Method and system for preventing work accidents in real time based on three-dimensional images

Abstract

The present invention describes a method for preventing work accidents in real time based on three-dimensional images, comprising: receiving a set of requirements that must be met by people in a monitored work area; capturing three-dimensional images of the monitored work area; creating three-dimensional regions of interest in the monitored work area; and processing the three-dimensional images of the three-dimensional regions of interest created with artificial intelligence techniques to identify whether the set of requirements is being met, said processing of the three-dimensional images with artificial intelligence techniques comprising analyzing pixel data and information in three-dimensional coordinates and in color coordinates.

Description

"METHOD AND SYSTEM FOR REAL-TIME WORK ACCIDENT PREVENTION BASED ON THREE-DIMENSIONAL IMAGES" TECHNICAL FIELD

[0001] The present invention relates to a method and system for preventing workplace accidents in real time based on three-dimensional images. More specifically, the present invention relates to a method and system that process three-dimensional images using artificial intelligence techniques, of the deep learning type, to prevent workplace accidents in real time. BACKGROUND OF THE INVENTION

[0002] Compliance with safety standards and proper work execution are crucial to preventing workplace accidents. Therefore, some existing solutions aim to assist people in the workplace in complying with these standards and performing their work properly to avoid accidents.

[0003] Document KR 102136070 Bl, entitled "DISASTER AND ACCIDENT PREVENTION DEVICE IN CONSTRUCTION SITE USING AI IMAGE ANALYSIS", published on July 21, 2020, refers to a device for disaster and accident prevention on a construction site through image analysis by artificial intelligence, capable of, in a dangerous zone of a construction site, issuing an alert by analyzing an image of a worker approaching the danger zone. The device includes a housing with internal storage space; a camera installed outside the housing and photographing an image of a predetermined angle range; a control part available in the storage space and analysis of the photographed image. Real-time camera; and a loudspeaker selectively emitting a warning announcement to the control party.

[0004] The document WO 2019 / 058379 Al, entitled "SYSTEMS AND METHODS FOR PREVENTING WORK ACCIDENTS", published on March 28, 2019, describes systems and methods used for preventing work accidents. The system determines at least one characteristic of a task scheduled to occur in an industrial environment. The system uses first synergy data from at least three types of safety-related information and at least one task characteristic to determine that a predicted risk score for the scheduled task is below a first threshold. After that, the system obtains real-time information indicative of human error from at least one employee associated with the task. The system uses second synergy data from at least three types of safety-related information and real-time information to determine if an actual risk score for the task has changed relative to the predicted risk score.When the task's actual risk score exceeds a second threshold, the system initiates corrective action to prevent an accident.

[0005] US patent 10,643,080 B2, entitled "ARTI-FICIAL INTELLIGENCE AND IMAGE PROCESSING-BASED WORK FORCE SAFETY," published on May 5, 2020, describes a workplace safety method for an industrial processing facility comprising a work safety system based on artificial intelligence and image processing that receives image data from cameras viewing work zones, including a A camera shows an individual in a work area. From the image data, the individual's current location is determined. A current minimum requirement for Personal Protective Equipment (PPE) is determined based on the current location, referencing a database of work environments with hazardous conditions and determining the current work environment with a current hazardous condition and the PPE required for the current hazardous condition. The image data is analyzed to identify the PPE currently worn by the individual. When it is determined that the individual is not currently meeting the PPE requirements, by comparing the PPE currently worn with the minimum PPE requirements, an alert will be generated in response to the unsafe condition.

[0006] US patent 2012 / 0146789 Al, titled "AUTOMATED MONITORING AND CONTROL OF SAFETY IN A PRODUCTION AREA," published on June 14, 2012, describes a machine vision process that monitors and controls safe work practices in a production area by capturing and processing image data relating to personal protective equipment (PPE) worn by individuals, the movement of various articles, and the conformations related to the movement of individuals and other objects in the production area. The data is analyzed to determine if there is a violation of a predetermined minimum image, movement, or conformation value for a predetermined time period. The determination of a safety violation triggers the computer activation of a safety control device.

[0007] However, one problem with the state of the art is the fact that it only uses two-dimensional cameras to capture images. The images to be processed will be processed in only two spatial coordinates (width and length) and three color coordinates (red, green, and blue).

[0008] An additional problem with the state of the art is that it does not take into account the actions performed by people in a workplace. SUMMARY OF THE INVENTION

[0009] One objective of the present invention is to provide a method for preventing workplace accidents in real time based on three-dimensional images that avoids the problems of the prior art.

[0010] This objective is achieved through a method for preventing workplace accidents in real time based on three-dimensional images, comprising: receiving a set of requirements that must be met by people in a monitored work area; capturing three-dimensional images of the monitored work area; creating three-dimensional regions of interest in the monitored work area; and processing the three-dimensional images of the created three-dimensional regions of interest with artificial intelligence techniques to identify whether the set of requirements is being met, said processing of three-dimensional images with artificial intelligence techniques comprising analyzing pixel data and information in three-dimensional coordinates and in color coordinates.

