Safety monitoring device, safety monitoring method, and safety monitoring program

Abstract

To provide a safety monitoring system, a safety monitoring method and a safety monitoring program for detecting both the working state of a worker and the wearing state of a protector, and for monitoring the safety of the worker based on the detection result.SOLUTION: A safety monitoring device includes a prevention detection unit that detects a prevention state in a work environment as to whether or not a safety measure is taken based on information from at least one first sensor, a work detection unit that detects a work state in the work environment as to whether or not predetermined work is being executed based on information from at least one second sensor, and a determination unit that determines whether or not work is possible in the work environment based on the prevention state and the work state.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a safety monitoring device, a safety monitoring method, and a safety monitoring program for monitoring the safety of workers in a factory. [Background technology]

[0002] Workers performing specific tasks in factories are required to wear appropriate safety equipment to ensure safety. Patent Document 1 proposes a work status monitoring system that can determine whether the status of safety equipment is appropriate depending on the worker's location. Patent Document 2 proposes clothing and a warning system that can maintain a good working environment at the workplace. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2017-093515 [Patent Document 2] Japanese Patent Application Publication No. 2019-020921 Summary of the Invention [Problem to be solved by the invention]

[0004] However, while the above-mentioned technology can separately detect the state of the protective equipment worn by the worker and the work status of the worker, it is difficult to perform more comprehensive and consistent safety monitoring.

[0005] Non-limiting examples of the present disclosure contribute to providing a safety monitoring system, a safety monitoring method, and a safety monitoring program that monitor worker safety more comprehensively and consistently. [Means for solving the problem]

[0006] A safety monitoring system according to one embodiment of the present disclosure includes a prevention detection unit that detects a prevention state indicating whether or not safety measures are being taken based on information from a first sensor, a work detection unit that detects a work state indicating whether or not a specified task is being performed based on information from a second sensor, and a determination unit that determines whether or not work is possible based on the prevention state and the work state. [Effects of the Invention]

[0007] The present disclosure provides for more comprehensive and consistent worker safety monitoring. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram illustrating an example of the configuration of a safety monitoring system according to the first embodiment. [Figure 2] FIG. 2 is a diagram illustrating an example of the configuration of the preventive detection unit. [Figure 3] FIG. 3 is a diagram illustrating an example of the configuration of the operation detection unit. [Figure 4] FIG. 4 is a diagram illustrating the prevention state and the working state input to the determining unit. [Figure 5] FIG. 5 is a diagram showing an example of a determination result obtained when both the preventive detection unit and the operation detection unit use the non-independent detection method. [Figure 6] FIG. 6 shows an example of a determination result obtained when both the preventive detection unit and the work detection unit use the independent detection method. [Figure 7] FIG. 7 is a flowchart showing the processing of the safety monitoring system according to the first embodiment. [Figure 8] FIG. 8 is a diagram showing an example of the configuration of a safety monitoring system according to the first modification of the first embodiment. [Figure 9] FIG. 9 is a diagram illustrating an example of the configuration of a safety monitoring system according to the second embodiment. [Figure 10]FIG. 10 is a diagram illustrating an example of a determination result by the determination unit using priority information that gives priority to the detection result by the preventive detection unit over the detection result by the operation detection unit. [Figure 11] FIG. 11 is a diagram illustrating an example of a determination result by the determination unit using priority information that gives priority to the detection result by the work detection unit over the detection result by the preventive detection unit. [Figure 12] FIG. 12 is a diagram illustrating an example of a determination result by the determination unit using priority information that prioritizes a detection result of "prevention: not present" over a detection result of "prevention: present" by the prevention detection unit. [Figure 13] FIG. 13 is a diagram illustrating an example of a determination result of the determination unit using priority information that prioritizes "prevention: yes" over "prevention: no" of the prevention detection unit. [Figure 14] FIG. 14 is a diagram illustrating an example of a determination result by the determination unit using priority information indicating that "Work: Safety" is given priority over "Work: Caution" by the work detection unit. [Figure 15] FIG. 15 is a diagram illustrating an example of a determination result by the determination unit using priority information indicating that "Work: Caution" is given priority over "Work: Safety" by the work detection unit. [Figure 16] FIG. 16 is a diagram illustrating an example of a determination result of the determination unit using threshold information in which the threshold for the preventive detection unit is lowered and the threshold for the operation detection is increased. [Figure 17] FIG. 17 is a diagram illustrating an example in which the determination unit makes a determination using threshold information that increases the threshold of the preventive detection unit and decreases the threshold of the operation detection unit. [Figure 18] FIG. 18 is a diagram illustrating an example in which the determination unit makes a determination using threshold information in which the threshold for "prevention: not" is set higher than the threshold for "prevention: with" of the prevention detection unit. [Figure 19] FIG. 19 is a diagram illustrating an example in which the determination unit makes a determination using threshold information that sets the threshold for "prevention: not" higher than the threshold for "prevention: with" of the prevention detection unit. [Figure 20]FIG. 20 is a diagram illustrating an example in which the determination unit makes a determination using threshold information in which the threshold for "task: caution" of the task detection unit is set higher than the threshold for "task: safety." [Figure 21] FIG. 21 is a diagram illustrating an example in which the determination unit makes a determination using threshold information in which the threshold for "task: caution" of the task detection unit is set higher than the threshold for "task: safety." [Figure 22] FIG. 22 is a diagram for explaining an example in which the determining unit interpolates the results of the momentary indeterminate state of the preventive state or the working state using time-series information. [Figure 23] FIG. 23 is a diagram for explaining an example in which an unsafe state is given priority. [Figure 24] FIG. 24 is a diagram for explaining an example in which a safe state is given priority. [Figure 25] FIG. 25 is a diagram illustrating an example in which the result of the prevention detection unit is corrected by the result of the work detection unit. [Figure 26] FIG. 26 is a diagram illustrating an example in which the result of the operation detection unit is corrected by the result of the prevention detection unit. [Figure 27] FIG. 27 is a diagram illustrating an example of the detection result of the prevention detection unit, the detection result of the operation detection unit, and the determination result of the determination unit in the safety monitoring system according to the third embodiment. [Figure 28] FIG. 28 is a diagram illustrating an example of hardware for realizing the safety monitoring system. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings as appropriate. However, more detailed explanation than necessary may be omitted. For example, detailed explanation of already well-known matters or redundant explanation of substantially identical configurations may be omitted. This is to avoid unnecessary redundancy in the following explanation and to facilitate understanding by those skilled in the art. The accompanying drawings and the following description are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

[0010] (Embodiment 1) Hereinafter, with reference to FIGS. 1 to 8, a configuration example and an operation example of the safety monitoring system 1 according to the first embodiment of the present disclosure will be described.

[0011] As shown in FIG. 1, the safety monitoring system 1 includes a prevention detection unit 2, an operation detection unit 3, and a determination unit 4.

[0012] The prevention detection unit 2 detects the preventive state in the work environment, that is, whether or not at least one safety measure is being taken. Here, the preventive state refers to whether or not preventive measures are being taken appropriately to prevent a decrease in safety in the work environment.

[0013] In the first embodiment, the preventive state refers to, for example, a state regarding whether or not a worker is wearing protective equipment appropriate for the work. Examples of protective equipment include protective glasses for protecting the worker's eyes, a helmet for protecting the head from falling objects, a face shield for protecting the entire face from flying debris and chemicals into the eyes, earplugs for hearing protection in noisy environments, protective masks for preventing inhalation of dust, fine particles, and harmful gases, safety shoes for protecting the feet from falling heavy objects, protective gloves for protecting the hands from chemicals, heat, electricity, etc., a safety belt for preventing falls when working at heights, protective clothing for protecting the entire body from dirt and chemicals, etc.

[0014] Furthermore, the preventive state may be other states regarding whether or not safety measures are taken in the work environment, in addition to whether or not the worker wears protective equipment, such as whether or not the work area is adequately ventilated, whether the temperature is properly maintained, whether or not there is sufficient lighting, whether or not objects are properly stored, and whether or not an escape route is available.

[0015] The preventive state may be determined by one safety measure or by multiple safety measures. For example, the preventive state may be determined by one safety measure, namely, wearing safety glasses, or by two safety measures, namely, wearing safety glasses and safety shoes, or by three safety measures, namely, wearing safety glasses and safety shoes and an appropriate temperature.

[0016] The safety measures to be detected are set arbitrarily based on the safety of the related work, the existence of applicable laws and regulations, the type of work and number of processes, the experience and skills of the workers, the work environment, the type of safety equipment used and its performance, data on past accidents and problems, work hours and workload, etc.

[0017] Regarding the safety measures of the workers to be detected, the safety measures of one worker may be detected, or in the case of a work performed by multiple workers in cooperation, the safety measures of each of them may be detected. For example, in a production line in a factory, the safety measures of each of worker A who installs part A, worker B who installs part B, and worker C who installs part C may be detected using, for example, multiple sensors. By detecting the safety measures of multiple workers, if it is detected that a safety measure has not been taken for one of them, appropriate measures such as stopping the production line or issuing a warning can be taken, thereby improving the safety of the working environment.

[0018] The preventive detection unit 2 receives as input the output from at least one first sensor 7 and performs preventive detection based on this. The connection between the preventive detection unit 2 and the first sensor 7 is wireless or wired. For wired connections, for example, a USB cable or an Ethernet cable can be used. Wired connections provide highly stable signals and are less susceptible to noise, allowing for highly reliable data transmission. Furthermore, wired connections are suitable for long-distance transmission, allowing for reliable collection of sensor data even over a wide area of ​​a factory.

[0019] On the other hand, Wi-Fi, Bluetooth (registered trademark), etc. can be used for wireless connections. Wireless connections have the advantage of eliminating the need for wiring when multiple sensors are used and providing greater flexibility in sensor installation locations. For example, by using a Wi-Fi connection, sensors can be placed over a wide area within a factory and data can be transmitted to the preventive detection unit 2. Furthermore, a Bluetooth (registered trademark) connection is suitable for short-distance data transmission and is effective for collecting sensor data within individual work areas.

