KR20210006722A - Apparatus, method and computer program for determining whether safety equipment is worn - Google Patents

Abstract

Provided is a device for determining whether safety equipment is worn, which comprises: an image receiving unit which receives an image of a worker from a camera; a body unit detection unit which detects a plurality of body parts from the received image based on a previously learned first deep learning model; a safety equipment area detection unit which detects a plurality of safety equipment areas from the received image based on a previously learned second deep learning model; and a determination unit configured to determine whether the worker wears the safety equipment based on the detected plurality of body parts and the plurality of safety equipment areas.

Description

A device, method, and computer program that determines whether or not safety equipment is worn {APPARATUS, METHOD AND COMPUTER PROGRAM FOR DETERMINING WHETHER SAFETY EQUIPMENT IS WORN}

The present invention relates to an apparatus, a method, and a computer program for determining whether to wear safety equipment.

In industrial sites, an average of 7 workers die every day from industrial accidents. Unlike natural disasters, industrial accidents can be prevented in advance if only a few safety rules are followed. Basically, safety rules require workers to wear safety shoes, hard hats, safety clothing, safety devices, and protective gear.

In order to prevent such industrial accidents, a technology for determining whether a worker is wearing safety equipment has been developed, and Korean Patent No. 10-1212657, a prior art, has initiated an intelligent safety control system with respect to the technology for determining whether or not a worker is wearing safety equipment. Are doing.

Conventionally, it was possible to determine whether or not the worker wears safety equipment through the safety equipment attached with the Bluetooth module.

In addition, it was possible to determine whether or not a worker is wearing safety equipment using BLE (Bluetooth Low Energy) technology and a search beacon module.

However, when using Bluetooth, BLE and beacon signals, there is a hassle of attaching a search beacon module to each safety device, and if these modules are attached to small safety equipment such as safety gloves and safety masks, the operator can use safety equipment. It had the disadvantage that it was difficult to work with wearing it.

In addition, when determining whether or not the worker is wearing safety equipment using the signals of these modules, the signal transmission distance between the module transmitting the signal and the worker terminal is simply calculated to determine whether the worker is wearing safety equipment. It had the disadvantage that the judgment result of whether the equipment was worn or not was not accurate.

In addition, additional devices such as a gateway module, a worker terminal, and a manager terminal are required in order to determine whether or not a worker is wearing safety equipment using signals from the modules.

An attempt is made to provide a device, a method, and a computer program for determining whether to wear safety equipment to determine whether a worker is wearing safety equipment using a general RGB camera.

Wearing safety equipment that determines whether the worker wears safety equipment with high accuracy by detecting the operator's body part using the first deep learning model and the safety equipment area to be worn by the operator using the second deep learning model It is intended to provide an apparatus, method, and computer program for determining whether or not.

However, the technical problem to be achieved by this embodiment is not limited to the above-described technical problems, and other technical problems may exist.

As a means for achieving the above-described technical problem, an embodiment of the present invention provides an image receiving unit for receiving an image photographed of a worker from a camera, a plurality of images from the received image based on a pre-learned first deep learning model. A body part detection unit that detects a body part, a safety equipment area detection unit that detects a plurality of safety equipment areas from the received image based on a pre-learned second deep learning model, and the detected plurality of body parts and the plurality of safety It is possible to provide a device for determining whether to wear safety equipment, including a determination unit that determines whether or not to wear the safety equipment for the worker based on the equipment area.

In another embodiment of the present invention, receiving an image of a worker from a camera, detecting a plurality of body parts from the received image based on a pre-learned first deep learning model, and pre-learning second Detecting a plurality of safety equipment areas from the received image based on a deep learning model, and determining whether to wear the safety equipment for the worker based on the detected plurality of body parts and the plurality of safety equipment areas It may provide a method for determining whether to wear safety equipment including the step of.

In another embodiment of the present invention, when the computer program is executed by a computing device, the computer program receives an image of a worker from a camera, and a plurality of body parts from the received image based on a pre-learned first deep learning model. And detecting a plurality of safety equipment areas from the received image based on a pre-learned second deep learning model, and based on the detected plurality of body parts and the plurality of safety equipment areas, A computer program stored in a medium including a sequence of instructions for determining whether the safety equipment is worn may be provided.

The above-described problem solving means are merely exemplary and should not be construed as limiting the present invention. In addition to the above-described exemplary embodiments, there may be additional embodiments described in the drawings and detailed description of the invention.