[0011] Additionally, the method according to the present invention consists in the fact that the set of requirements comprises one or more of: identifying the number of people in the monitored work area, use of personal protective equipment, execution of dangerous gestures, and distance between objects and people.

[0012] Furthermore, the method according to the present invention also consists of identifying the depth of objects in the processed three-dimensional images, seeking to identify their distances in relation to people and the risks related to approaching objects.

[0013] One advantage of the method according to the present invention is that, when alerting about a specific deviation, the system is able to provide information about the reason for the alert, thus facilitating its correction given the universe of existing deviations.

[0014] Furthermore, the method according to the present invention consists of issuing a real-time alert in case of non-compliance with any requirement, wherein the alert issued comprises the following variations of signal emission: audible alert, visual alert, and vibration alert.

[0015] The method according to the present invention also consists of creating and sending files to a web platform containing a summary of the procedures performed by people in the monitored work area. The files are analyzed, and a score is generated based on the procedures performed, with the score being generated by quantifying the number of non-compliances with the set of requirements.

[0016] The present invention also defines a system for preventing workplace accidents in real time based on three-dimensional images, comprising: stereoscopic cameras, adapted to capture three-dimensional images of a monitored work area; warning devices, adapted to issue a real-time alert in case of non-compliance with any requirement; and a processor with instructions to execute the method defined in the present invention.

[0017] An additional advantage of the present invention consists in analyzing data and information from three-dimensional images, that is, it analyzes three spatial coordinates (width, length, and depth) and three color coordinates (red, green, and blue). In this way, the present invention enables the training of neural networks capable of analyzing the depth of each object in the detected images and, in turn, their distances in relation to people and the risks related to approaching objects.

[0018] Another advantage of the present invention is the fact that the image processing is also based on the actions performed by people in the workplace to generate an alert that will draw the attention of anyone present. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The objectives and advantages of the present invention will become clearer through the following detailed description of the non-limiting examples and drawings presented at the end of this document:

[0020] Figure 1 illustrates the flowchart of the method for preventing workplace accidents in real time based on... three-dimensional images according to a preferred embodiment of the present invention.

[0021] Figure 2 schematically illustrates the system for preventing workplace accidents in real time based on three-dimensional images according to a preferred embodiment of the invention.

[0022] Figure 3A illustrates a possible implementation of the method and system according to a preferred embodiment of the present invention.

[0023] Figure 3B illustrates a possible implementation of the method and system according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0024] Although the present invention may be susceptible to different embodiments, a preferred embodiment is shown in the following detailed discussion, it being understood that the present description should be considered an exemplification of the principles of the invention and that the present invention is not intended to be limited to what has been illustrated and described herein.

[0025] According to Figure 1, the method for preventing workplace accidents in real time based on three-dimensional images comprises an initial step of receiving a set of requirements that must be met by people in a monitored work area.

[0026] The set of requirements includes the work rules that will be verified during the execution of a job. These rules are customizable and can be changed depending on the job. For example, in a job at heights, the use of a piece of equipment... Personal protective equipment, such as a helmet, is necessary; it's possible to add a rule to identify if a person is wearing a helmet. In another example, if a certain task requires only two people, it's possible to create a rule to check if the number of people in the work area respects the required limit.

[0027] Therefore, the set of requirements includes various possible rules, such as the number of people in the monitored work area, the use of personal protective equipment, the execution of dangerous gestures, the distance between objects and people, among others.

[0028] According to Figure 1, the method also includes a step to capture 200 three-dimensional images of the monitored workspace.

[0029] The capture of three-dimensional images is performed using stereoscopic cameras positioned in a work environment on cones or tripods. Optionally, the cameras can also be mounted on walls. The cameras must be positioned to generate the fewest possible blind spots so that there is a correct analysis of the images. Furthermore, since stereoscopic cameras can capture three-dimensional images, the generated images provide information on the spatial position of each captured pixel.

[0030] Thus, after the cameras capture three-dimensional images of the monitored workspace, 300 three-dimensional regions of interest are created where work is performed, which will be analyzed by the method. Therefore, the method ignores the remaining regions of the monitored workspace where no work is being performed.

[0031] As illustrated in Figure 1, the method then processes 400 three-dimensional images of the three-dimensional regions of interest created with artificial intelligence techniques to identify whether the set of requirements is being met.

[0032] Deep learning-type artificial intelligence algorithms are used in the method to analyze pixel data and information in three-dimensional coordinates (width, length, and depth) and in RGB color coordinates (red, green, and blue). Thus, each pixel of the captured images comprises six coordinates, three spatial coordinates and three color coordinates.