[0020] The first sensor 7 is a variety of sensors such as a passive sensor or an active sensor. Examples of the first sensor 7 that provides information for determining whether or not a worker is wearing safety equipment include a camera sensor such as a high-resolution camera or an infrared camera that uses image recognition technology to detect whether or not a worker is wearing a helmet or protective glasses, an RFID tag and reader that attaches an RFID tag to the worker's safety equipment and reads it with an RFID reader to confirm whether or not the equipment is being worn, a pressure sensor that detects whether or not the worker is working while seated in a work chair, and a proximity sensor such as an ultrasonic sensor or an infrared sensor that detects whether or not the safety equipment is being worn by reacting when the worker's body or equipment approaches the sensor.

[0021] Examples of sensors for detecting whether safety measures are being taken in the work environment include temperature sensors such as thermo cameras and thermistors that monitor whether the temperature of the work environment is being maintained appropriately, light sensors that detect whether the work area is sufficiently lit, gas sensors such as semiconductor gas sensors and electrochemical gas sensors that detect the presence of harmful gases or unsafe substances, sound sensors such as microphones and sound level meters that measure noise levels in the work environment, vibration sensors such as accelerometers and vibrometers that monitor vibrations in work equipment and the environment, and various other sensors such as LiDAR sensors that measure the distance to objects or people in the work environment and the shapes of objects and people.

[0022] Furthermore, the first sensor 7 that can be connected to the preventive detection unit 2 may be an area sensor. An area sensor is a sensor system for detecting the presence and movement of objects and people within a specified area. Because area sensors can cover a wide area at once, a single sensor can monitor multiple workers and a large work area, eliminating the need to install multiple sensors and reducing costs and installation labor.

[0023] The first sensor 7 may be single or multiple. When there is a single sensor, the system configuration is simplified and installation and maintenance are easy. With a single sensor, the sensor installation position and wiring are simple, and initial costs can be reduced. In addition, data analysis is relatively easy, and the overall system operates quickly.

[0024] On the other hand, if multiple sensors are used, it is possible to detect the safety measures of multiple workers, improving the accuracy and reliability of detection. Furthermore, by combining multiple sensors, if one sensor fails, it is possible for other sensors to compensate. Furthermore, by using different types of sensors, it is possible to collect multifaceted data on the preventive status, providing a basis for more accurate judgment of the preventive status.

[0025] For example, when detecting whether protective eyewear is being worn, using a camera sensor and an infrared sensor together can detect both the light reflection characteristics and shape, improving detection accuracy. Furthermore, by installing multiple sensors, data from each sensor can be cross-verified, reducing the risk of false detection or malfunction. By integrating and cross-checking data obtained from multiple sensors, a more reliable determination of the protective status can be made.

[0026] The preventive detection unit 2 receives as input the data detected by the first sensor 7 and performs preventive detection. In preventive detection by the preventive detection unit 2, for example, if the first sensor 7 is a camera sensor, the camera sensor captures an image of the work area and acquires image data of the worker.

[0027] The image data includes the worker's face and the state of protective equipment. The preventive detection unit 2 applies a face detection algorithm to the image data to identify the position of the worker's face. After detecting the face, the preventive detection unit 2 also applies an image recognition algorithm to check the wearing status of protective equipment such as protective glasses, and determines whether the protective equipment is being worn properly.

[0028] The preventive detection unit 2 may omit the process of identifying the position of the worker's face and detect the wearing state of the protective equipment from the image data. Furthermore, if the sensor is a temperature sensor, for example, the preventive detection unit 2 acquires data on the temperature around the worker measured by the temperature sensor, analyzes the data according to the preventive detection algorithm, and determines whether safety measures are being taken for the temperature around the worker.

[0029] As shown in FIG. 2, the preventive detection unit 2 includes a detection unit 21 and a processing unit 22.

[0030] There are two types of detection methods for the prevention state by the detection unit 21: non-independent detection and independent detection. In the case of non-independent detection, the probabilities of the two prevention states, "prevention: yes" and "prevention: no", are detected non-independently, and the probabilities of each state are output to the processing unit 22. "Non-independent" means that there is some relationship between the probability of "prevention: yes" and the probability of "prevention: no".

[0031] In the case of independent detection, the two prevention states, "prevention: yes" and "prevention: no", are detected independently, and the probability of each state is output to the processing unit 22. In independent detection, the probability of "prevention: yes" and the probability of "prevention: no" are detected independently, so in principle, their total can be up to 200%.

[0032] The processing unit 22 outputs one prevention state based on the calculated probability. In the case of non-independent detection, one of two prevention states, "prevention: yes" and "prevention: no", is output depending on the probability. In the case of independent detection, one of three prevention states, "prevention: yes", "prevention: no", and "prevention: uncertain", is output depending on the probability.

[0033] Here, the preventive state "Prevention: Yes" refers to a state in which safety measures have been taken. For example, this refers to a state in which a worker is wearing protective eyewear as protective equipment, or a state in which the lighting in the work area is sufficient. The preventive state "Prevention: No" refers to a state in which safety measures have not been taken. For example, this refers to a state in which a worker is not wearing protective eyewear as protective equipment, or a state in which the lighting in the work area is not bright enough for work. The preventive state "Prevention: Uncertain" refers to a state in which it is not clear whether safety measures have been taken. For example, this refers to a state in which it is not clear whether a worker is wearing protective eyewear as protective equipment, or a state in which it is not clear whether the lighting in the work area is not bright enough for work. The preventive state "Prevention: Uncertain" may be output when the difference in probability between the two preventive states "Prevention: Yes" and "Prevention: No" is within a predetermined value.

[0034] The detection unit 21 can select either independent or non-independent detection depending on the situation. For example, this can be the case when the detection accuracy is affected by changes in the lighting conditions or work environment in the factory. If the lighting in the factory is sufficient, independent detection is applied, and the wearing status of protective equipment can be detected with high accuracy.

[0035] On the other hand, when lighting is insufficient or the worker is working in a dark place, more accurate judgment is possible by using independent detection to detect multiple states including "Prevention: Uncertain." Furthermore, the detection unit 21 can select non-independent detection or independent detection depending on the intensity of the worker's movements and the type of work.

[0036] For example, when a worker is performing dynamic work that requires a lot of movement, independent detection can be used to flexibly respond to momentary movements and environmental changes. Conversely, when performing static work such as sitting down, non-independent detection can be used to obtain more stable detection results.

[0037] In independent detection, when detecting whether a factory worker is wearing protective glasses, the detection unit 21 may output a detection result that the probability of "prevention: yes" is 75% and the probability of "prevention: no" is 75%. In this case, the sum of the probabilities of the two states exceeds 100%, indicating that multiple states may exist for the same data.

[0038] For example, if an image detected by a camera sensor only captures a portion of the protective glasses worn by a worker, it may be possible that the worker is wearing the protective glasses correctly as required, but because only a portion of the image is captured, the worker is wearing them in an insufficient manner.

[0039] Independent detection allows for simultaneous detection of multiple conditions, enabling a more detailed understanding of the situation, which can reflect the complexities of reality. Note that when the probability of "prevention: yes" and the probability of "prevention: no" are roughly the same, the priority information in the second embodiment described below can be used.

[0040] In non-independent detection, when detecting whether a factory worker is wearing protective glasses, the detection unit 21 may output detection results with a 75% probability of "prevention: yes," a 15% probability of "prevention: no," and a 10% probability of "prevention: uncertain." In this case, the probabilities of each of the three states are not independent, and the total is 100%.

[0041] This indicates that the states are mutually exclusive, reflecting, for example, that a worker may be wearing safety glasses properly, not wearing them at all, or wearing them poorly.

[0042] In non-independent detection, each state is not independent, so a clear and exclusive judgment is possible, simplifying analysis. Note that if the probabilities of "prevention: yes," "prevention: no," and "prevention: uncertain" are similar, the priority information in the second embodiment described below can be used. On the other hand, in independent detection, each state is independent, so the output is non-exclusive, which can complicate analysis.

[0043] If the first sensor 7 is a camera sensor, the probability provided by the detection unit 21 is found by analyzing the video data acquired from the camera sensor. For example, the probability is calculated using the following procedure: Video data is collected from the camera sensor. Then, a model for calculating the probability of a preventive state is applied to the video data. For example, a deep learning algorithm is used to generate this model. The probability of a preventive state (wearing protective glasses) is calculated using this model.

[0044] For example, in non-independent detection, the probability that the protective glasses are worn is calculated as 85%, and the probability that the protective glasses are not worn is calculated as 15%. In independent detection, the probability that the protective glasses are worn is calculated as 85%, and the probability that the protective glasses are not worn is calculated as 45%. For example, if the protective glasses have multiple structural and color characteristics, the probabilities obtained from the multiple characteristics may be integrated to calculate the final probability of the preventive state.

[0045] The probability calculated by the detection unit 21 may be calculated based on information from a single sensor, may be calculated based on information from multiple sensors, or may be calculated based on information from one of multiple sensors.

[0046] When calculating probability from a single sensor, relying on a single sensor means that detection accuracy is heavily dependent on the performance of the sensor, and there is a risk of false positives or oversights. On the other hand, when using multiple sensors, combining information from different types of sensors improves the accuracy of the probability. For example, by combining a camera sensor and a temperature sensor, judgments can be made based on temperature change information in addition to visual information, making it possible to calculate probabilities with greater accuracy.

[0047] The trigger and timing of sampling by the detection unit 21 to acquire data from the first sensor 7 may be selected arbitrarily depending on the detection target. The trigger to start and stop sampling may be, for example, a time-based trigger that starts sampling at 9:00 AM when factory work starts and stops at 5:00 PM when work ends, an action-based trigger that starts and stops sampling in response to a worker turning on or off a light, or a sensor detection trigger that starts and stops sampling in response to the detection of a worker by the first and second sensors 7 and 8 or another sensor.

[0048] The sampling interval may be set arbitrarily depending on the type of work. For example, in the case of dynamic work involving vigorous movement, it is preferable to set the sampling interval to a relatively short time. On the other hand, in the case of static work while sitting in a chair, the sampling interval may be set to a relatively long time.

[0049] The processing unit 22 outputs the detected prevention status to the determination unit 4. In the case of non-independent detection, one of two prevention statuses, "prevention: yes" and "prevention: no", is output. In the case of independent detection, one of three prevention statuses, "prevention: yes", "prevention: no", and "prevention: undetermined", is output.

