Protective equipment wearing status monitoring method and system based on artificial intelligence

Abstract

The present invention relates to the field of data processing technology, and in particular to a method and system for monitoring the wearing status of protective equipment based on artificial intelligence. The method comprises the steps of: obtaining each body region in a pre-acquired photoelectric data matrix of protective equipment, a feature vector corresponding to the body region, and the probability that each key data point in the body region belongs to a corresponding human key point category; determining the regional effectiveness of the body region; obtaining a preset sample to build a base library, obtaining the cosine similarity of the feature vector of the body region and the sample in the base library, and determining the wearing probability of the body region; calculating the wearing probability of the body region after calibration; obtaining the wearing status of the protective equipment by comparing the wearing probability of the body region after calibration with a preset threshold, so as to effectively improve the accuracy of the monitoring result obtained based on the wearing probability after calibration.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for monitoring the wearing status of protective equipment based on artificial intelligence. Background Art

[0002] Protective equipment is the first line of defense to protect individuals from harm in the work environment. It plays a vital role in ensuring the safety and health of workers. Monitoring the wearing status of protective equipment can monitor in real time whether the protective equipment is used correctly, effectively prevent safety hazards caused by improper wearing of protective equipment, strengthen the implementation of safety regulations, and ensure the compliance of work processes. Automated monitoring can effectively improve the efficiency of safety management, reduce human errors, and provide strong support for building an accident-free work environment.

[0003] In the prior art, there are many applications for monitoring the wearing status of user protective equipment. For example, the patent application document with publication number CN114625290A discloses a human-machine interface and a related interaction method, which detects the face of a person trying to interact with the HMI and also detects the gesture made by the person relative to the HMI. The information of the detected face is analyzed to determine whether the person is wearing personal protective equipment, and corresponding processing is performed based on the detection result. Among them, the processing circuit can be configured to perform one or more image analysis and / or image recognition technologies to detect whether the person is wearing personal protective equipment.

[0004] Although the above-mentioned prior art provides steps for obtaining the wearing status of protective equipment through image analysis and / or image recognition technology, it does not record how to accurately obtain the wearing status of the user's protective equipment, and it is subsequently impossible to obtain accurate interaction results based on this.

[0005] Based on this, how to accurately obtain the wearing status of protective equipment is a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0006] In order to solve the technical problem of how to accurately obtain the wearing status of protective equipment, the present invention provides a wearing status monitoring method and system for protective equipment based on artificial intelligence.

[0007] In the first aspect, the present invention provides a protective equipment wearing status monitoring method based on artificial intelligence, which adopts the following technical solutions:

[0008] A protective equipment wearing status monitoring method based on artificial intelligence includes the following steps:

[0009] Obtain each body region in the photoelectric data matrix of the protective equipment obtained in advance and the feature vector corresponding to the body region; obtain the probability that each key data point in the body region belongs to the corresponding human key point category; determine the regional effectiveness of the body region, and the regional effectiveness is positively correlated with the probability that each key data point in the body region belongs to the corresponding human key point category; obtain a preset sample to build a base library, obtain the cosine similarity of the feature vector of the body region and the samples in the base library, determine the wearing probability of the body region, and the wearing probability is positively correlated with the cosine similarity; calculate the wearing probability of the body region after calibration:

[0010] ;

[0011] In the formula, Indicates The wearing probability after calibration of body regions, Indicates The wearing probability of each body region, Indicates The wearing status of the protective equipment is obtained by comparing the wearing probability of the body area after calibration with a preset threshold.

[0012] The present invention determines the body area by acquiring key data points in the matrix, and combines the cosine similarity between the feature vectors of the body area and the base library samples, thereby effectively improving the efficiency of the monitoring results of determining the wearing status of the protective equipment; in addition, the present invention also takes into account the occlusion in the photoelectric data acquisition and the accuracy of the key data points will affect the accuracy of the obtained wearing probability, so by acquiring the regional effectiveness of the body area to calibrate the wearing probability, the accuracy of the monitoring results obtained based on the calibrated wearing probability can be effectively improved.

