construction, it is easy to go to the riverside, and the presence of heavy machinery can be dangerous, and workers may sink in the swamp. By using the image information 110 as an axis and combining it with the position information 111, it is possible to analyze and quantify the behavior of workers and working machines.

Any method may be used to specify the work start point as long as the time and ID can be specified. For example, barcode reading, video start time, RFID read time, etc. can be used.

As the position information 111, as described in FIG. 3, not only GPS but also position information with a centimeter error using correction signals of L1 (meter class) and L6 (centimeter class) of quasi-zenith satellites can be used. is possible. Since location information from satellites cannot be captured indoors, RFID and beacons can be used together. Also, the position information 111 associated with the action may be extracted based on recognition of a specific worker or work machine based on English and calculation of its spatial coordinates.

According to this embodiment, in the case of workers, it is possible to take measures such as not allowing ordinary people to enter the vicinity of dangerous machines and limiting the number of persons in charge. Based on the attribute information 113, it is also possible to detect that a person who is not qualified to drive (unlicensed, license expired) is approaching the forklift.

Personnel evaluation, safety evaluation, and the like can be performed by using the point values calculated according to the present embodiment.

100 work determination system 101 image information acquisition unit 102 position information acquisition unit 103 image information processing unit 104 attribute information extraction unit 105 attribute information DB
106 work analysis learning unit 107 standard work pattern storage unit 108 point calculation unit 109 notification unit 110 image information 111 position information 112 processed image information 113 attribute information 114 standard work pattern 115 matching degree 116 point value 117 notification 118 mail

Claims (10)

a work information acquisition unit that acquires work information related to work performed by a worker or a work machine;
an attribute information extraction unit that extracts attribute information representing attributes of the work from the work information;
estimating a work pattern consisting of a combination of a plurality of work elements constituting the work based on the work information acquired by the work information acquisition unit, and the attribute information extraction unit for the estimated work pattern and the work information; A work analysis learning unit that determines a standard work pattern for each attribute information based on the combination with the attribute information extracted by and stores it in a standard work pattern storage unit;
estimating a work pattern consisting of a combination of a plurality of work elements constituting the work based on the work information acquired by the work information acquisition unit, and applying the estimated work pattern to the work information to extract the attribute information; a degree of matching is calculated by comparing with the standard work pattern stored in the standard work pattern storage unit having the same attribute information as the attribute information extracted by the unit, and a point value corresponding to the degree of matching is calculated. a point calculation unit;
work discrimination system. 2. The work determination system according to claim 1, wherein said work information includes video information obtained by imaging work performed by said worker or said work machine. 3. The work determination system according to claim 2, wherein said work information includes processed image information obtained by performing predetermined processing on said image information. further comprising an attribute information database that stores a data set of an external image of the worker or the work machine and attributes to which the worker or the work machine belongs;
The attribute information extracting unit extracts, from among the data sets stored in the attribute information database, data in which an appearance image of the data set matches an appearance image obtained by performing appearance recognition processing on the video information. extract attribute information corresponding to the tuple,
The work discrimination system according to claim 2 or 3. 5. The work determination system according to claim 4, wherein said appearance image is a face image obtained by executing face recognition processing, which is said appearance recognition processing, on video information corresponding to said worker. 5. The work determination system according to claim 4, wherein said appearance image is a machine contour image obtained by executing machine contour recognition processing, which is said appearance recognition processing, on video information corresponding to said work machine. 7. The work determination system according to any one of claims 2 to 6, wherein said work information further includes position information indicating a work position of said worker or said work machine. 8. The work determination system according to any one of claims 1 to 7, wherein said work analysis learning unit determines a standard work pattern for each attribute information by machine learning. 9. The work determination system according to claim 8, wherein said work analysis learning unit determines a standard work pattern for each attribute information by machine learning using a deep neural network. 10. The work determination system according to any one of claims 1 to 9, further comprising a notification unit that performs either or both of an alarm notification and an email notification according to the point value calculated by the point calculation unit.

Images