[0050] The task detection unit 3 detects the task status of a worker in the task environment, i.e., whether a specific task is being performed. The task status refers to whether or not a worker is performing a specific task in the task environment. Examples of tasks include dynamic tasks, which involve the worker's vigorous and extensive physical movements, and static tasks, which involve the worker's relatively small movements and maintaining the same posture for long periods of time. Examples of dynamic tasks include product assembly, welding, and packaging. Examples of static tasks include soldering electronic components, assembling precision equipment, inspecting product quality, and monitoring automated production lines.

[0051] A predetermined task may be determined by one work process or by multiple work processes. For example, in the case of an assembly task in which one product is assembled from multiple parts A, B, and C, if the work process includes a first process for incorporating part A, a second process for incorporating part B, and a third process for incorporating part C, the first process may be detected as the predetermined task, or the first to third processes may be detected as the predetermined task.

[0052] The work detection unit 3 receives as input the output from at least one second sensor 8 and detects the work status based on this. The work detection unit 3 and the second sensor 8 are connected wirelessly or via a wire, similar to the connection between the first sensor 7 and the preventive detection unit 2, so a detailed description will be omitted.

[0053] The at least one second sensor 8 is a passive sensor, an active sensor, etc. Any sensor that can be connected to the work detection unit 3 can be selected as long as it can directly or indirectly detect the dynamic work and static work of the worker.

[0054] Examples of sensors include infrared sensors that use infrared rays to detect the presence and movement of objects to determine whether a worker is working or not; ultrasonic sensors that emit ultrasonic waves and receive reflected waves to detect the distance and presence of objects to determine whether a worker is working or not; pressure sensors that detect the pressure applied by an object to determine whether a worker is sitting in a chair or not; proximity sensors that detect the approach of an object to determine whether a worker is at a specified distance and whether a worker has approached a machine; optical sensors that detect the intensity and reflection of light; magnetic sensors that detect changes in magnetic fields; temperature sensors that detect the temperature of an object; and LiDAR sensors that measure the distance to objects or people in the work environment and the shape of objects and people.

[0055] The second sensor 8 is not limited to a sensor that directly detects the worker's work motion, but may be a switch that indirectly detects the work motion. For example, a switch may be placed on a tool stand or tool box, and turned on when a tool is lifted.

[0056] The second sensor 8 that can be connected to the work detection unit 3 may be an area sensor. An area sensor is a sensor system for detecting the presence and movement of objects and people within a specified area. Area sensors can cover a wide area at once, making it possible to monitor multiple workers and a large work area with a single sensor, thereby eliminating the need to install multiple sensors and reducing costs and installation labor.

[0057] The second sensor 8 may be a single sensor or multiple sensors. For example, in the case of dynamic work in which the worker moves vigorously, multiple sensors may be provided to detect the predetermined work from multiple angles. On the other hand, in the case of static work in which the worker moves relatively little, a single sensor may be used.

[0058] The at least one first sensor 7 associated with the preventive detection unit 2 and the at least one second sensor 8 associated with the work detection unit 3 may be separate sensors provided independently of each other, or may be the same shared sensor. By combining the first sensor 7 and the second sensor 8 into a single shared sensor, cost reduction, simplified installation, data consistency and integrity, and efficient data processing are possible.

[0059] The task detection unit 3 receives as input the data detected by the second sensor 8 and performs task detection. For example, if the second sensor 8 is a camera sensor, the task detection unit 3 detects the task by having the camera sensor capture an image of the task area and acquire image data including the worker's actions. This includes the worker's movements and the usage status of the tools the worker holds. A task detection algorithm is applied to the image data to analyze the worker's movements. This determines whether the task being performed by the worker corresponds to a specified task (e.g., soldering, assembling parts, etc.).

[0060] As shown in FIG. 3, the operation detection unit 3 includes a detection unit 31 and a processing unit 32.

[0061] Similar to the detection unit 21 of the preventive detection unit 2, there are two methods for detecting the pre-work state by the detection unit 31: non-independent detection and independent detection. In the case of non-independent detection, the probabilities of the two states, "Work: Caution" and "Work: Safe", are detected non-independently, and the probabilities of each state are output to the processing unit 32. In the case of independent detection, the probabilities of the two preventive states, "Work: Caution" and "Work: Safe", are detected independently, and the probabilities of each state are output to the processing unit 32.

[0062] The processing unit 32 outputs one work status based on the calculated probability. In the case of non-independent detection, one of two preventive statuses, "Work: Caution" and "Work: Safe", is output depending on the probability. In the case of independent detection, one of three work statuses, "Work: Caution", "Work: Safe", and "Work: Undefined", is output depending on the probability.

[0063] Here, the working state of "Work: Caution" refers to a state in which the worker is performing a specified work. For example, if the specified work is soldering, this means that the worker is performing soldering work. The working state of "Work: Safe" refers to a state in which the worker is not performing the specified work. For example, if the specified work is soldering, this means that the worker is not performing soldering work.

[0064] The work state "Work: Uncertain" refers to a state in which it is not clear whether the worker is performing a specified work. For example, it means a state in which it is difficult to detect whether the worker is performing soldering work. The work state "Work: Uncertain" is output when the difference in probability between the two work states "Work: Yes" and "Work: No" is within a specified value.

[0065] The detection unit 31 can select between independent and non-independent detection depending on the situation. Examples of such situations include dynamic work, static work, and when detection accuracy is affected by changes in the lighting conditions or work environment in the factory. In the case of static work where the movement of workers is relatively small, non-independent detection is applied, allowing for stable and highly accurate detection of the work status.

[0066] On the other hand, in the case of dynamic work in which the worker's movements are relatively large, using independent detection to detect multiple states including "Prevention: Uncertain" allows for flexible response to momentary movements and more accurate judgment. Furthermore, the detection unit 31 can select a detection method depending on the type of work as well as the intensity of the worker's movements.

[0067] In independent detection, when detecting whether a factory worker is working with a soldering iron, the detection unit 31 may obtain a detection result that the probability of "work: yes" is 75% and the probability of "work: no" is 75%. In this case, the sum of the probabilities of the two states exceeds 100%, indicating that multiple states may exist simultaneously for the same data.

[0068] For example, a situation where a soldering iron is being used may be confused with a situation where a tool similar to a soldering iron is being used. By detecting multiple states simultaneously, a more detailed understanding of the situation is possible, which can reflect the complexities of real-world situations.

[0069] In non-independent detection, when detecting whether a factory worker is using a soldering iron, the detection unit 31 may output a detection result in which the probability of "operation: present" is 75%, the probability of "operation: absent" is 15%, and the probability of "operation: uncertain" is 10%. In this case, the probabilities of each of the three states are not independent, and the sum of these three states is 100%. This indicates that each state is mutually exclusive. For example, this reflects the fact that a worker may either be using a soldering iron, not using a soldering iron at all, or be indefinite. In non-independent detection, because each state is not independent, clear and exclusive judgments are possible, which can simplify analysis. On the other hand, in independent detection, because each state is independent, the output is non-exclusive, which can complicate analysis.

[0070] The processing unit 32 outputs the detected work status to the determination unit 4. In the case of non-independent detection, one of two work status states, "Work: Caution" and "Work: Safe", is output. In the case of independent detection, one of three work status states, "Work: Caution", "Work: Safe", and "Work: Undefined", is output.

[0071] As shown in Figure 4, the judgment unit 4 accepts input of the prevention state detected by the prevention detection unit 2 ("Prevention: Yes" or "Prevention: No" or "Prevention: Undefined") and the work state detected by the work detection unit 3 ("Work: Caution" or "Work: Safe" or "Work: Undefined"), and outputs one judgment result based on the combination of each prevention state and each work state.

[0072] As shown in FIG. 5, the determination unit 4 performs a matrix analysis by using the two detection results of the preventive detection unit 2 as columns and the two detection results of the work detection unit 3 as rows.

[0073] If the result of the prevention detection unit 2 is "Prevention: Yes" and the result of the work detection unit 3 is "Work: Deadline", the determination unit 4 outputs a determination result of "Work: Possible". The determination result of "Work: Possible" indicates that the worker may continue the specified work.

[0074] If the result of the prevention detection unit 2 is "Prevention: No" and the result of the work detection unit 3 is "Work: Caution", the determination unit 4 outputs a determination result of "Work: Not Possible". The determination result of "Work: Not Possible" indicates that the worker should stop the specified work.

[0075] If the result of the prevention detection unit 2 is "Prevention: Yes" and the result of the work detection unit 3 is "Work: Safe", the determination unit 4 outputs a determination result of "Work: Standby". The determination result of "Work: Standby" indicates that the worker has not yet performed the specified work and is waiting.

[0076] If the result of the prevention detection unit 2 is "prevention: not performed" and the result of the work detection unit 3 is "work: not performed", the determination unit 4 similarly outputs the determination result of "work: waiting".

[0077] In this way, the decision unit 4 outputs one of the four decision results of the 2x2 matrix.

[0078] As shown in FIG. 6, the determination unit 4 performs a matrix analysis by treating the three detection results of the preventive detection unit 1 as columns and the three detection results of the work detection unit 2 as rows.

[0079] Independent detection has a 3x3 matrix compared to the 2x2 matrix of non-independent detection shown in Figure 5. The indefinite items in independent detection will be explained below.

[0080] If the result of the prevention detection unit 2 is "Prevention: Uncertain" and the result of the work detection unit 3 is "Work: Caution", the determination unit 4 outputs a determination result of "Work: Unavailable". The determination result of "Work: Unavailable" indicates that the worker should stop the specified work.

[0081] If the result of the prevention detection unit 2 is "Prevention: Uncertain" and the result of the work detection unit 3 is "Work: Safe", the determination unit 4 outputs a determination result of "Work: Standby". The determination result of "Work: Standby" indicates that the worker has not yet performed the specified work and is waiting.

[0082] If the result of the activity detection unit 3 is "activity: indeterminate," the determination unit 4 outputs a determination result of "activity: unknown" regardless of the result of the preventive detection unit 2. The determination result of "activity: unknown" indicates that it is unknown whether the worker is performing the specified activity or not.

[0083] The determination unit 4 outputs one of four determination results from the 2x2 matrix for non-independent detection or one of nine determination results from the 3x3 matrix for independent detection as a determination result. As shown in Fig. 1, the determination result from the determination unit 4 may be output to an external device 9 connected to the determination unit 4 via a wired or wireless connection.