According to any one of the above-described problem solving means of the present invention, a device for determining whether to wear safety equipment to determine whether or not a worker is wearing safety equipment using a general RGB camera without attaching a separate Bluetooth module to the safety equipment. , Method and computer program can be provided.

By determining whether a worker is wearing safety equipment with only a device for determining whether to wear safety equipment and a general RGB camera, the device, method, and computer program for determining whether to wear safety equipment that does not require additional devices such as gateway beacons, operator terminals, and manager terminals are used. Can provide.

Wearing safety equipment that determines whether the worker wears safety equipment with high accuracy by detecting the operator's body part using the first deep learning model and the safety equipment area to be worn by the operator using the second deep learning model A device, a method, and a computer program may be provided.

It is possible to provide an apparatus, method, and computer program for determining whether to wear safety equipment that enables the operator to determine whether or not the operator is wearing safety equipment even if the operator does not stand facing the front of the camera.

When a camera photographs a plurality of workers, an apparatus, a method, and a computer program for determining whether to wear safety equipment may be provided to simultaneously determine whether or not the plurality of workers wear safety equipment.

1 is a block diagram of an apparatus for determining whether to wear safety equipment according to an embodiment of the present invention.
2 is an exemplary diagram showing a first deep learning model according to an embodiment of the present invention.
3 is an exemplary diagram for explaining a process of detecting a body part from an image photographed of a worker according to an embodiment of the present invention.
4 is an exemplary diagram showing a second deep learning model according to an embodiment of the present invention.
FIG. 5 is an exemplary diagram for explaining a process of determining whether a worker wears safety equipment based on a body part and a safety equipment area according to an embodiment of the present invention.
6 is a flowchart of a method of determining whether to wear safety equipment in the apparatus for determining whether to wear safety equipment according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art can easily implement the present invention. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and similar reference numerals are assigned to similar parts throughout the specification.

Throughout the specification, when a part is said to be "connected" to another part, this includes not only "directly connected" but also "electrically connected" with another element interposed therebetween. . In addition, when a part "includes" a certain component, it means that other components may be further included, and one or more other features, not excluding other components, unless specifically stated to the contrary. It is to be understood that it does not preclude the presence or addition of any number, step, action, component, part, or combination thereof.

In the present specification, the term "unit" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Further, one unit may be realized using two or more hardware, or two or more units may be realized using one hardware.

In the present specification, some of the operations or functions described as being performed by the terminal or device may be performed instead by a server connected to the terminal or device. Likewise, some of the operations or functions described as being performed by the server may also be performed by a terminal or device connected to the server.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an apparatus for determining whether to wear safety equipment according to an embodiment of the present invention. Referring to FIG. 1, the apparatus 100 for determining whether to wear safety equipment includes an image receiving unit 110, a body part detecting unit 120, a safety equipment area detecting unit 130, a determining unit 140, and a control unit 150. I can.

The image receiving unit 110 may receive an image of a worker from a camera. For example, the image receiving unit 110 may receive an image in which the entire body of a worker is photographed from a general RGB camera.

The body part detection unit 120 may detect a plurality of body parts from the received image based on the pre-learned first deep learning model. The first deep learning model will be described in detail with reference to FIG. 2.

2 is an exemplary diagram showing a first deep learning model according to an embodiment of the present invention. Referring to FIG. 2, the first deep learning model may be pre-trained to estimate a worker's pose.

The pre-learned first deep learning model 210 uses the received image as an input 200, and a connection line between at least one key point 230, 231 and at least one key point 230, 231 among a plurality of body parts ( 232) can be output 220.

For example, the pre-learned first deep learning model can take as an input a 3-channel color image composed of an image (RGB) photographed of a worker received by the image receiving unit 110 and adjust the size of the image (HxW). have. The size of the image may be adjusted to H=368 and W=W/H*368 so that H is 368.

The pre-learned first deep learning model may extract features of an image by appropriately applying a convolution layer, PReLU, and pooling. For example, in the first deep learning model, a 3x3 kernel size is applied to the convolution layer, and max pooling of 2x2 and 2x2 kernel sizes is applied to the pooling. Image features can be extracted using (Max Pooling).

Thereafter, the first deep learning model previously learned extracts a feature for detecting at least one key point among a plurality of body parts by appropriately applying a convolution layer, a PReLU, and a concatenation, and then the connection line It is possible to extract features for detecting. In this case, the previously learned first deep learning model may be repeatedly applied in a format consisting of a convolution layer, PReLU, and concatenation, for example, six times.

