JP2018081549A - Model creation device, terminal device, hazardous place discovery system, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a model creation device, a terminal device, a hazardous place discovery system, and a program that can mechanically discover a hazardous place at a work site.SOLUTION: The model creation device includes a model creating unit that executes machine learning using learning data in which information indicating a hazardous event at a hazardous place is affixed to an image of the hazardous place at a work site, thereby creating a hazardous place detection model for detecting the hazardous place.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a model creation device, a terminal device, a dangerous spot finding system, and a program.

  At the work site, the site supervisor needs to visually check whether there are any dangerous parts on the work site every day so that workers can work safely. However, because the field supervision work is wide-ranging and busy, there may be cases where a dangerous spot is missed. Therefore, there is a need for a method for mechanically detecting the dangerous location without the site supervisor directly viewing it.

Japanese Patent Laid-Open No. 2003-242202

  For example, as described in Patent Document 1, a method for predicting a danger to a work procedure at a construction site is conventionally known, but a method for detecting a dangerous point at a work site such as a construction site is known. Absent.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide a model creation device, a terminal device, a dangerous spot detection system, and a program that can mechanically detect a dangerous spot at a work site. is there.

  One aspect of the present invention is a risk for detecting a dangerous location by executing machine learning using learning data in which information indicating a dangerous event at the dangerous location is attached to an image of a dangerous location at a work site. It is a model creation apparatus provided with the model creation part which creates a location detection model.

  One aspect of the present invention is the model creation device described above, wherein the risk location detection model is a model of risk factors that cause the risk event.

  One aspect of the present invention is the above-described model creation device, in which the model creation unit performs the machine learning by deep learning using a convolutional neural network.

  According to one aspect of the present invention, the display unit includes an imaging unit that captures a work site as a captured image, a display unit that displays the captured image, and the dangerous point detection model generated by the model creation device. And a detection unit that detects the presence or absence of a dangerous spot in the captured image of the work site displayed on the terminal.

  One aspect of the present invention is the terminal device described above, and when it is determined that there is a location corresponding to the modeled risk factor, warning information indicating that the location is the dangerous location is provided as the warning information. A display control unit to be displayed on the display unit is further provided.

  One aspect of the present invention is the above-described terminal device, wherein the dangerous point detection model is a model of a risk factor that causes the dangerous event, and the terminal device is displayed on the display unit. A determination unit that determines whether or not a location corresponding to the modeled risk factor exists in the captured image of the work site, and a case where it is determined that a location corresponding to the modeled risk factor exists Further includes a display control unit that causes the display unit to display warning information indicating that the part is the dangerous part.

  One aspect of the present invention is the above-described terminal device, wherein the display control unit displays the warning information at the location when it is determined that the location corresponding to the modeled risk factor exists. To do.

  One aspect of the present invention is the above-described terminal device, wherein the warning information is information on the dangerous event caused by the risk factor.

  One aspect of the present invention is a dangerous spot finding system including the above-described model creation device and the above-described terminal device.

  One embodiment of the present invention is a program for causing a computer to function as the above-described model creation device.

  One embodiment of the present invention is a program for causing a computer to function as the terminal device described above.

  As described above, according to the present invention, a dangerous point at a work site can be found mechanically.

It is a figure which shows an example of schematic structure of the dangerous spot discovery system 1 which concerns on one Embodiment of this invention. It is a figure which shows an example of the learning data which concern on one Embodiment of this invention. It is an image figure where the field supervisor is imaging the work site using the terminal device 5 which concerns on one Embodiment of this invention. It is a figure explaining the flow of the process which discovers the dangerous spot of the dangerous spot discovery system 1 which concerns on one Embodiment of this invention.

  Hereinafter, the present invention will be described through embodiments of the invention, but the following embodiments do not limit the invention according to the claims. In addition, not all the combinations of features described in the embodiments are essential for the solving means of the invention. In the drawings, the same or similar parts may be denoted by the same reference numerals and redundant description may be omitted. In addition, the shape and size of elements in the drawings may be exaggerated for a clearer description.