[0084] Examples of external devices 9 connected to the judgment unit 4 include an alert device that issues an audio or visual alert depending on the judgment result, a notification device that sends a notification to a supervisor or manager depending on the judgment result, a power control device that forcibly turns off the power to work equipment depending on the judgment result, a storage device that saves the judgment result as a log, a monitoring device that monitors and displays the work status in real time, an access control device that controls access to the work area depending on the judgment result, a feedback device that provides feedback to the worker and encourages improvement, an environmental control device that controls the temperature, humidity, lighting, etc. of the work environment depending on the judgment result, and a robot assist device that uses a robot to assist the worker based on the judgment result.

[0085] 7, the safety monitoring process is started and ended in steps S10 and S14, respectively. The triggers for starting and ending the safety monitoring process may be selected arbitrarily depending on the detection target. For example, it may be a time-based trigger that starts sampling at 9:00 AM when the factory shift starts and stops at 5:00 PM when the shift ends, an action-based trigger that starts and stops in response to a worker turning on or off the workplace lights, or a sensor-detected trigger that starts and stops in response to the detection of a worker by the first and second sensors or another sensor.

[0086] Furthermore, it may be an equipment operation trigger that starts when a specific device or facility related to the work starts operating and ends when it stops, or a manual trigger that an administrator or worker manually instructs to start or end monitoring, or an event-based trigger that starts or ends in response to a specific event such as an abnormality detection or alarm occurrence.

[0087] Note that after the start or end of safety monitoring processing, the two states "Work: Caution" and "Work: Safe" alternate and coexist in a short period of time, which can result in "Work: Undefined." If the frequency of changes in the detection results "Work: Caution" and "Work: Safe" from the work detection unit 3 within a predetermined period of time is equal to or greater than a certain level (for example, if the mixed state continues for an instant), the determination unit 4 may consider this to be the start or end of work, and determine the state to be "Work: Caution."

[0088] In response to the start of step S10, detection of a preventive state by the prevention detection unit 2 in step S11 and detection of a work state by the work detection unit 3 in step S12 are performed. It is preferable to start steps S11 and S12 at the same time. By collecting data at the same time, data on the preventive state and the work state are synchronized, consistent data is obtained, and a correlation between the two can be found during analysis.

[0089] In step S13, the judgment unit 4 makes a judgment, and one of the four judgment results from the 2x2 matrix for non-independent detection or one of the nine judgment results from the 3x3 matrix for independent detection is output as the judgment result.

[0090] Regarding the probabilities provided by the detection unit 21, if the probability of the two preventive states is below a predetermined value, for example, the probability of "prevention: yes" is 10% and the probability of "prevention: no" is 10%, if the probability of the two preventive states is above a predetermined value, for example, the probability of "prevention: yes" is 90% and the probability of "prevention: no" is 90%, if the probability of the two preventive states maintains an intermediate value, for example, the probability of "prevention: yes" is 50% and the probability of "prevention: no" is 50%, or if the probabilities of the two preventive states, "prevention: yes" and "prevention: no" fluctuate greatly over time, it is assumed that there is a malfunction in the first sensor 7, and the judgment unit 4 may make a judgment of "work: no".

[0091] Similarly, with regard to the probabilities provided by the detection unit 31 described later, if the probabilities of the two work states are below a predetermined value, for example, the probability of "Work: Caution" is 10% and the probability of "Work: Safe" is 10%, it is assumed that there is a malfunction in the second sensor 8, and the judgment unit 4 may make a judgment of "Work: Unavailable."

[0092] Using the probabilities provided by the detection units 21 and 31, it is possible to quickly estimate malfunctions in the first sensor 7 and the second sensor 8, respectively, improving the reliability of the entire system and enabling action to be taken before a problem due to a malfunction occurs. Inappropriate work based on a malfunctioning sensor can be prevented, ensuring the safety of workers and reducing the risk of accidents.

[0093] (Modification 1 of Embodiment 1: Learning Unit 5) As shown in Fig. 8, safety monitoring system 10 includes a prevention detection unit 2, an activity detection unit 3, and a determination unit 4, and differs from safety monitoring system 1 in Fig. 1 in that work detection unit 3 is provided with a learning unit 5. Although Fig. 8 shows that work detection unit 3 is provided with the learning unit 5, the learning unit 5 may be provided in prevention detection unit 2, or in both prevention detection unit 2 and work detection unit 3.

[0094] The preventive state detected by the preventive detection unit 2 is determined based on whether the worker is wearing protective equipment appropriate for the work and / or whether safety measures have been taken in the work environment. For example, there is a possibility that the protective equipment being detected may be confused with an item similar in shape, etc.

[0095] For example, if a worker wears glasses on a daily basis to assist with vision and then wears protective glasses over the glasses when working with a soldering iron, the preventive detection unit 2 may mistakenly recognize the everyday glasses as protective glasses, or vice versa, making an erroneous detection.

[0096] Furthermore, in the work environment, there is a possibility that the target object may be confused with an object similar in shape, etc. Such false detection may result in workers working without wearing protective eyewear, which could be a safety issue and may cause unnecessary warnings or alerts, reducing work efficiency.

[0097] Furthermore, the task detection unit 3 may confuse the task being detected with a similar task. For example, if a worker is operating a smartphone at hand, the task detection unit 3 may mistakenly detect the smartphone operation as soldering iron work, which also requires manual work. Such erroneous detection may result in the soldering iron being turned on despite the worker operating the smartphone at hand, or the soldering iron being turned off despite the worker being using the soldering iron, resulting in unnecessary confirmation work and potentially posing a safety concern.

[0098] The learning unit 5 is provided to appropriately distinguish between imminent matters that the prevention detection unit 2 may mistakenly detect and safety measures that are the target of detection, and to appropriately distinguish between imminent similar actions that the work detection unit 3 may mistakenly detect and specified work that is the target of detection.

[0099] The learning unit 5 is composed of a trained model generation unit (not shown) that generates a trained model, and an inference unit (not shown) that uses the trained model to detect the preventive state and / or the working state. The trained model generation unit uses data obtained by the first sensor and / or the second sensor as training data to train a model of the preventive state and / or the working state. The trained model generation unit generates a trained model as a result of this model training. The trained model may be used to infer both the preventive state and the working state, or there may be two learning models used to infer the preventive state and the working state, respectively.

[0100] The training data may be based on data obtained before the system is put into operation, or may be based on data obtained within a specified period of time after the system has been put into operation. The trained model generation unit uses machine learning to train a model of the prevention state and / or the operation state. Deep learning or the like may be used as the machine learning model. The trained model generation unit outputs the trained model obtained as a result of the model training to the inference unit. The trained model generation unit may store the trained model in memory 103, which will be described later. In this case, the inference unit reads out the trained model from memory 103. The inference unit of the learning unit 5 performs a detection process on data sequentially generated from the first sensor and / or the second sensor using a trained model for the preventive state and / or the working state. The inference unit may output the result of the detection process to a display device 105 described later. The inference unit may also store the detection result in the memory 103.

[0101] In addition to the data obtained by the first sensor and / or the second sensor, the trained model generation unit may additionally learn data as training data regarding objects or states that are easily mistaken for having safety measures in place, objects or states that are easily mistaken for having no safety measures in place, objects or actions that are easily mistaken for having performed a predetermined task, and objects or actions that are easily mistaken for having not performed a predetermined task.The trained model generation unit can also use data regarding the work environment, such as temperature, humidity, illuminance, and noise level, as training data to build a model that also takes into account the effects of changes in the environment.

[0102] The learning unit 5 significantly reduces false positives, thereby reducing the risk of workers performing work without wearing appropriate protective equipment. In addition, by confirming the appropriate work status, it reduces the number of accidental power on / offs of work equipment and unnecessary alerts and interruptions due to false positives. This allows workers to work efficiently and reduces their stress. Furthermore, it reduces the occurrence of the detection result "Work: Undefined" that can be output from the work detection unit.

[0103] As described above, according to the first embodiment, the preventive state in the work environment, i.e., whether or not safety measures have been taken, and the work state in the work environment, i.e., whether or not a predetermined task is being performed, are detected, and the preventive state and the work state are correlated to determine whether or not the task can be performed. This ensures thorough safety measures for workers, and by preventing any deterioration in safety, the safety of the work environment is improved, while the efficiency of the worker's work can be improved. Furthermore, the inclusion of a learning unit can reduce false detections and improve detection accuracy.

[0104] (Embodiment 2) 9 to 26, a configuration example and an operation example of the safety monitoring system 100 according to the second embodiment of the present disclosure will be described. As shown in FIG. 9, the safety monitoring system 100 includes a prevention detection unit 20, an operation detection unit 30, a determination unit 40, and a determination auxiliary information acquisition unit 60.

[0105] The preventive detection unit 20, the work detection unit 30, and the determination unit 40 are similar to the preventive detection unit 2, the work detection unit 3, and the determination unit 4 in the first embodiment, and therefore details of the similar configurations and functions will be omitted.

[0106] The judgment auxiliary information acquisition unit 60 generates judgment auxiliary information based on the setting values ​​input by the user and the information output from the preventive detection unit 2 and the work detection unit 3, and outputs the information to the judgment unit 40. The setting values ​​input by the user may be preset values ​​(standard setting values) that have been set in advance, or may be setting values ​​that can be interactively adjusted using a UI (user interface).

[0107] The setting values ​​input by the user include, for example, thresholds for the detection probability of the preventive detection unit 2 and the work detection unit 3, the priority of the detection results, and the validity period for obtaining the detection results. The information output from the preventive detection unit 2 and the work detection unit 3 includes the detection results, interim results, log information, sensor information, etc. from these detection units.

[0108] The judgment auxiliary information generated by the judgment auxiliary information acquisition unit 60 may be at least one of priority order information, threshold setting information, and time series information, or may be any combination of priority order information, threshold setting information, and time series information. By combining priority order information, threshold setting information, and time series information, it becomes possible to more accurately judge the safety state of the worker.