For example, in a format consisting of a convolution layer, PReLU, and concatenation, the structure of a convolution layer and PReLU are connected three times, and feature maps extracted from PReLU are concatenated. ) Can be performed. For example, when 128 feature maps are extracted for each Conv→PReLU structure, the result of concatenation can be derived as '128+128+128=384'.

The pre-learned first deep learning model applies a 1x1 kernel size convolution layer and PReLU to the features of strong key points extracted through several iterations, and then the 1x1 kernel size. 21 feature maps may be extracted by reapplying the convolution layer. Here, the 21 feature maps may include 20 key points and 1 negative.

The pre-learned first deep learning model applies a 1x1 kernel size convolution layer and PReLU to the robust connection line features extracted through several iterations, and then a 1x1 kernel size convolution. 38 feature maps can be extracted by re-applying the layer (Convolution Layer). Here, the 38 feature maps may be composed of a total of 19x2=38 as a probability feature map of a connection line from key point A to a key point B for 19 connection lines and a probability feature map of a connection line from key point B to key point A.

The pre-learned first deep learning model may detect key points and connection lines from the final 59 feature maps (21+38=59). In this case, in the case of a key point, the location of the local maximum in the key point feature maps (20) of each body part becomes the location of the key point, and the negative feature map may be excluded. In addition, in the case of a connection line, when two key points derived earlier are substituted in two feature maps representing one connection line, the connection line can be detected by connecting the corresponding two key points belonging to the portion with the highest connection line probability. .

Returning to FIG. 1 again, the body part detection unit 120 may detect the body part based on the detected at least one key point and a connection line between the at least one key point. The process of detecting the body part will be described in detail with reference to FIG. 3.

3 is an exemplary diagram for explaining a process of detecting a body part from an image photographed of a worker according to an embodiment of the present invention. Referring to FIG. 3, the body part detection unit 120 includes a head 300, a nose 301, a neck 303, a right shoulder 305, a left shoulder 306, a right elbow, a left elbow, and a right wrist ( 311), left wrist (314), right fingertip (312), left fingertip (315), lower abdomen (308), right pelvis (307), left pelvis (309), right knee (317), left knee (321) , The right ankle 318, the left ankle 322, the right toe 319, the left toe 323 detects key points corresponding to, and calculates an area for each body part based on the detected at least one key point and a connection line. I can.

The area for each body part is the upper face area 302, the lower face area 304, the trunk area 310, the right hand area 313, the left hand area 316, the right lower leg area 320, and the left lower leg area. (324) may be included.

The area above the face 302 is composed of a square area circumscribed to a circle having the connection line between the head key point 300 and the nose key point 301 as a diameter, and the area under the face 304 is the nose key point 301 and the neck key point. (303) is composed of a square area circumscribed to a circle having a connection line as a diameter, and the trunk area 310 is a right shoulder key point 305, a neck key point 303, a left shoulder key point 306, and a right pelvic key point 307 ), the bottom key point 308, the left pelvic key point 309 is composed of the smallest rectangular area including all, the right hand area 313 is a diameter connecting line between the right wrist key point 311 and the right fingertip key point 312 Consisting of a square area circumscribed to a circle, the left hand area 316 is composed of a square area circumscribed to a circle whose diameter is the connection line between the left wrist key point 314 and the left hand fingertip key point 315, The area 320 is composed of the smallest rectangular area extending the base to include the right toe key point 319 from a square circumscribed to a circle having a diameter of the connecting line between the right knee key point 317 and the right ankle key point 318, and , The left lower leg area 324 is the smallest rectangle in which the base is extended to include the left toe key point 323 from a square circumscribed to a circle having a connection line between the left knee key point 321 and the left ankle key point 322 as a diameter. It can be composed of areas.

The body part detection unit 120 may detect a total of 7 body parts, such as an upper face, a lower face, a trunk, a right hand, a left hand, a right lower leg, and a left lower leg, based on the calculated area for each body part.

The body part detection unit 120 may be set by a manager to determine whether to wear safety equipment. For example, the administrator sets body part detection to on for the top, bottom, right lower leg, and left lower leg among a total of 7 body parts, and detects body parts for the torso, right hand, and left hand. It can be set to off. In this case, the body part detection unit 120 may prevent proceeding to the next step when the body part region set to on is not detected.