  In the entire specification, when a part includes a component, “includes”, “have”, or “comprises”, this does not exclude other components unless otherwise stated. This means that it can further include the following components.

The dangerous spot finding system according to an embodiment of the present invention is a system that can find a dangerous spot at a work site from an image of a work site such as a construction site taken by a terminal device used by a user. In other words, the dangerous spot finding system according to an embodiment of the present invention is a system that can evaluate the risk of the work site from, for example, an image of the work site such as a construction site taken by a terminal device used by a user. is there. Here, the risk indicates a dangerous event. That is, to evaluate the risk at the work site is to specify a dangerous event at the dangerous location when the dangerous location exists at the work site.
Hereinafter, a dangerous spot finding system 1 according to an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram showing an example of a schematic configuration of a dangerous spot finding system 1 according to an embodiment of the present invention. As shown in FIG. 1, the dangerous spot finding system 1 includes an imaging device 2, a communication network 3, a learning server 4, and a terminal device 5. The learning server 4 may be an example of a model creation device.

The imaging device 2 transmits information to the learning server 4 via the communication network 3.
The imaging device 2 images a work site such as a construction site. The imaging device 2 may be a camera held by a person or a fixed point camera such as a surveillance camera. For example, the imaging apparatus 2 is implemented using an imaging element such as a CCD (Charge Coupled Device) or a CMOS (Complementary Metal Oxide Semiconductor) and various members such as a lens for controlling the formation of a subject image on the imaging element. It is an apparatus that images a space and generates a captured image. Further, the image captured by the imaging device 2 at a work site such as a construction site may be a still image or a moving image.
The imaging device 2 transmits the captured image to the learning server 4 via the communication network 3.

The communication network 3 may be a wireless communication transmission line, a wired communication, or a combination of a wireless communication transmission line and a wired communication transmission line. The communication network 3 may be a mobile communication network such as a mobile phone line network, a wireless packet communication network, the Internet, a dedicated line, or a combination thereof.
Note that the communication network 3 between the imaging device 2 and the learning server 4 can be omitted if the image captured by the imaging device 2 can be transferred to the learning server 4 using a portable recording medium. It is. For example, the storage medium is an externally mounted storage medium such as a hard disk, a magnetic disk, a magneto-optical disk, an optical disk, a USB (Universal Serial Bus) memory, or a memory card.

The learning server 4 transmits information to the terminal device 5 via the communication network 3.
The learning server 4 includes a communication unit 41, a processing unit 42, and a model creation unit 43.
The communication unit 41 performs wireless or wired communication with the imaging device 2 and the terminal device 5 via the communication network 3.
The processing unit 42 acquires an image from the imaging device 2 via the communication unit 41. Then, the processing unit 42 preprocesses the image acquired from the imaging device 2. Here, the preprocessing is to process an image so as to facilitate machine learning described later.
For example, the processing unit 42 extracts an image in which a dangerous spot is shown from a plurality of images acquired from the imaging device 2. Further, the processing unit 42 extracts, for each image, a risk factor that causes a dangerous event at the dangerous part as a feature amount from the extracted image showing the dangerous part.

  The model creation unit 43 creates a dangerous spot detection model for detecting a dangerous spot at the work site by executing machine learning using the image preprocessed by the processing unit 42 as learning data. Machine learning in the present embodiment is supervised learning, for example. FIG. 2 is a diagram illustrating an example of learning data according to an embodiment of the present invention.