[0109] Each piece of information complements the others, effectively eliminating momentary false positives and uncertain states, improving overall detection accuracy. In addition, because auxiliary judgment information can be arbitrarily combined, flexible settings can be made to suit each work environment and work content. This makes it easier to respond to a variety of work sites and situations, providing a highly versatile system. Using multiple pieces of auxiliary judgment information enables comprehensive judgment without relying on individual detection results.

[0110] This improves the reliability of the entire system and further strengthens worker safety monitoring. By utilizing priority information and threshold setting information, high-risk situations can be detected quickly and accurately, enabling appropriate responses. This helps prevent accidents and problems and ensures safety at the work site. By utilizing time-series information, it is possible to deal with temporary fluctuations and make stable judgments. This prevents unnecessary work interruptions.

[0111] (Example 1 of priority information) The priority information is the priority assigned to multiple detection results of the preventive detection unit 2, the priority assigned to multiple detection results of the work detection unit 3, and the priority assigned to the detection results of the preventive detection unit 2 and the detection results of the work detection unit 3. The priority is information that serves as a criterion for evaluating the detection results of the preventive detection unit 2 and the work detection unit 3 and deciding which result should be used first in order to ensure the safety of workers to the maximum extent possible.

[0112] Priorities are set based on, for example, an arbitrary numerical value set by the user, the risk of not taking safety measures, the existence of applicable laws and regulations, the type of work and its safety, the worker's experience and skills, the work environment, the type of protective equipment used and its performance, data on past accidents and troubles, work time and workload, the accuracy of preventive detection and work detection, etc.

[0113] In addition, the priority may be, for example, given priority to the detection results of the preventive detection unit in a work environment with a high risk of accidents, and given priority to the detection results of the work detection unit in a work environment with a low risk of accidents. Furthermore, the priority may be changed depending on the time of day. For example, given priority to the detection results of the work detection unit during the daytime when it is relatively bright, and given priority to the detection results of the preventive detection unit during nighttime work when it is relatively dark.

[0114] (Example 2 of priority information: Example of prioritizing the detection results of preventive detection unit 2) As shown in Figure 10, when the result that can be detected by the activity detection unit 3 is "activity: uncertain," if the prevention detection unit 2 detects "prevention: none," the normal judgment would result in the judgment result of the judgment unit 4 being "activity: unknown." However, the result of "prevention: none" by the prevention detection unit 2 is given priority, and the judgment result of the judgment unit 4 is corrected from the normal judgment of "activity: unknown" to the judgment result of "activity: impossible," and output. This allows the prevention state to be given priority.

[0115] Examples of situations in which the detection results of the prevention detection unit 2 should be given priority over the detection results of the work detection unit 3 include situations in which starting work without wearing protective equipment increases the risk of accidents or injuries, or situations in which wearing protective equipment is required by law. In this way, by prioritizing the prevention state of the prevention detection unit 2, the risk of accidents or injuries due to not wearing safety equipment can be avoided.

[0116] (Example 3 of priority information: Example of prioritizing the detection result of the work detection unit 3) As shown in Figure 11, when the work detection unit 3 detects "Work: Caution" and the possible result detected by the prevention detection unit 2 is "Prevention: Uncertain," if it were a normal judgment, the judgment unit 4 would have determined "Work: Unavailable." However, the detection result of the work detection unit 3 is given priority, and the judgment unit 4 corrects the judgment result from the normal judgment of "Work: Unavailable" to "Work: Available" and outputs it. This allows the work status to be given priority.

[0117] An example of a situation in which the detection result of the work detection unit 3 may be prioritized is when handling high-temperature materials, where interrupting work can actually reduce safety, and it may be safer to continue the work while the work detection unit 3 indicates "Work: Caution." Also, when mixing chemicals, stopping the work midway may result in the generation of unsafe chemically reactive substances, and if the work detection unit 3 indicates "Work: Caution," indicating that the worker should continue the work, it may be safer to prioritize the work detection unit 3 over the results of the prevention detection unit 2. In this way, when interrupting work in a given task involves a high risk, the detection result of the work detection unit may be prioritized.

[0118] (Example 4 of priority information: "Prevention: No" takes precedence over "Prevention: Yes") As shown in FIG. 12, in the case of independent detection, the prevention detection unit 2 outputs two prevention states, "prevention: yes" and "prevention: no", along with their respective probabilities.

[0119] If the detection results of the prevention detection unit 2 are both "Prevention: Yes" and "Prevention: No" with a probability of 75%, and the detection result of the work detection unit is "Work: Caution", there are two possible results that can be output from the judgment unit 4: "Work: Possible" and "Work: Not Possible". By prioritizing "Prevention: No", the judgment unit 4 can output "Work: Not Possible". This reduces false positives and leads to worker safety.

[0120] Furthermore, in non-independent detection, if the probabilities of the two states "Prevention: Yes" and "Prevention: No" are the same, the judgment result that can be output from the judgment unit 4 will be "Prevention: Uncertain." However, in the case of "Prevention: Uncertain," the judgment unit 4 will output "Work: Not Possible," which leads to the safety of workers.

[0121] In this example, when the probability of wearing safety glasses is the same as that of not wearing them, the judgment is made to prioritize the state of not wearing safety glasses in order to ensure work safety. By making such a judgment, the risk of the worker continuing to work without taking safety measures can be avoided.

[0122] (Example 5 of priority information: "Prevention: Yes" takes precedence over "Prevention: No" and "Prevention: Uncertain") 13, in the case of independent detection, the prevention detection unit 2 outputs two prevention states, "Prevention: Yes" and "Prevention: No," along with their respective probabilities. If the detection results of the prevention detection unit 2, "Prevention: Yes" and "Prevention: No," both have a probability of 75%, and the detection result of the work detection unit 3 is "Work: Caution," there are two possible determination results that can be output from the judgment unit 4: "Work: Possible" and "Work: Not Possible." By prioritizing "Prevention: Yes," the judgment unit 4 can output "Work: Possible."

[0123] Furthermore, in independent detection, if the probabilities of the two states "Prevention: Yes" and "Prevention: No" are the same value, the judgment result that can be output from the judgment unit 4 will be "Prevention: Uncertain", and the judgment unit 4 will output "Work: Not Possible". However, by prioritizing "Prevention: Yes" over "Prevention: Uncertain", the judgment unit 4 can output "Work: Possible".

[0124] This allows the worker to continue working without interruption. Examples of situations that could arise include when the worker is wearing safety glasses but is not wearing them correctly, or when an obstruction makes it difficult for the sensor to fully detect the safety glasses. This allows for both work efficiency and safety, ensuring the continuous safety of workers.

[0125] (Example 6 of priority information: "Work: Safety" takes priority over "Work: Caution" and "Work: Unknown") As shown in FIG. 14, in the case of independent detection, the activity detection unit 3 outputs two activity states, "activity: caution" and "activity: safety," along with their respective probabilities.

[0126] If the probability of "Work: Caution" and "Work: Safe" are both 75% and the detection result of the prevention detection unit 2 is "Prevention: Yes", there are two possible judgment results that can be output from the judgment unit 4: "Work: Possible" and "Work: Standby". By prioritizing "Work: Safe", the judgment unit 4 can output "Work: Standby".

[0127] A situation in which "work: safety" should be prioritized would be to reduce false positives that indicate that a worker is performing a specified task when in fact the worker is not, thereby ensuring the worker's safety.

[0128] Furthermore, in non-independent detection, if the probabilities of the two states "Work: Caution" and "Work: Safe" are the same value, the judgment result that can be output from the judgment unit 4 will be "Work: Uncertain", and the judgment unit 4 will output "Work: Unknown". However, by prioritizing "Work: Safe" over "Work: Unknown", the judgment unit 4 can output "Work: Waiting".

[0129] (Example 7 of priority information: "Work: Caution" takes priority over "Work: Safety" and "Work: Unknown") 15, in the case of independent detection, the work detection unit 3 outputs two work states, "Work: Caution" and "Work: Safe", along with their respective probabilities. If the detection results of "Work: Caution", "Work: Caution" and "Work: Safe", both have probabilities of 75%, and the detection result of the prevention detection unit 2 is "Prevention: Yes", there are two possible judgment results that can be output from the judgment unit 4: "Work: Possible" and "Work: Standby", and by prioritizing "Work: Caution", the judgment unit 4 can output "Work: Possible".

[0130] An example of a situation in which "Work: Caution" should be prioritized is when abnormal sounds or vibrations are detected during machine maintenance. Even if the probability of "Work: Caution" and "Work: Safety" are equal, prioritizing "Work: Caution" may allow work to continue, detect the abnormality early, and prevent breakdowns or accidents.

[0131] Furthermore, in non-independent detection, if the probabilities of the two states "Task: Caution" and "Task: Safe" are the same value, the judgment result that can be output from the judgment unit 4 will be "Task: Uncertain," and the judgment unit 4 will output "Task: Unknown." However, by prioritizing "Task: Caution" over "Task: Unknown," the judgment unit 4 can output "Task: Possible."

[0132] (Example 1 of threshold information) The judgment assistance information may be threshold information. "Threshold information" refers to a predetermined reference value (threshold) for the detection results obtained from the preventive detection unit 2 and the work detection unit 3. This information serves as a judgment standard for ensuring the safety of workers, and setting an appropriate threshold enables accurate detection and judgment. The threshold information is set based on, for example, an arbitrary numerical value set by the user, the risk of not wearing protective gear, the existence of applicable laws and regulations, the type of work and its safety, the worker's experience and skills, the work environment, the type of protective gear used and its performance, data on past accidents and problems, work time and workload, the accuracy of preventive detection and work detection, physical quantities obtained from sensors, etc.

[0133] (Example 2 of threshold information: Lowering the threshold of the preventive detection unit 2 and raising the threshold of the work detection unit 3) As shown in Figure 16, we will explain a situation in which the prevention detection unit 2 outputs a detection result of "Prevention: Yes" when the probability of detecting "Prevention: Yes" is equal to or greater than the threshold of 50%, and the work detection unit 3 outputs a detection result of "Work: Caution" when the probability of detecting "Work: Caution" is set to a threshold of 80% or greater.

[0134] For example, when detecting the wearing of protective glasses, the prevention detection unit 2 determines that "prevention is required" with a relatively low probability (50% or more), while the work detection unit 3 determines that the work status is "work: caution" with a high probability (80% or more). With this setting, the prevention detection unit 2 more proactively detects the wearing of protective glasses, and this low threshold setting makes it more likely to determine that "prevention is required" even with some uncertainty. Because the work detection unit 3 sets a high probability to determine that the work status is "work: caution," it can reduce false positives and detect "work: caution" when there is a high possibility that the work is actually being performed.