Returning to FIG. 1 again, the safety equipment area detection unit 130 may detect a plurality of safety equipment areas from the received image based on the previously learned second deep learning model.

Here, the safety equipment may include, for example, a hard hat, a safety vest, a mask, safety gloves, and safety boots, and the second deep learning model is used to detect a safety equipment area through instance segmentation. Can be pre-learned. Here, instance segmentation refers to an example of an object detection method that distinguishes different instances of the same class. For example, when an image related to a cycling race is learned using a deep learning model, the conventional semantic segmentation distinguishes only the bicycle and the athlete from the image related to the cycling race, whereas the instance segmentation is the athlete 1 and the athlete. Objects can be detected by classifying different instances for the same class, such as 2, player 3,..., bike 1, bike 2, bike 3, .... If the type of safety equipment is added, the second deep learning model may perform additional learning on image data related to the safety equipment to be added.

The safety equipment area detection unit 130 may detect a safety equipment identification area in which an object corresponding to the safety equipment is classified in pixel units and a rectangular area surrounding the safety equipment identification area as a safety equipment candidate area.

The safety equipment area detection unit 130 may be set by a manager to determine whether to wear safety equipment. For example, the manager may set whether to wear safety equipment for hard hats, safety gloves, and safety boots to on, and set whether to wear safety equipment for vests and masks to off.

Here, the pre-learned second deep learning model will be described in detail with reference to FIG. 4.

4 is an exemplary diagram showing a second deep learning model according to an embodiment of the present invention. Referring to FIG. 4, the second deep learning models 410 and 420 that have been previously learned are input 400 as an image received from a camera in the image receiving unit 110, and a plurality of safety equipment candidate areas 411 in a rectangular area. , A first sub deep learning model 410 outputting 412 and a second sub deep learning model 420 classifying safety equipment for each pixel in a corresponding rectangular area.

The pre-learned second deep learning models 410 and 420 may input an image of each of the plurality of safety equipment candidate regions 411 and 412 and output at least one safety equipment identification region 430. Here, the pre-learned second deep learning models 410 and 420 identify candidate areas 411 and 412 of a plurality of safety devices using the first sub deep learning model 410, and the second sub deep learning model Using 420, the safety equipment identification region 430 may be detected and output through the classification of the corresponding safety equipment for each pixel of the candidate regions 411 and 412.

Returning to FIG. 1 again, the determination unit 140 may determine whether to wear safety equipment for an operator based on the detected plurality of body parts and a plurality of safety equipment areas. Here, a process of determining whether the safety equipment is worn will be described in detail with reference to FIG. 5.

FIG. 5 is an exemplary diagram for explaining a process of determining whether a worker wears safety equipment based on a body part and a safety equipment area according to an embodiment of the present invention.

Referring to FIG. 5, the determination unit 140 may determine whether to wear safety equipment for a worker based on a plurality of body parts and a plurality of safety equipment areas.

The determination unit 140 may further include an overlap area detection unit 141 and a calculation unit 142.

The overlapping area detection unit 141 may detect at least one overlapping area between a plurality of body parts and a plurality of safety equipment areas.

For example, the overlapping area detection unit 141 includes an area above the face 500 and a hard hat detection area 501, an area under the face 510 and a mask detection area 511, a body area 520, and a safety vest detection area ( 521), right hand area 530 and safety gloves detection area, left hand area 540 and safety gloves detection area, right lower leg area 550 and safety boots detection area 551, left lower leg area 560 and safety An overlapping area between the boots detection areas 561 may be detected.

In this case, the safety glove detection area of the safety equipment area does not separate the right-hand area and the left-hand area, and the safety boots detection area 561 does not separate the right lower leg area and the left lower leg area.

For example, the overlapping area detection unit 141 detects the hard hat identification area 570 through the upper face area 500 and the hard hat detection area 501, and identifies the area under the face 510 and the mask detection area 511. The mask identification area 571 is detected through the body area 520 and the safety vest detection area 521, the safety vest identification area 572 is detected, and the right lower leg area 550 and the safety boots detection area ( The right lower leg safety boots identification area 573 may be detected through 551, and the left lower leg safety boots identification area 574 may be detected through the left lower leg area 560 and the safety boots detection area 561. .