  As shown in FIG. 2, the learning data is data in which information (correct answer label) indicating a dangerous event at the dangerous location is attached to an image (input image) of the dangerous location at the work site. For example, when an opening is reflected in the input image preprocessed by the processing unit 42, a correct answer label indicating that the dangerous event is “falling” is attached to the image. For example, when a worker who is not wearing a safety belt is shown in the input image pre-processed by the processing unit 42, a correct answer label indicating that the dangerous event is “no safety belt” is attached to the image. Has been. Further, for example, when a crane work machine that lifts a suspended load is shown in the input image preprocessed by the processing unit 42, the correct label indicating that the dangerous event is “dropped suspended load” in the image. Is attached. For example, when an input image preprocessed by the processing unit 42 shows a state in which objects are piled up, a correct answer label indicating that the dangerous event is “collapse” is attached to the image. . The process of attaching a correct answer label indicating a dangerous event to the input image may be performed by the user. In this case, for example, an input unit is provided in the model creation unit 43. The input unit receives a user instruction from the user. This user instruction instructs which danger event correct answer label is attached to which input image. The input unit may be a keyboard, a touch panel, or a microphone.

  The model creation unit 43 performs machine learning using learning data in which an input image that is an image of a dangerous spot at a work site is attached with a correct answer label that is information indicating a dangerous event of the dangerous spot. Create a dangerous spot detection model to detect a spot. In the present embodiment, the model creation unit 43 performs machine learning by deep learning using a convolutional neural network.

  For example, the risk location detection model is a model of risk factors that cause a dangerous event. In this dangerous spot detection model, risk factors are modeled for each of various dangerous events. For example, when the dangerous event is “fall”, the risk factor is that there is an opening. For example, when the dangerous event is “no safety belt”, the risk factor is that the worker does not wear the safety belt. For example, when the dangerous event is “dropping a suspended load”, the risk factor is that there is a crane working machine that lifts the suspended load. For example, when the dangerous event is “collapse”, the risk factor is that there is a state where things are piled up.

  The model creation unit 43 transmits the created dangerous spot detection model to the terminal device 5 via the communication network 3. In this case, the information on the dangerous event is included in the dangerous point detection model. If the dangerous part detection model created by the model creation unit 43 can be transferred to the terminal device 5 using a portable recording medium, the model creation unit 43 (learning server 4) and the terminal The communication network 3 with the device 5 can be omitted. For example, the storage medium is an externally mounted storage medium such as a hard disk, a magnetic disk, a magneto-optical disk, an optical disk, a USB memory, or a memory card.

  The terminal device 5 is, for example, a terminal used by a site supervisor who is a user. For example, the terminal device 5 is a mobile phone such as a smartphone, a tablet terminal, a glasses-type wearable device, a PC (Personal Computer), or the like.

As illustrated in FIG. 1, the terminal device 5 includes a communication unit 51, an imaging unit 52, a detection unit 53, a display control unit 54, and a display unit 55.
The communication unit 51 performs wireless or wired communication with the learning server 4 via the communication network 3.
The imaging unit 52 captures the work site as a captured image. The display unit 55 displays the captured image captured by the imaging unit 52.

The detection unit 53 detects the presence or absence of a dangerous location in the captured image of the work site displayed on the display unit 55 using the dangerous location detection model acquired from the model creation unit 43 via the communication unit 51.
For example, the detection unit 53 determines in the captured image of the work site displayed on the display unit 55 whether or not there is a location corresponding to the risk factor modeled by the dangerous location detection model. The presence / absence of a dangerous spot in the captured image of the work site displayed at 55 is detected. That is, the detection unit 53 is displayed on the display unit 55 when there is a part corresponding to the risk factor modeled in the dangerous part detection model in the captured image of the work site displayed on the display unit 55. It is determined that a dangerous location exists in the captured image at the work site.

  If the detection unit 53 determines that there is a location corresponding to the risk factor, the display control unit 54 displays the warning information 6 at the location. In this case, the warning information is information on a dangerous event caused by the risk factor. For example, if the detection unit 53 determines that there is a risk factor modeled in the dangerous part detection model, for example, a part with an opening, in the captured image of the work site displayed on the display unit 55, The characters “Fall” are displayed on the display screen of the display unit 55 corresponding to the location. FIG. 3 is an image diagram in which the field supervisor images the work site using the terminal device 5. As shown in FIG. 3, the site supervisor copies a captured image of the work site on the display unit 55 of the terminal device 5 and mechanically finds a dangerous spot on the work site by checking the captured image. Can do.

