TWI836618B - Cloud system for industrial safety, method for detecting industrial safety event and computer-readable medium - Google Patents

Abstract

The present invention discloses a cloud system for industrial safety, a method for detecting industrial safety event and a computer-readable medium using artificial intelligence image recognition technology, which basically allows users to connect to the cloud system from a remote network through a browser of a terminal setting device, so as to set and select the desired industrial safety detection service. Users only need to provide their own network cameras and terminal alarm devices, and then they can use the related modules of the industrial safety cloud, including a setting and selection interface module, a surveillance area framing module, a human body posture detection module, a multi-object detection module, and an alarm output module, to automatically and quickly achieve various industrial safety detection services.

Description

Industrial safety cloud system, industrial safety incident detection method and computer-readable media

The present invention relates to an industrial safety image detection technology that combines artificial intelligence (AI) image detection and recognition with network cloud technology applications, and in particular to an industrial safety cloud system, industrial safety event detection method and computer-readable medium that apply the industrial safety image detection technology.

In the past, in order to monitor the safety of construction sites, cameras were used to transmit images to surveillance cameras to reduce accidents at construction sites. However, it is impossible for people to monitor the screen 24/7 to prevent unsafe behaviors at construction sites. Therefore, a technology was proposed to form a virtual warning zone based on the position of the pyramid at the site in the image, and the machine vision automatically detects the entry and exit of personnel and vehicles in this virtual warning zone, so as to form an electronic fence.

In addition, in order to ensure whether the construction environment is safe and whether construction workers are wearing appropriate safety devices, the competent authorities often require construction workers to take photos of the on-site environment to prove the safety of the construction environment. This approach not only requires a lot of manpower or time for filming, but also makes it impossible to monitor the construction site in real time. For example, some construction workers may remove their safety devices after the filming work is completed, thus creating a loophole in industrial safety. To this end, there is a technology that proposes to use images or images of the construction area to Determining whether the construction site is safe includes determining whether the construction personnel's behavior complies with the safety code of conduct, determining whether the construction personnel's wearing devices comply with the wearing device specifications, and determining whether the protective equipment on site complies with the protective equipment specifications.

However, these previous technologies can only be debugged and set by professional engineers for specific construction environments. Their installation costs are high and time-consuming, and they cannot automatically and quickly provide various industrial safety detection services.

The purpose of the present invention is to provide a cloud service that provides the most complete and best industrial safety image monitoring technology for general users to select and debug by themselves, so that automatic industrial safety image monitoring becomes an instant, convenient and usable service.

The present invention provides an industrial safety cloud system using artificial intelligence image recognition technology. Basically, the relevant modules of the industrial safety cloud are used, including a setting selection interface module, a warning zone framing module, a multi-object detection module, and a human posture detection module. detection module and alarm output module, and cooperate with terminal setting devices, Internet Protocol (IP) network cameras and network alarm devices in specific areas to automatically and quickly achieve various industrial safety detections service.

According to an implementation aspect, the present invention provides an industrial safety cloud system for receiving real-time captured on-site images of a construction site, and performing safety monitoring and alarming based on the on-site images. The industrial safety cloud system includes: a warning zone framing module , used to frame the warning area in the on-site image according to user settings; the human posture detection module is used to detect the postures and movements of the construction workers in the warning area; the multi-object detection module is used to detect the Prescribed objects worn by construction workers in the warning area; and an alarm output module used to determine the alarm based on at least one of the posture, the action and the prescribed object according to the user settings. Whether an industrial safety incident occurs in the warning zone, so that when an industrial safety incident occurs, an alarm message will be sent according to the user's settings.

According to another implementation aspect, the present invention provides a method for detecting industrial safety incidents, which includes: receiving real-time captured on-site images of a construction site; framing a warning area in the on-site image according to user settings; and detecting construction in the warning area. The posture and movement of the personnel; detecting the prescribed objects worn by the construction workers in the warning area; according to the user settings, judging whether work has occurred in the warning area based on at least one of the posture, the movement and the prescribed object. safety incident; and when it is determined that the industrial safety incident occurs, an alarm message is sent according to the user's settings.

According to another implementation aspect, the present invention provides a computer-readable medium, which is used in a computing device, computer or server and stores instructions to execute the above-mentioned industrial safety incident detection method.