[0135] (Threshold information example 3: Increasing the threshold of the preventive detection unit 2 and decreasing the threshold of the work detection unit 3) As shown in Figure 17, we will explain a situation in which the prevention detection unit 2 outputs a detection result of "Prevention: Yes" when the probability of detecting "Prevention: Yes" is equal to or higher than the threshold of 80%, and the work detection unit 3 outputs a detection result of "Work: Caution" when the probability of detecting "Work: Caution" is set to a threshold of 50% or higher.

[0136] For example, when detecting the wearing of protective equipment, the prevention detection unit 2 judges it as "Prevention: Yes" with a relatively high probability (80% or more), while the work detection unit 3 judges the work status as "Work: Caution" with a low probability (50% or more). With this setting, the prevention detection unit 2 will not detect the wearing of protective equipment unless there is a high probability, which ensures that the protective equipment is being worn properly, further enhancing the safety of the worker. On the other hand, the work detection unit 3 detects "Work: Caution" relatively easily, which means that it detects that the work is being done relatively continuously, thereby reducing interruptions to work.

[0137] (Example 4 of threshold information: Setting the threshold for "Prevention: Yes" higher than the threshold for "Prevention: No") As shown in Figure 18, the following setting conditions are explained: when the probability of "Prevention: Yes" in the prevention detection unit 2 is equal to or greater than a threshold of 90%, the detection result of "Prevention: Yes" is output; when the probability of "Prevention: Yes" is less than the threshold of 90%, the detection result of "Prevention: Yes" is not output; when the probability of "Prevention: No" is equal to or greater than a threshold of 50%, the detection result of "Prevention: No" is output; and when the probability of "Prevention: No" is less than the threshold of 50%, the detection result of "Prevention: Yes" is not output.

[0138] If the detection result from the prevention detection unit 2 shows an 80% probability of "prevention: yes" and a 60% probability of "prevention: no," the 80% probability of "prevention: yes" is less than the 90% threshold for "prevention: yes," but the 60% probability of "prevention: no" is greater than or equal to the 50% threshold for "prevention: no," so in this case the judgment unit 4 outputs a detection result of "work: no." By setting the threshold for "prevention: yes" in the prevention detection unit 2 higher, it is possible to impose stricter restrictions on the wearing of protective equipment. This reduces the risk of continuing work without wearing protective equipment, ensuring the safety of workers. By setting a high threshold, it is possible to reduce the number of cases where "prevention: yes" is mistakenly determined, ensuring the safety of workers.

[0139] (Example 5 of threshold information: Setting the threshold for "Prevention: No" higher than the threshold for "Prevention: Yes") As shown in Figure 19, the following setting conditions are explained: when the probability of "prevention: yes" in the prevention detection unit 2 is equal to or greater than a threshold of 50%, the detection result of "prevention: yes" is output; when the probability of "prevention: yes" is less than the threshold of 50%, the detection result of "prevention: yes" is not output; when the probability of "prevention: no" is equal to or greater than a threshold of 90%, the detection result of "prevention: no" is output; and when the probability of "prevention: no" is less than the threshold of 90%, the detection result of "prevention: yes" is not output.

[0140] If the detection results from the prevention detection unit 2 indicate a 60% probability of "prevention: yes" and an 80% probability of "prevention: no," the 60% probability of "prevention: yes" is above the 50% threshold for "prevention: yes," but the 80% probability of "prevention: no" is below the 90% threshold for "prevention: no," in this case, the prevention detection unit 2 prioritizes the "prevention: yes" detection result, and the judgment unit 4 outputs a judgment result of "work: possible." Setting a lower threshold for "prevention: yes" in the prevention detection unit 2 allows for more flexible responses to the wearing status of preventive equipment. This reduces the risk of continuing work without wearing protective equipment, ensuring the safety of workers.

[0141] (Example 6 of threshold information: An example in which the threshold for "Work: Caution" of the work detection unit 3 is set higher than the threshold for "Work: Safety") As shown in Figure 20, the setting situation is explained in which the detection result "Work: Caution" is output when the probability of "Work: Caution" in the work detection unit 3 is equal to or greater than the threshold of 90%, the detection result "Work: Caution" is not output when the probability of "Work: Caution" is less than the threshold of 90%, the detection result "Work: Caution" is output when the probability of "Work: Safe" is equal to or greater than the threshold of 50%, and the detection result "Work: Safe" is not output when the probability of "Work: Safe" is less than the threshold of 50%.

[0142] If the detection results from the work detection unit 3 show an 80% probability of "Work: Caution" and a 60% probability of "Work: Safe", the 80% probability of "Work: Caution" is less than the 90% threshold for "Work: Caution", but the 60% probability of "Work: Safe" is equal to or greater than the 50% threshold for "Work: Safety", so in this case, a determination of "Work: Standby" is output based on the detection result of "Work: Safe" from the work detection unit 3. Setting a high threshold for "Work: Caution" in the work detection unit 3 is appropriate for cases where a relatively high probability of being detected as a specified work action by a worker is required.

[0143] (Example 7 of threshold information: Set the threshold for "Work: Safety" higher than the threshold for "Work: Caution") As shown in Figure 21, the setting situation is explained in which the detection result "Work: Caution" is output when the probability of "Work: Caution" in the work detection unit 3 is equal to or greater than a threshold of 50%, the detection result "Work: Caution" is not output when the probability of "Work: Caution" is less than the threshold of 50%, the detection result "Work: Caution" is output when the probability of "Work: Safe" is equal to or greater than a threshold of 90%, and the detection result "Work: Safe" is not output when the probability of "Work: Safe" is less than the threshold of 90%.

[0144] If the detection result from the work detection unit 3 is that the probability of "Work: Caution" is 60% and the probability of "Work: Safe" is 80%, the probability of "Work: Caution" of 60% is greater than the threshold of "Work: Caution" of 50%, but the probability of "Work: Safe" of 80% is less than the threshold of "Work: Safe" of 90%.In this case, based on the detection result of "Work: Caution" from the work detection unit 3, the judgment unit 4 outputs the judgment result of "Work: Possible".

[0145] Setting a high threshold for "Work: Safe" in the work detection unit 3 is effective in cases where a worker is likely to be detected as not working despite actually working. For example, this can be the case when a worker is performing work that requires delicate movements such as assembling precision equipment, when performing intermittent work, when performing work with little physical movement, when working while moving, when performing work that requires quick movements, or when multiple workers are performing different tasks.

[0146] (Example 1 of time series information) The judgment assistance information may be time-series information of the detection results of the preventive detection unit 2 and the work detection unit 3. "Time-series information" is information that takes into account temporal fluctuations by collecting the detection results of the preventive detection unit 2 and the work detection unit 3 over a predetermined validity period. This allows the system to make stable judgments by taking into account instantaneous fluctuations and temporary abnormalities. The validity period of the time-series information from the various detection units is determined by the frequency at which the first and second sensors 7 and 8 collect data, the frequency at which workers put on and take off protective equipment, settings by the administrator or user, etc.

[0147] For example, the time series information may be the detection results for the past 10 seconds from the time when an instantaneous fluctuation or temporary abnormality occurred among the detection results output from the preventive detection unit 2 and / or the operation detection unit 3. Alternatively, the time series information may be the detection results for the past 5 seconds from the time when an instantaneous fluctuation or temporary abnormality occurred among the detection results output from the preventive detection unit 2, or the detection results for the past 6 seconds from the time when an instantaneous fluctuation or temporary abnormality occurred among the detection results output from the preventive detection unit 2.

[0148] (Time series information example 2: Example of interpolating the results of an instantaneous, uncertain state using time series information) The time-series information in Example 2 is the determination result of the determination unit 4 for a predetermined valid time going back from the present time. The predetermined valid time may be set arbitrarily by the user, for example, or may be set to a relatively short time in the case of dynamic work in which the movements of the predetermined work are relatively large, or to a relatively long time in the case of dynamic work in which the movements of the predetermined work are relatively small.

[0149] As shown in FIG. 22, the detections by the preventive detection unit 2 and the work detection unit 3 are not necessarily stable, and the detection result from the preventive detection unit 2 may momentarily become "Prevention: Uncertain" and / or the detection result from the work detection unit 3 may momentarily become "Work: Uncertain." The detection results by the preventive detection unit 2 and the work detection unit 3 may momentarily become unstable, for example, when another person passes between the sensor and the worker, when detection is not performed due to a sudden change in lighting, when the worker makes a sudden movement, or when the sensor malfunctions. Even if the detection results from the preventive detection unit 2 and the work detection unit 3 become unstable, the system is configured to use the determination results within the valid time as time-series information to interpolate the momentarily unstable detection results into a time series.

[0150] For example, if the detection result of the determination unit 4 over 10 seconds is "Work: Possible," and the detection result from the prevention detection unit 2 momentarily becomes "Prevention: Uncertain" and / or the detection result from the work detection unit 3 becomes "Work: Uncertain" for 1 second, which is shorter than the sampling period of the prevention detection unit 2 and the work detection unit 3, the determination unit 4 corrects the momentary results of "Prevention: Uncertain" and "Work: Uncertain" based on the chronological determination results going back through the past valid time, and outputs the determination result of "Work: Possible." By processing the detection results chronologically, momentary uncertain states that are shorter than the sampling period of the prevention detection unit 2 and the work detection unit 3 can be avoided, a correct determination can be made, and fluctuations in the detection of the prevention detection unit 2 and the work detection unit 3 can be reduced, ensuring the safety of the worker.

[0151] (Example 3 of time series information: Prioritizing unsafe conditions) The time-series information in Example 3 is the determination result of the determination unit 4 for a predetermined valid time going back from the present time. As shown in Fig. 23, the working environment or the condition of the worker may suddenly change, and an unsafe situation (a "Work: Unavailable" state) may momentarily occur. Specifically, if the worker suddenly falls and loses protective equipment, but then gets up and puts the protective equipment back on, the determination result may momentarily output as "Work: Unavailable," but a determination result such as "Work: Unknown" or "Work: Unknown" may also be output.