  Below, the flow of the process which discovers the dangerous spot of the dangerous spot discovery system 1 which concerns on one Embodiment of this invention is demonstrated. FIG. 4 is a diagram for explaining the flow of processing for finding a dangerous spot in the dangerous spot finding system 1 according to the embodiment of the present invention.

The imaging device 2 images a work site such as a construction site (step S101). This captured image is an image for learning. Then, the imaging device 2 transmits the captured image to the learning server 4 via the communication network 3.
The processing unit 42 acquires a captured image from the imaging device 2 via the communication unit 41 (step S102). And the process part 42 pre-processes the image acquired from the imaging device 2 (step S103). For example, the processing unit 42 extracts, for each image, a risk factor that causes a dangerous event in the dangerous part as a feature amount from an image showing the extracted dangerous part.

The model creation unit 43 creates a dangerous location detection model for detecting a dangerous location at the work site by executing machine learning using the image preprocessed by the processing unit 42 as learning data (step S104). . In the learning data, for example, information indicating a dangerous event at the dangerous place is attached to an image of the dangerous place at the work site. That is, the model creation unit 43 performs supervised learning and performs machine learning by deep learning using a convolutional neural network.
The model creation unit 43 transmits the created dangerous spot detection model to the terminal device 5 via the communication network 3.

The terminal device 5 acquires the dangerous spot detection model created by the model creation unit 43 via the communication network 3 (step S105).
The imaging unit 52 images the work site by a user operation such as work supervision (step S106). The display unit 55 displays the captured image captured by the imaging unit 52.

  The detection unit 53 detects the presence / absence of a dangerous location in the captured image of the work site displayed on the display unit 55 using the dangerous location detection model acquired from the model creation unit 43 via the communication unit 51 (step S107). . For example, the detection unit 53 determines whether there is a location corresponding to the risk factor modeled by the dangerous location detection model in the captured image of the workplace displayed on the display unit 55. Detect the presence or absence of dangerous spots. For example, when the captured image of the work site displayed on the display unit 55 is present at a location corresponding to the opening, the detection unit 53 determines that there is a “falling” dangerous location at the work site. In addition, for example, the detection unit 53, in the captured image of the work site displayed on the display unit 55, when there is a location corresponding to a worker who is not wearing a safety belt, It is determined that there is a danger spot “None”. In addition, for example, when a portion corresponding to the crane work machine that lifts the suspended load exists in the captured image of the work site displayed on the display unit 55, the detection unit 53 performs “hanging” at the work site. It is determined that there is a danger point of “load drop”. In addition, for example, when there is a portion corresponding to a state in which objects are piled up in the captured image of the work site displayed on the display unit 55, the detection unit 53 has a risk of “collapse” at the work site. It is determined that the location exists.

  For example, if the detection unit 53 determines that there is a location corresponding to the risk factor described above, the display control unit 54 displays the risk of the risk factor on the display screen of the display unit 55 corresponding to the location. The event is displayed as warning information (step S108).

  As described above, the model creation unit 43 according to an embodiment of the present invention performs machine learning using learning data in which information indicating a dangerous event at a dangerous place is attached to an image of a dangerous place at a work site. Accordingly, a model creation unit 43 that creates a dangerous spot detection model for detecting the dangerous spot is provided. As a result, the dangerous spot detection model can be mechanically found from an image captured on the terminal device 5 by causing the terminal device 5 such as a tablet terminal or a glasses-type wearable device to execute as a program. It is.