10: Gong'an Cloud

11: Set up the optional interface module

12: Warning zone frame module

13:Human posture detection module

14:Multiple object detection module

15: Alarm output module

20: Terminal setting device

30: Webcam

40: Terminal alarm device

111:Network camera settings

112: Terminal alarm device setting

113: Warning zone frame selection and setting

114: Human posture detection selection settings

115:Multiple object detection model selection settings

116: Alarm output selection setting

121: Full image warning zone setting

122: Manual frame warning zone setting

123: Automatically detect and frame triangular cones to form a warning zone

131: Selection of human posture detection model

132: Fall detection selected

133: Climbing detection selection

141: Multiple object detection type settings

142: Selection of multi-object detection model

143:Multiple object detection model library

Figure 1 is a diagram of the architecture of the industrial security cloud system based on artificial intelligence image recognition technology of the present invention.

FIG. 2 is an explanatory diagram of the setting and selection interface module of the present invention.

Figure 3 is an illustration of the warning zone frame module of the present invention.

Figure 4 is an explanatory diagram of the human body posture detection module of the present invention.

Figure 5 is an illustration of the multi-object detection module of the present invention.

The following describes the technical content of the present invention through a specific concrete implementation form. People familiar with this technology can easily understand the advantages and effects of the present invention from the content disclosed in this manual. However, the present invention can also be implemented or applied through other different specific implementation forms.

Figure 1 is an architectural diagram of the industrial safety cloud system that applies artificial intelligence image recognition according to the present invention. In order to allow general users to choose and debug by themselves, as shown in Figure 1, the industrial safety cloud system uses cloud technology including multiple modules. 10. Cooperate with the terminal setting device 20, the network camera 30, and the terminal alarm device 40 to achieve an industrial safety cloud system that applies artificial intelligence image recognition. The terminal setting device 20, the network camera 30 and the terminal alarm device 40 can all be prepared by the user. Users can use the terminal setting device 20, such as a computer, laptop or mobile device that can connect to the Internet, to use related setting services through website browsing or application (APP) of the Industrial Safety Cloud 10. And according to the previous settings, the real-time on-site images of the construction site taken from the network camera 30 are sent back to the Industrial Safety Cloud 10 through the network for the Industrial Safety Cloud 10 to perform detection and identification, and then the detection and identification results are used according to the previous settings. The terminal alarm device 40 is triggered via the network to activate the alarm. In one embodiment, the network camera 30 can be various IP cameras with a website address, and the terminal alarm device 40 can be an IP electronic signage with a website address, an IP alarm with a website address, and/or a mobile device, etc., wherein , IP electronic signboards and IP alarms are mainly installed at industrial safety sites to directly warn and remind workers at industrial safety sites. Mobile devices such as mobile phones and tablets can use various instant messaging software such as Line or Messenger. Diversified text messages can be used to immediately alert relevant personnel responsible for industrial safety.

In one embodiment, the industrial safety cloud 10 is a cloud service (Cloud Service) that provides customized chemical safety detection, allowing users to remotely connect from the Internet to flexibly set the specific industrial safety detection services required. Mainly Provides mutual communication connection settings and optional interface module 11, warning zone framing module 12, human posture detection module 13, multi-object detection module 14 and alarm output module 15 for connection and control, among which , the industrial safety cloud system is used to receive real-time on-site images of the construction site, and perform customized safety monitoring and alarms based on the on-site images. Each module 11~15 in Industrial Safety Cloud 10 can be implemented as hardware, firmware or software, or as a cloud service or microservice.

As shown in FIG2 , the functions of the configuration selection interface module 11 basically include network camera settings 111, terminal alarm device settings 112, warning zone frame selection and settings 113, human posture detection selection settings 114, multi-object detection model selection settings 115, and alarm output selection settings 116. In one embodiment, the network camera settings 111 can be set to send a camera control command to the network IP address of the network camera 30 at a specific time, so as to send the scene image taken by the network camera 30 back to the industrial security cloud 10 via the network for detection and identification. The terminal alarm device setting 112 is to set the type, alarm message style and network IP address of the terminal alarm device 40, so that the industrial safety cloud 10 can, according to the setting, when a specific event or object that needs to be alarmed is detected, make the alarm output module 15 transmit the alarm message to the terminal alarm device 40 via the network to trigger the alarm. The warning area frame selection and setting 113 provides the relevant settings of the warning area frame module 12. The alarm output selection setting 116 is to cooperate with the human posture detection selection setting 114 and the multi-object detection model selection setting 115 to make the alarm selection setting for specific events or objects, so as to select a specific output alarm style for the specific event or object detection result. The above settings 111~116 can be collectively referred to as user settings. The setting selection interface module 11 can accept the connection request of the user's terminal setting device 20 so that the user can provide user settings through the terminal setting device 20.