[0152] In such cases, if a "Work: Unavailable" judgment is made even momentarily within a predetermined time, the judgment is made "Work: Unavailable" to ensure the safety of the worker. The predetermined time is set based on, for example, the type and safety of the work, the worker's experience and skills, the work environment, the type and performance of safety equipment used, data on past accidents and problems, the number of work steps, the work time and workload, the accuracy of preventive detection and work detection, etc. Furthermore, if a predetermined work is composed of multiple steps with different safety levels, unsafe conditions can be prioritized according to safety. For steps with a relatively high level of safety, a normal judgment can be made, and for steps with a relatively low level of safety, unsafe conditions can be prioritized.

[0153] (Example 4 of time series information: Prioritizing safety conditions) The time-series information in Example 4 is the determination result of the determination unit 4 for a predetermined valid time going back from the present time. As shown in Fig. 24, there are cases where the determination result continues to be "Work: Standby", then momentarily becomes "Work: Possible", and then returns to "Work: Standby" again. For example, if the predetermined work is welding, the period during which the welding work is being prepared will be "Work: Standby", and when an arc for welding work is instantaneously generated, the period will be "Work: Possible", and when no arc is generated, the state will be "Work: Standby".

[0154] In such cases, if a "Work: Possible" judgment is made even momentarily within a specified valid time, it can be determined to be "Work: Possible" even if the usual judgment would be "Work: Waiting." This reduces unnecessary "Work: Waiting" judgments. By preventing unnecessary work interruptions, work efficiency is improved. Work can be continued even in the event of momentary fluctuations, maintaining worker productivity. Furthermore, by reducing unnecessary waiting time, overall work efficiency is improved.

[0155] (Example 5 of time-series information: Correct the results of the preventive detection unit 2 with the results of the work detection unit 3) The time-series information in Example 5 is the result of the work detection unit 3 for a predetermined valid time going back from the present time. As shown in Figure 25, there is a possibility that ambiguity may occur in the prevention detection unit 2. For example, if the area becomes temporarily dark due to a power supply problem or a malfunction of the lighting equipment, the first sensor 7 may have difficulty accurately recognizing the protective equipment, and the result may temporarily be determined as "Prevention: Uncertain."

[0156] In such a case, if the detection result of the prevention detection unit 2 temporarily changes from "Prevention: Yes" to "Prevention: Uncertain," but the detection result of the work detection unit 3 remains "Work: Caution" without being affected by the temporary change, the prevention detection unit 2 maintains the "Prevention: Yes" state before the temporary change. For example, the result of the prevention detection unit 2 is corrected by the result of the work detection unit 3. This allows the judgment unit 4 to maintain the judgment result of "Work: Possible," suppress the generation of unnecessary alerts, and avoid unnecessary confusion for workers and supervisors. This prevents work from being interrupted due to temporary fluctuations in the prevention detection unit 2, and maintains work efficiency.

[0157] (Example 6 of time-series information: Example of correcting the results of the work detection unit 3 with the results of the preventive detection unit 2) The time-series information in Example 6 is the result of the preventive detection unit 2 for a predetermined valid time going back from the present time. As shown in Fig. 26, the detection result from the operation detection unit 3 may continue as "Operation: Caution" and then temporarily become "Operation: Undefined." For example, if the predetermined operation (the operation to be detected) is soldering, the detection result may be "Operation: Caution" during the period when the soldering operation is being performed, and the detection result may be "Operation: Undefined" during the period when the soldering tool is temporarily returned to the tool holder.

[0158] In such a case, if the detection result of the prevention detection unit 2 remains "Prevention: Yes" within a predetermined valid time, the judgment unit 4 corrects the result to "Work: Possible" even if the normal judgment result would be "Work: Uncertain." This makes it possible to avoid unnecessary "Work: Uncertain" states while ensuring the safety of workers, prevent work interruptions, and maintain work efficiency.

[0159] As described above, according to this embodiment, by combining judgment auxiliary information (priority information, threshold setting information, and time series information), it is possible to eliminate momentary erroneous detections and unstable states, improve overall detection accuracy, and adapt to a variety of work sites.

[0160] The judgment unit 4 uses priority information to prioritize appropriate detection results in high-risk situations, ensuring worker safety. It can also adjust the priority according to user settings and changes in the work environment, making it possible to adapt to a variety of work sites and situations, improving the versatility of the system. It can also eliminate false detections and uncertain states and make accurate judgments. Prioritizing the results of the work detection unit allows workers to continue work safely in situations where interrupting work would actually decrease safety.

[0161] Furthermore, by using threshold setting information to make judgments, the judgment unit 4 can eliminate false detections and uncertain states, and by setting appropriate thresholds, it is possible to strictly judge the wearing status of protective equipment and the safety status of work, thereby further ensuring the safety of workers. Furthermore, since the thresholds can be set arbitrarily by the user, flexible settings according to the work environment and work content are possible, and it is possible to respond to a variety of work sites and situations.

[0162] Furthermore, in high-risk work or environments, strict threshold settings can quickly detect unsafe situations, making it possible to prevent accidents and problems before they occur. By appropriately setting the thresholds of the work detection unit 2 to reduce work interruptions, it becomes easier to continue work and unnecessary work interruptions can be prevented, thereby improving work efficiency.

[0163] Furthermore, by using time-series information to make judgments, the judgment unit 4 can ignore momentary fluctuations and temporary abnormalities and make stable judgments, thereby reducing fluctuations in the detection results of the preventive detection unit 2 and the work detection unit 3 and making accurate judgments. Furthermore, by interpolating the detection results of momentary unstable states with time-series information, false detections can be eliminated and stable judgments can be made even against temporary malfunctions of the sensor or sudden changes in the environment.

[0164] Furthermore, by utilizing time-series information, temporary fluctuations can be dealt with and stable judgments can be made, preventing unnecessary interruptions to work and improving work efficiency. By mutually correcting the detection results of the preventive detection unit 2 and the work detection unit 3 based on time-series information, detection accuracy is improved and more accurate judgments can be made.

[0165] (Embodiment 3) The prevention detection unit 2 according to the first embodiment detects two states, "prevention: with" and "prevention: without," or three states, "prevention: with," "prevention: without," and "prevention: undefined." As shown in Fig. 27, the prevention detection unit 2 according to the third embodiment is configured to detect things or states that "are likely to be mistaken if prevention is not taken" and things or states that "are likely to be mistaken if prevention is taken," instead of "prevention: undefined."

[0166] An item or state that is "likely to be mistaken for not taking preventive measures" refers to an item or state that may be mistaken for "no preventive measures" even when appropriate preventive measures are in place. For example, when using transparent protective glasses as safety equipment, the safety equipment, which is difficult for the preventive measures detection unit 2 to detect because of its transparency, falls under the category of an item or state that is "likely to be mistaken for not taking preventive measures." An item or state that is "likely to be mistaken for taking preventive measures" refers to an item or state that may be mistaken for "preventive measures" even when no preventive measures are in place. For example, when a worker wears regular glasses instead of protective glasses, the shape of the protective glasses and regular glasses is similar, so the safety equipment that the preventive measures detection unit is likely to mistaken for "taking preventive measures" falls under the category of an item or state that is "likely to be mistaken for taking preventive measures."

[0167] Furthermore, the activity detection unit 3 according to the first embodiment detects two states, "activity: caution" and "activity: safety," or three states, "activity: caution," "activity: safety," and "activity: uncertain." The activity detection unit 3 according to the third embodiment is configured to detect things or states where "it is easy to make mistakes when working" and things or states where "it is easy to make mistakes when working," instead of "activity: uncertain."

[0168] An object or state that is "likely to be mistaken for not performing work" refers to an object or state that may be mistaken for not performing work, even though the worker is performing the work to be detected. For example, when a worker is performing soldering work, the action of holding solder is mistakenly detected as "not performing work," so the action of holding solder falls under the category of an object or state that is "likely to be mistaken for not performing work." An object or state that is "likely to be mistaken for performing work" refers to an object or state that may be mistaken for performing work, even though the worker is not performing the work to be detected. For example, when a worker is performing soldering work, the action of holding a pencil may be mistaken for "performing work," so the action of holding a pencil falls under the category of an object or state that is "likely to be mistaken for performing work."

[0169] The judgment unit 4 outputs one of the 4x4 judgment results by performing a matrix analysis in the same way as the safety monitoring system according to the first embodiment, based on the four detection results from the preventive detection unit 2 and the four detection results from the work detection unit 3. Note that for the safety monitoring system according to the third embodiment, the judgment result may be adjusted using the judgment auxiliary information according to the second embodiment.

[0170] In addition to the 4x4 matrix, a 5x5 matrix of "Prevention: Undefined" and "Work: Undefined" may also be used.

[0171] As described above, according to this embodiment, by detecting a state in which "it is easy to make a mistake if no prevention is taken" and a state in which "it is easy to make a mistake if prevention is taken," it is possible to reduce erroneous prevention judgments and more accurately detect whether or not a preventive tool is being worn. Furthermore, by detecting a state in which "it is easy to make a mistake if no work is being performed" and a state in which "it is easy to make a mistake if work is being performed," it is possible to improve the accuracy of judgments on whether or not work is being performed, reduce erroneous detections of work, and accurately grasp the actual work situation.

[0172] (Fourth embodiment) This embodiment relates to a technique for improving the detection accuracy of the preventive detection unit 2 and the work detection unit 3 when multiple workers are working in the same area. In embodiment 4, an information acquisition unit is further added to the safety monitoring systems 1, 10, and 100 shown in Figs. 1, 8, and 9.

[0173] When multiple workers are in the same work area, an image captured by the camera sensor of the prevention detection unit 2 may simultaneously capture both "people taking precautions" and "people not taking precautions." This may cause the prevention detection unit 2 to make an indeterminate judgment. In such cases, the judgment unit 4 makes a judgment based on the positional relationship between the worker and the workbench. Specifically, the judgment unit 4 prioritizes the results of a person taking precautions who is closer to the workbench, or a person who is at the same depth but closer to the workbench in the left-right direction. In this case, the judgment unit 4 considers a person detected as "not taking precautions" to be not involved in the work, and prioritizes the results of a person detected as "taking precautions." Information on the positional relationship between the worker and the workbench is acquired by the information acquisition unit.