The calculation unit 142 may calculate a determination value for whether the safety equipment is worn based on at least one safety equipment identification area included in the detected at least one overlapping area and a width of the safety equipment identification area among the overlapping areas. . The calculation unit 142 may calculate a determination value using Equation 1 below. Here, the area of the safety equipment identification area among the overlapping areas may be derived based on the number of pixels in the overlapping area.

The determination unit 140 may determine whether to wear safety equipment based on whether the calculated determination value is equal to or greater than a preset threshold value. For example, assume that the preset threshold is set to 0.5. When the determination value calculated between the area on the face and the hard hat detection area is 0.8, the determination unit 140 may determine that the worker is wearing a hard hat. For another example, when the determination value calculated between the body region and the safety vest detection region is 0.4, the determination unit 140 may determine that the worker has not worn the safety vest. For another example, when the safety equipment area detected by the safety equipment area detection unit 130 does not exist, the determination unit 140 may determine that the worker does not wear any safety equipment.

The control unit 150 may control a work site access system so that a worker can access the work site based on whether or not the determined safety equipment is worn. For example, the controller 150 may control the work site access system so that the operator can access the work site when all the safety equipment set by the manager is worn by the operator. For another example, the control unit 150 may control the work site access system so that the operator cannot enter the work site if any of the safety equipment set by the manager is not worn by the operator.

The apparatus 100 for determining whether the safety equipment is worn may be executed by a computer program stored in a medium including a sequence of instructions for determining whether the safety equipment is worn. When the computer program is executed by a computing device, it receives an image of a worker from a camera, detects a plurality of body parts from the received image based on a pre-learned first deep learning model, and a pre-learned second deep learning model. Includes a sequence of commands that detects a plurality of safety equipment areas from the received image based on the running model, and determines whether or not the worker is wearing safety equipment based on the detected plurality of body parts and the plurality of safety equipment areas can do.

6 is a flowchart of a method of determining whether to wear safety equipment in the apparatus for determining whether to wear safety equipment according to an embodiment of the present invention. The method of determining whether the safety equipment is worn in the apparatus 100 for determining whether to wear the safety equipment shown in FIG. 6 is time-series by the apparatus 100 for determining whether to wear the safety equipment according to the embodiments shown in FIGS. 1 to 5. It includes steps that are processed as. Therefore, even if omitted below, the method of determining whether the safety equipment is worn by the apparatus 100 for determining whether to wear the safety equipment according to the embodiment illustrated in FIGS. 1 to 5 is also applied.

In step S610, the apparatus 100 for determining whether to wear safety equipment may receive an image of a worker from a camera.

In operation S620, the apparatus 100 for determining whether to wear safety equipment may detect a plurality of body parts from the received image based on the previously learned first deep learning model.

In operation S630, the apparatus 100 for determining whether to wear safety equipment may detect a plurality of safety equipment areas from the received image based on the previously learned second deep learning model.

In operation S640, the apparatus 100 for determining whether to wear safety equipment may determine whether to wear safety equipment for an operator based on the detected plurality of body parts and a plurality of safety equipment areas.

In the above description, steps S610 to S640 may be further divided into additional steps or combined into fewer steps, according to an embodiment of the present invention. In addition, some steps may be omitted as necessary, or the order of steps may be switched.

A method of determining whether to wear safety equipment in the apparatus for determining whether to wear safety equipment described with reference to FIGS. 1 to 6 is a form of a recording medium including a computer program stored in a medium executed by a computer or instructions executable by a computer. It can also be implemented as In addition, the method of determining whether to wear safety equipment in the apparatus for determining whether to wear safety equipment described through FIGS. 1 to 6 may be implemented in the form of a computer program stored in a medium executed by a computer.

Computer-readable media can be any available media that can be accessed by a computer, and includes both volatile and nonvolatile media, removable and non-removable media. Further, the computer-readable medium may include a computer storage medium. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data.

The above description of the present invention is for illustrative purposes only, and those of ordinary skill in the art to which the present invention pertains will be able to understand that it can be easily modified into other specific forms without changing the technical spirit or essential features of the present invention. will be. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not limiting. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as being distributed may also be implemented in a combined form.

The scope of the present invention is indicated by the claims to be described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention do.