As shown in FIG3 , the functions of the warning zone framing module 12 basically include full image warning zone setting 121, manual framed warning zone setting 122, and automatic detection framed triangle to enclose a warning zone 123. Functions 121-123 correspond to three different warning zones, and which warning zone is actually used can be specified by the warning zone framing selection and setting 113. In one embodiment, the full image warning zone setting 121, that is, the entire image area of the on-site image captured by the network camera 30 is set as the warning zone for subsequent industrial safety event detection. Manual framed warning zone setting 122 can be based on the previous setting of the network IP address of the network camera, requesting the network camera 30 to transmit the on-site image, and providing a frame editing interface for the user to manually frame the warning zone for subsequent industrial safety incident detection. Automatic detection and framing of triangles to form a warning zone 123 is to use machine vision technology or other artificial intelligence detection technology to automatically detect the area surrounded by triangles and connecting rods in the on-site image, and automatically frame the area as the warning zone for subsequent industrial safety incident detection.

In addition, in one embodiment, the pyramid and the connecting rod can be extended to any warning object that can be used to define or mark the warning area. In addition to the above-mentioned pyramid and connecting rod, wire mesh, fences, various preset marking objects, ropes or cables of a specific style, or lines of a specific style on the wall or the ground, etc. can all be used as warning objects. If the warning area frame selection and setting 113 selects such a warning area, the warning area frame module 12 uses machine vision technology or other artificial intelligence detection technology to instantly detect the area surrounded by the warning object in the scene image, and uses the area as the warning area. In this way, users can directly arrange warning objects at the construction site to select a warning zone, and users can adjust the warning objects (such as any of the aforementioned warning objects, or a combination of any of the aforementioned warning objects) at any time to change the shape, size or position of the warning zone.

As shown in FIG. 4 , the functions of the human posture detection module 13 basically include human posture detection model selection 131 , fall detection selection 132 , and climbing detection selection 133 . In one embodiment, the human pose detection model selection 131 uses human pose (Human Pose) artificial intelligence deep learning (Deep Learning) recognition technology to identify human posture with its deep learning model, or uses other artificial intelligence detection technologies. To detect if someone falls or makes climbing movements. The user can select an applicable human posture artificial intelligence detection model to detect falls or climbing according to the detection accuracy and speed requirements, and select whether to perform fall detection through the fall detection selection 132 according to the application needs, and Use the climbing detection selection 133 to select whether to perform climbing motion detection, and use the previous settings of the human posture detection selection setting 114 and the alarm output selection setting 116 to perform relevant alarm output to the relevant alarm device.

In addition, in one embodiment, the human posture detection module 13 can detect more types of human postures and human movements, and the user can select dangerous postures and/or dangerous movements that may cause industrial safety incidents (such as climbing, stretching arms, jumping, running, jogging and/or falling postures and/or movements) through the human posture detection selection setting 114. If the human posture detection module 13 detects that the construction personnel in the warning area present the posture and/or movement selected by the human posture detection selection setting 114, the alarm output module 15 determines that an industrial safety incident has occurred, and sends an alarm message according to the alarm output selection setting 116. The alarm output selection setting 116 may include the object and method of the alarm output module 15 to send the alarm message, wherein the object of the alarm message may be one or more terminal alarm devices 40, one or more users, and/or one or more management personnel of the construction site who will receive the alarm message, and the method of sending the alarm message may be a short message of an instant messaging software, a mobile phone text message, and/or a telephone voice call.

As shown in FIG. 5 , the functions of the multi-object detection module 14 basically include multi-object detection type setting 141 , multi-object detection model selection 142 , and multi-object detection model library 143 . In one embodiment, the multi-object detection type setting 141 selects one or more specified objects on the construction workers for image detection and recognition according to the industrial safety needs, and then uses the artificial intelligence detection model selection function in the multi-object detection model selection 142 , select an applicable multi-object artificial intelligence detection model from the multi-object detection model library 143 according to the specified object type, detection accuracy and speed requirements, and use the previous multi-object detection model selection settings 115 and alarm output Select the setting of Setting 116 to do relevant alarm output for the relevant alarm device. Among them, the types of prescribed items may include at least one of the following items, such as construction site hats, reflective clothing, work clothes, construction site shoes, work gloves, and protective glasses, etc. The multi-object artificial intelligence detection models in the multi-object detection model library 143 can use any artificial intelligence detection technology. In the future, new artificial intelligence detection model libraries can continue to be developed according to needs to detect more types of objects or Make more accurate and effective detection.