  As described above, since a worker or a user such as a site supervisor can mechanically find a dangerous spot from an image photographed on the terminal device 5 even in a busy work spot, he / she overlooks the dangerous spot on the work spot. It can be easily discovered without any problems. Further, since the dangerous spot can be found mechanically by using the dangerous spot detection model, it is also possible to find a dangerous spot that has been missed only by human eyes. In addition, it is possible to quickly notify a person who is not familiar with the work site (for example, a new visitor, a visitor, etc.).

  In the above-described embodiment, as a risk factor that causes a dangerous event at a dangerous location, one risk factor is included in one dangerous event (for example, if the dangerous event is “fall”, the risk factor is an opening). However, the present invention is not limited to this. That is, the model creation unit 43 may extract a plurality of feature amounts such as the size and position of the opening from the learning data, and create a dangerous location detection model using the extracted feature amounts.

  Further, in the above-described embodiment, the detection unit 53 has a feature amount of an image captured at the work site by the imaging unit 52 of the terminal device 5 and a feature amount modeled by the dangerous spot detection model being a certain level or more. When there is a similarity or a matching point of a certain level or more, it may be determined that there is a dangerous place in the captured image at the work site.

  You may make it each implement | achieve each of the learning server 4 and the terminal device 5 in embodiment mentioned above with a computer. In that case, a program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be read into a computer system and executed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory inside a computer system serving as a server or a client in that case may be included and a program held for a certain period of time. Further, the program may be a program for realizing a part of the above-described functions, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system. You may implement | achieve using programmable logic devices, such as FPGA (Field Programmable Gate Array).

  For example, the learning server 4 according to the present embodiment may be configured by installing a program for performing various processes for controlling the model creating unit 43 and causing the computer to execute the program. Also. For example, the terminal device 5 according to the present embodiment may be configured by installing a program for performing various processes for controlling the detection unit 53 and the display control unit 54 and causing the computer to execute the program.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

DESCRIPTION OF SYMBOLS 1 Dangerous part detection system 2 Imaging device 3 Communication network 4 Learning server 5 Terminal device 41, 51 Communication part 42 Processing part 43 Model preparation part 52 Imaging part 53 Detection part 54 Display control part 55 Display part

Claims (11)

  A model that creates a dangerous part detection model for detecting a dangerous part by executing machine learning using learning data with information indicating a dangerous event of the dangerous part in the image of the dangerous part at the work site A model creation device including a creation unit.   The model creation apparatus according to claim 1, wherein the risk location detection model is a model of a risk factor that causes the dangerous event.   The model creation device according to claim 1, wherein the model creation unit performs the machine learning by deep learning using a convolutional neural network. An imaging unit for imaging a work site as a captured image;
A display unit for displaying the captured image;
The presence or absence of a dangerous place in the captured image of the work site displayed on the display unit is detected using the dangerous place detection model generated by the model creation device according to any one of claims 1 to 3. A detector to perform,
A terminal device comprising: The dangerous point detection model is a model of risk factors that cause the dangerous event,
The detector is
In the captured image of the work site displayed on the display unit, the danger in the captured image of the work site displayed on the display unit by determining whether or not there is a location corresponding to the modeled risk factor The terminal device according to claim 4, wherein presence / absence of a location is detected.   If it is determined that there is a location corresponding to the modeled risk factor, the terminal device further includes a display control unit that displays warning information indicating that the location is the dangerous location on the display unit. The terminal device according to claim 5 provided.   The terminal device according to claim 6, wherein, when it is determined that there is a location corresponding to the modeled risk factor, the display control unit displays the warning information at the location.   The terminal device according to claim 7, wherein the warning information is information on the dangerous event caused by the risk factor. The model creation device according to any one of claims 1 to 3,
The terminal device according to any one of claims 4 to 8,
Hazardous area detection system equipped with.   The program for functioning a computer as a model production apparatus as described in any one of Claims 1-3.   The program for functioning a computer as a terminal device as described in any one of Claims 4-8.

Images