[0174] When the information acquisition unit detects the presence of multiple workers in the sensor image, it determines whether the multiple workers are wearing protective equipment and further measures the distance between the multiple workers and the workbench. The judgment unit 4 makes a final judgment on the presence or absence of prevention based on the state probability from the prevention detection unit 2 and the "presence or absence of safety equipment being worn and the distance from the workbench" from the information acquisition unit.

[0175] If it is determined that multiple workers are in the same area, the workers are notified by the external device 9. This notification can prompt worker A, who is not wearing goggles, to leave the work area. If both workers are wearing goggles, the worker farthest from the work platform can be prompted to leave the area, or a notification can be sent that multiple workers are working in the same area.

[0176] As shown in FIG. 28, the safety monitoring systems 1, 10, and 100 according to embodiments 1 to 4 are realized by a computer system having a processing circuit including a processor 102 and a memory 103, a communication device 101, an input device 104, a display device 105, and an interface device (not shown).

[0177] Examples of the processor 102 include a CPU (Central Processing Unit), a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, a DSP (Digital Signal Processor), or a system LSI (Large Scale Integration). Examples of the memory 103 include non-volatile or volatile semiconductor memories such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable ROM), and EEPROM (Electrically EPROM), magnetic disks, flexible disks, optical disks, compact disks, minidisks, and DVDs (Digital Versatile Discs).

[0178] This is realized by the processor 102 executing a program for operating as the preventive detection unit 2, the work detection unit 3, the judgment unit 4, the learning unit 5, the judgment auxiliary information acquisition unit 60, and the information acquisition unit according to the first to fourth embodiments. This program is pre-stored in the memory 103. The processor 102 reads out and executes this program from the memory 103, thereby operating as the preventive detection unit 2, the work detection unit 3, the judgment unit 4, the learning unit 5, the judgment auxiliary information acquisition unit 60, and the information acquisition unit according to the first to fourth embodiments.

[0179] This program also causes a computer to execute the procedures or methods of the processes performed by the preventive detection unit 2, the work detection unit 3, the judgment unit 4, the learning unit 5, the judgment auxiliary information acquisition unit 60, and the information acquisition unit according to embodiments 1 to 4. The program may be provided by a storage medium on which the program is stored, or by other means such as a communication medium.

[0180] The memory 103 is used to store data from the first sensor 7 and the second sensor 8, the detection algorithms of the preventive detection unit 2 and the work detection unit 3, the detection results, the learned model of the learning unit 5, the judgment results of the judgment unit 4, the setting values ​​50 input to the judgment auxiliary information acquisition unit 60, judgment auxiliary information, etc. The memory 103 is also used as a temporary memory when the processor 102 executes various processes.

[0181] The input device 104 may be connected to, for example, an external microphone, a keyboard, a mouse, a keypad, or a touch panel.

[0182] The interface device (not shown) is an interface for outputting the obtained determination results to the outside.

[0183] The display device 105 is a device that displays, for example, the detection results of the preventive detection unit and the work detection unit, and the judgment result of the judgment unit. The display device 105 is, for example, an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) display. The display device can be omitted as necessary.

[0184] The communication device 101 is a device for performing wireless or wired communication with external devices such as the first sensor 7, the second sensor 8, and the external device 9.

[0185] In the above-described embodiments, the notation "... part" used for each component may be replaced with other notations such as "... circuitry," "... assembly," "... device," "... unit," or "... module."

[0186] The configurations shown in the above embodiments are merely examples, and may be combined with other known technologies, or different embodiments may be combined with each other. It is also possible to omit or modify parts of the configurations as long as they do not deviate from the gist of the invention. [Industrial Applicability]

[0187] The present disclosure is useful for safety monitoring systems. [Explanation of symbols]

[0188] 1, 10, 100 Safety Monitoring System 2, 20 Preventive detection unit 3, 30 Work detection unit 4, 40 Judgment section 5. Learning Department 7 First Sensor 8 Second Sensor 9 External device 21 Detection unit 22 Processing section 31 Detection unit 32 Processing section 50 setting value 60 Judgment auxiliary information acquisition unit 101 Communication equipment 102 processors 103 memory 104 Input Device 105 Display device

Claims

1. a prevention detection unit that detects a prevention state in the work environment, based on information from at least one first sensor, as to whether or not a safety measure is being taken; an operation detection unit that detects an operation state in the operation environment, based on information from at least one second sensor, as to whether a predetermined operation is being performed; a determination unit that determines whether work is possible in the work environment based on the prevention state and the work state; A safety monitoring device comprising:

2. The prevention detection unit includes a prevention state learning unit that performs model learning of the prevention state using data from the at least one first sensor and obtains, as a result of the model learning, a prevention state learned model that is used to detect the prevention state; and / or The task detection unit includes a task state learning unit that performs model learning of the task state using data from the at least one second sensor and obtains, as a result of the model learning, a task state learned model to be used for detecting the task state.

2. The safety monitoring device according to claim 1.

3. a judgment assistance information acquisition unit that generates judgment assistance information based on a set value, the prevention state, and the work state; the determination assistance information is priority information associated with the prevention status and / or the operation status; the prevention state and the working state each include a plurality of states, and the priority order information is information indicating which state should be given priority when the plurality of states are determined to have the same degree of probability; the determination unit determines whether the work is possible or not based on the determination auxiliary information, the prevention state, and the work state.

2. The safety monitoring device according to claim 1.

4. The safety monitoring device according to claim 3, characterized in that the priority information is a priority assigned to a first preventive state indicating that safety measures have been taken and a second preventive state indicating that safety measures have not been taken, among the plurality of preventive states.

5. the priority information is a priority assigned to a first work state indicating that a predetermined work is being performed and a second work state indicating that the predetermined work is not being performed, among the plurality of work states; 4. The safety monitoring device according to claim 3.

6. a judgment assistance information acquisition unit that generates judgment assistance information based on a set value, the prevention state, and the work state; the judgment assistance information is threshold information associated with the prevention state and / or the work state, the prevention state and the working state each include a plurality of states, and the threshold information is information on a threshold used to determine whether or not each of the plurality of states applies; the determination unit determines whether the work is possible or not based on the determination auxiliary information, the prevention state, and the work state.

2. The safety monitoring device according to claim 1.

7. The threshold value set for the state included in the preventive state is smaller than the threshold value set for the state included in the working state.

7. The safety monitoring device according to claim 6.

8. The threshold value set for the state included in the preventive state is greater than the threshold value set for the state included in the working state.

7. The safety monitoring device according to claim 6.

9. The threshold information is, among the plurality of states included in the preventive state, a first precautionary threshold set to a first precautionary state indicating that the safety measures are being taken; a second precautionary threshold set to a second precautionary state indicating that the safety measure has not been taken; Including, The first prevention threshold is less than the second prevention threshold.

7. The safety monitoring device according to claim 6.

10. The threshold information is, among the plurality of states included in the preventive state, a first precautionary threshold set to a first precautionary state indicating that the safety measures are being taken; a second precautionary threshold set to a second precautionary state indicating that the safety measure has not been taken; Including, the first prevention threshold is greater than the second prevention threshold; 7. The safety monitoring device according to claim 6.

11. The threshold information is, among the plurality of states included in the work state, a first work threshold value set to a first work state indicating that the predetermined work is being performed; and a second work threshold set to a second work state indicating that the predetermined work is not being performed; and Including, the first operating threshold is less than the second operating threshold; 7. The safety monitoring device according to claim 6.

12. The threshold information is, among the plurality of states included in the work state, a first work threshold value set to a first work state indicating that the predetermined work is being performed; and a second work threshold set to a second work state indicating that the predetermined work is not being performed; and Including, The first operating threshold is greater than the second operating threshold.

7. The safety monitoring device according to claim 6.

13. a judgment assistance information acquisition unit that generates judgment assistance information based on a set value, the prevention state, and the work state; the judgment assistance information is time-series information associated with the prevention state and / or the work state, the prevention state and the work state each include a plurality of states, and the time-series information is information on whether the work is possible or not at a plurality of points in time in the past; the determination unit determines whether the work is possible or not based on the determination auxiliary information, the prevention state, and the work state.

2. The safety monitoring device according to claim 1.

14. The possibility of the work is as follows: a first work availability status indicating that the worker is able to continue performing the work; and a second work availability status indicating that the worker is unable to continue performing the work; and a third work availability status indicating that it is unclear whether the worker can continue to perform the work; and Including, The time-series information includes the first work availability status or the second work availability status within a predetermined period.

14. The safety monitoring device according to claim 13.

15. a judgment assistance information acquisition unit that generates judgment assistance information based on a set value, the prevention state, and the work state; the judgment assistance information is time-series information associated with the prevention state and / or the work state, the prevention state and the work state each include a plurality of states, and the time-series information is information indicating what state the prevention state and / or the work state was at a plurality of points in time in the past, the determination unit determines whether the work is possible or not based on the determination auxiliary information, the prevention state, and the work state.

2. The safety monitoring device according to claim 1.

16. The preventive detection unit A state in which it is easy to mistakenly believe that the safety measures have been taken; A state in which it is easy to make a mistake if the safety measures are not taken; Detecting 16. A safety monitoring device according to any one of claims 1 to 15.

17. The work detection unit An object or state that is likely to be mistaken for the specified task being performed; An object or state that is likely to be mistaken for not having performed the specified task; Detecting 16. A safety monitoring device according to any one of claims 1 to 15.

18. An information acquisition unit that acquires information on distances between the plurality of workers and the workbench, the determination unit determines whether or not the work is possible based on the distance information and the preventive state detected for each of the plurality of workers.

2. The safety monitoring device according to claim 1.

19. Detecting a preventative condition in the work environment based on information from the at least one first sensor, whether or not a safety measure is being taken; detecting a work state in the work environment, based on information from at least one second sensor, as to whether a predetermined work is being performed; determining whether work is possible in the work environment based on the prevention state and the work state; A safety monitoring method comprising:

20. Detecting a preventative condition in the work environment based on information from the at least one first sensor, whether or not a safety measure is being taken; detecting a work state in the work environment, based on information from at least one second sensor, as to whether a predetermined work is being performed; determining whether work is possible in the work environment based on the prevention state and the work state; A program that causes a computer to execute the following.

Images