Specifically, the multi-object detection module 14 can detect which prescribed items are worn by construction workers in the warning area and which prescribed items are not worn, and the user can select and set 115 the selected construction workers must wear through the multi-object detection model. Required objects, such as one of the various stipulated objects mentioned above Kind or variety. If the multi-object detection module 14 detects that the construction workers in the warning area are not wearing the specified objects selected by the multi-object detection model selection setting 115, the alarm output module 15 determines that an industrial safety incident has occurred, and outputs the alarm according to the Select Setting 116 to send an alert message. If there is more than one specified object selected, the user can further select it through the multi-object detection model selection setting 115. As long as the construction workers in the warning zone do not wear any one of the specified objects, it will be judged that an industrial safety incident has occurred, or only in all cases. Only when none of the specified items are worn by construction workers in the warning zone will an industrial safety incident occur.

In one embodiment, the present invention further provides a method for detecting industrial safety incidents corresponding to the industrial safety cloud system. The steps of the method may correspond to the functions of the industrial safety cloud system and each module 11 to 15 in the above embodiments and diagrams.

In one embodiment, the present invention provides a computer-readable medium, such as a memory, a floppy disk, a hard disk or an optical disk. The computer-readable medium can be used in a computing device, computer or server, and stores instructions to execute the above-mentioned industrial safety incident detection method.

Compared with traditional technology, the present invention has the following advantages:

1. Industrial Safety Cloud is a cloud service that provides industrial safety event detection and alarm. Its purpose is to allow users to access the cloud directly from the Internet through a browser to use the specific industrial safety detection service they need, without having to purchase or install software on their own hardware. Therefore, there is no need to purchase and maintain expensive software and hardware and develop and update related advanced technologies. In addition to being able to apply the latest industrial safety detection technology, users can also save a lot of money, time, manpower and maintenance costs.

2. In addition to using machine vision's pyramid and linkage detection analysis to automatically estimate the warning zone, you can also manually frame the warning zone directly in the block diagram editing interface to detect industrial safety incidents.

3. Each function is modular and can be flexibly selected and applied in various scenarios.

4. Users can set the detection method and range by themselves, and can provide customized and differentiated application services according to user needs.

5. New event and object detection functions can continue to be developed, allowing users to add or select more different object detection and event detection.

6. When a situation that does not comply with industrial safety regulations is detected and an alarm is required, users can choose a convenient way to receive the alarm in a variety of ways.

7. It can simultaneously provide real-time monitoring, continuous tracking, fatigue-free and automatic recording of industrial safety violation detection functions.

The above detailed description is a specific description of a feasible embodiment of the present invention, but the embodiment is not intended to limit the patent scope of the present invention. Any equivalent implementation or modification that does not deviate from the technical spirit of the present invention should be included in the patent scope of this case.

10: Gongan Cloud

11: Set up the optional interface module

12: Warning zone framing module

13: Human posture detection module

14:Multiple object detection module

15:Alarm output module

20: Terminal setting device

30: Webcam

40: Terminal alarm device

Claims (7)

An industrial safety cloud system is used to receive real-time on-site images of construction sites, and perform safety monitoring and alarms based on the on-site images. The industrial safety cloud system includes: a setting selection interface module for accepting the user's terminal setting device The connection request is for the user to provide user settings through the terminal setting device; the warning area framing module is used to frame the warning area in the live image according to the user settings; the human posture detection module; the multi-object detection module; and an alarm output module, wherein if the human posture detection module detects that the construction workers in the warning area exhibit the posture and/or action selected by the user setting, or the multi-object detection module If it is detected that the construction worker is not wearing the specified items selected by the user setting, the alarm output module determines that an industrial safety incident occurs in the warning zone, and then sends an alarm message according to the user setting. The industrial safety cloud system as described in claim 1, wherein the warning zone framing module uses the entire image area of the on-site image as the warning zone according to the user setting. The industrial safety cloud system as described in claim 1, wherein the warning zone framing module uses the user's manually framed area in the scene image as the warning zone according to the user's setting. The industrial safety cloud system described in claim 1, wherein the warning area framing module is based on the user setting to instantly detect the area surrounded by warning objects in the scene image, so that the area is used as the warning area. The industrial safety cloud system as described in claim 1, wherein the user settings include the objects and methods for the alarm output module to transmit the alarm message. An industrial safety incident detection method includes: Receive real-time captured on-site images of the construction site; accept the connection request from the user's terminal setting device so that the user can provide user settings through the terminal setting device; frame the warning zone in the on-site image based on the user settings; and if detected If it is detected that the construction worker in the warning area exhibits the posture and/or action selected by the user setting, or it is detected that the construction worker is not wearing the specified items selected by the user setting, then it is determined that an industrial safety incident has occurred in the warning area. , and then send an alarm message according to the user settings. A computer-readable medium, used in a computing device, a computer or a server, stores instructions for executing the industrial safety incident detection method as described in claim 6.

Images