[0033] Artificial intelligence algorithms comprise a first module responsible for detecting deviations and risks in still images. The developed model, while capable of working in any scenario, has specialized application in the electrical power sector, being able to detect specific elements such as transformers, knife switches, and others. One implementation can analyze the use of personal protective equipment by each individual, also identifying incorrect positioning of said equipment. The algorithm management infrastructure was developed to allow for modification of the algorithm at runtime and the combination of various algorithms to cover different application areas. A second module consists of specialized models capable of verifying actions in real time. Developed for the creation of a real-time accident prevention system, it utilizes... Advanced artificial intelligence algorithms detect and recognize action patterns in videos. Using convolutional neural networks, deep learning models, and LSTM architecture, this module is capable of analyzing sequences of images in real-time streams and identifying a wide range of human actions, including movements, gestures, and interactions. One of the main features of this module is its ability to consider temporal information to identify the correct work procedure in hazardous areas. Through the analysis of sequential video frames, it can recognize movement patterns and identify specific actions, enriching the decision criteria in detecting situations of risk or imminent danger.

[0034] According to Figure 1, the method additionally involves identifying the depth of objects in the processed images, thus identifying the distance between objects and people.

[0035] Thus, after processing, calculations are made to determine the relative position of each detected object in relation to the others. The process involves camera calibration, feature matching, triangulation, filtering, and refinement. This allows for object recognition to provide three-dimensional position information of the detected objects based on recognition by the artificial intelligence used.

[0036] According to Figure 1, the method also includes issuing a real-time alert in case of non-compliance with any requirement of the set of requirements.

[0037] Warning devices are used to issue alerts when a requirement is not met. Thus, audible, visual, and vibration alerts are used, which are activated when a requirement is not met to draw the person's attention.

[0038] Finally, as illustrated in Figure 1, the method comprises a step of creating and sending 700 files to a web platform summarizing the procedures performed by people in the monitored work area. Subsequently, the files are analyzed and a score is generated based on the procedures performed, with the score being generated by quantifying the number of non-compliances with the set of requirements. Thus, the lower the risk and the more the rules and use of personal protective equipment are respected, the higher the score assigned. The results of the evaluations are sent to a cloud and can be viewed on a website.

[0039] As illustrated in Figure 2, a system for preventing workplace accidents in real time based on three-dimensional images according to the present invention comprises stereoscopic cameras 1000, adapted to capture three-dimensional images of a monitored work area, alert devices 1100, adapted to issue a real-time alert in case of non-compliance with any requirement, and a processor 1200 with instructions to execute the method for preventing workplace accidents in real time based on three-dimensional images.

[0040] Figure 3A illustrates a possible implementation of the method and system as defined in the present invention. As can be observed, the system receives a set of Requirements that must be met by people working in a monitored workspace. Stereoscopic cameras capture three-dimensional images of the monitored workspace and create a region of interest. According to Figure 3B, when a person enters the monitored workspace and fails to meet a requirement, the alert devices are activated and issue a real-time alert.

[0041] In addition to the embodiment presented above, the same inventive concept may be applied to other alternatives or possibilities for using the invention.

[0042] Although the present invention may be susceptible to different embodiments, a preferred embodiment is shown in the following detailed discussion, it being understood that the present description should be considered an exemplification of the principles of the invention and that the present invention is not intended to be limited to what has been illustrated and described herein.

Claims

CLAIMS 1. Method for preventing work accidents in real time based on three-dimensional images, characterized in that it comprises: receiving (100) a set of requirements that must be met by people in a monitored work area; capturing (200) three-dimensional images of the monitored work area; creating (300) three-dimensional regions of interest in the monitored work area; and processing (400) the three-dimensional images of the created three-dimensional regions of interest with artificial intelligence techniques to identify whether the set of requirements is being met, said processing of the three-dimensional images with artificial intelligence techniques comprising analyzing pixel data and information in three-dimensional coordinates and in color coordinates.

2. A method according to claim 1, characterized in that the set of requirements comprises identifying one or more of: the number of people in the monitored work area, the use of personal protective equipment, the execution of dangerous gestures, and the distance between objects and people.

3. Method, according to claim 1, characterized in that it further comprises identifying (500) the depth of objects in the processed three-dimensional images.

4. Method according to claim 1, characterized in that it further comprises issuing (600) a real-time alert in the event of non-compliance with any requirement.

5. A method according to the preceding claim, characterized in that the emitted alert comprises one or more audible alerts, visual alerts, and vibration alerts.

6. Method, according to claim 1, characterized in that it comprises preparing and sending (700) to a web platform files with a summary of the procedures performed by people in the monitored work area.

7. A method, according to any of the preceding claims, characterized in that the files are analyzed and a score is generated based on the procedures that were performed, with the score being generated by quantifying the amount of non-compliance with the set of requirements.

8. Real-time workplace accident prevention system based on three-dimensional images, characterized in that it comprises: stereoscopic cameras (1000), adapted to capture three-dimensional images of a monitored work area; alert devices (1100), adapted to issue a real-time alert in case of non-compliance with any requirement; and a processor (1200) with instructions to execute the method as defined in any of the claims 1 to 6.

Images