100: device for determining whether to wear safety equipment
110: image receiving unit
120: body part detection unit
130: safety equipment area detection unit
140: judgment unit
141: overlap area detection unit
142: calculation unit
150: control unit

Claims (19)

In the device for determining whether to wear safety equipment,
An image receiving unit for receiving an image of a worker from a camera;
A body part detection unit that detects a plurality of body parts from the received image based on a pre-learned first deep learning model;
A safety equipment area detection unit that detects a plurality of safety equipment areas from the received image based on a pre-learned second deep learning model; And
A device for determining whether to wear safety equipment, including a determination unit that determines whether or not to wear the safety equipment for the worker based on the detected plurality of body parts and the plurality of safety equipment areas.
The method of claim 1,
The pre-learned first deep learning model inputs the received image and outputs at least one key point among the plurality of body parts and a connection line between the at least one key point.
The method of claim 2,
The body part detection unit detects the body part based on the detected at least one key point and a connection line between the at least one key point.
The method of claim 1,
The pre-learned second deep learning model includes a first sub deep learning model that receives the received image as an input and outputs a plurality of safety equipment candidate regions, and a first sub-deep learning model that classifies the outputted plurality of safety equipment candidate regions for each pixel. A device for determining whether to wear safety equipment, including a 2 sub deep learning model.
The method of claim 4,
The pre-learned second deep learning model inputs an image for each of the plurality of safety equipment candidate regions and outputs at least one safety equipment identification region.
The method of claim 5,
The determining unit further comprises an overlapping area detection unit detecting at least one overlapping area between the plurality of body parts and the plurality of safety equipment candidate areas.
The method of claim 6,
The determination unit further comprises a calculator configured to calculate a determination value for whether the safety equipment is worn based on the at least one safety equipment identification region included in the detected at least one overlapping region and the number of pixels in the overlapping region. The device for determining whether to wear safety equipment.
The method of claim 7,
The determining unit is to determine whether the safety equipment is worn based on whether the calculated determination value is equal to or greater than a preset threshold value.
The method of claim 8,
A device for determining whether to wear safety equipment, further comprising a control unit for controlling a work site access system so that the worker can access the work site based on whether or not the safety equipment is worn.
In a method of determining whether to wear safety equipment in a device for determining whether to wear safety equipment,
Receiving an image of a worker from a camera;
Detecting a plurality of body parts from the received image based on a pre-learned first deep learning model;
Detecting a plurality of safety equipment areas from the received image based on a pre-learned second deep learning model; And
And determining whether to wear the safety equipment for the worker based on the detected plurality of body parts and the plurality of safety equipment areas.
The method of claim 10,
The pre-learned first deep learning model inputs the received image and outputs at least one key point among the plurality of body parts and a connection line between the at least one key point.
The method of claim 11,
The step of detecting the body part,
And detecting the body part based on the detected at least one key point and a connection line between the at least one key point.
The method of claim 10,
The pre-learned second deep learning model includes a first sub deep learning model that receives the received image as an input and outputs a plurality of safety equipment candidate regions, and a first sub-deep learning model that classifies the outputted plurality of safety equipment candidate regions for each pixel. 2 A method of determining whether to wear safety equipment, including a sub deep learning model.
The method of claim 13,
The pre-learned second deep learning model inputs an image for each of the plurality of safety equipment candidate regions and outputs at least one safety equipment identification region.
The method of claim 14,
The step of determining whether the safety equipment is worn,
And detecting at least one overlapping area between the plurality of body parts and the plurality of safety equipment candidate areas.
The method of claim 15,
The step of determining whether the safety equipment is worn,
Comprising the step of calculating a determination value for whether the safety equipment is worn based on the at least one safety equipment identification area included in the detected at least one overlapping area and the number of pixels in the overlapping area How to determine if the equipment is worn.
The method of claim 16,
The step of determining whether the safety equipment is worn,
And determining whether to wear the safety equipment based on whether the calculated determination value is equal to or greater than a preset threshold.
The method of claim 17,
The method of determining whether to wear safety equipment further comprising the step of controlling a work site access system so that the worker can access the work site based on whether the safety equipment is worn.
A computer program stored in a computer-readable medium comprising a sequence of instructions for determining whether to wear safety equipment,
When the computer program is executed by a computing device,
Receive an image of the operator from the camera,
Detecting a plurality of body parts from the received image based on the pre-learned first deep learning model,
Detecting a plurality of safety equipment areas from the received image based on the pre-learned second deep learning model,
A computer program stored in a medium comprising a sequence of instructions for determining whether to wear the safety equipment for the worker based on the detected plurality of body parts and the plurality of safety equipment areas.

Images