[0013] According to the artificial intelligence-based protective equipment wearing status monitoring method provided by the present invention, the acquisition of each body area in the pre-acquired photoelectric data matrix of the protective equipment also includes: real-time acquisition of photoelectric data by sensors, storing the photoelectric data in matrix form, and then performing denoising preprocessing to obtain the photoelectric data matrix.

[0014] The present invention takes into account that the initial data collected by the sensor is not conducive to subsequent data processing, and therefore improves the overall quality of the data through denoising preprocessing.

[0015] According to the artificial intelligence-based protective equipment wearing status monitoring method provided by the present invention, the probability of obtaining each key data point in the body area belonging to the corresponding human key point category includes: taking a historically collected matrix as a sample set, wherein the sample set contains multiple matrix samples; after marking the human key points in the matrix samples with the coordinates of the key data points, inputting the YOLOv8 model for training to obtain a trained YOLOv8 model; the trained YOLOv8 model extracts the key data points from the photoelectric data stored in the matrix form and the probability that the key data points belong to the corresponding human key point category.

[0016] According to the artificial intelligence-based protective equipment wearing status monitoring method provided by the present invention, the acquisition of each body area in the pre-acquired photoelectric data matrix of the protective equipment includes: presetting the category of human key points required to determine each of the body areas, and forming a key point group of the body area to obtain a plurality of key point groups of the body area; using the coordinates of the key data points in the SVM model to extract the corresponding regional matrix, if the key data points in the regional matrix meet the categories of all human key points in any key point group, then determining the regional matrix as the body area corresponding to the key point group.

[0017] According to the artificial intelligence-based protective equipment wearing status monitoring method provided by the present invention, the regional effectiveness of the body region satisfies the relationship:

[0018] ;

[0019] In the formula, Indicates The regional effectiveness of each body area, Indicates The number of key point groups in each body region, Indicates body region The number of human key points in a key point group, Indicates body region The first key point in the group The probability that a key data point belongs to the corresponding human key point category, Indicates the multiplication symbol.

[0020] According to the protective equipment wearing status monitoring method based on artificial intelligence provided by the present invention, the wearing probability of the body area satisfies the relationship:

[0021] ;

[0022] In the formula, Indicates The wearing probability of each body region, Indicates the number of samples in the base library, Indicates The feature vector of each body region, Indicates the first The feature vector of the samples, represents the cosine similarity function, Indicates the multiplication symbol, represents the linear normalization function.

[0023] The present invention provides a precise calculation method for the wearing probability of a body region. By calculating the cosine similarity between the feature vectors of the current body region and the body regions in the base database to characterize the degree of difference between the two, the wearing probability of the body region can be accurately obtained.

[0024] According to the artificial intelligence-based protective equipment wearing status monitoring method provided by the present invention, the wearing status of the protective equipment is obtained by comparing the wearing probability of the body area after calibration with the preset threshold, including: a preset threshold; if the wearing probability of the body area after calibration is greater than the threshold, the wearing status of the body area is obtained as worn; if the wearing probability of the body area after calibration is not greater than the threshold, the wearing status of the body area is obtained as not worn.

[0025] The present invention compares the calibrated wearing probability of the obtained body area with the threshold by means of a preset threshold, so as to accurately obtain different wearing states of the body area, and is helpful to make relevant processing based on this in the subsequent process, thereby reducing work risks.

[0026] According to the artificial intelligence-based protective equipment wearing status monitoring method provided by the present invention, the wearing status of the body area is obtained as not worn, and then it also includes: marking the body area corresponding to the not worn wearing status, and issuing a warning to the outside.

[0027] The present invention takes into account that there may be greater safety risks when workers directly work without wearing protective equipment. Therefore, the present invention sends out an early warning to remind workers to wear protective equipment in time, thereby effectively reducing safety risks.

[0028] In a second aspect, the present invention provides a protective equipment wearing status monitoring system based on artificial intelligence, which adopts the following technical solutions:

[0029] The protective equipment wearing status monitoring system based on artificial intelligence includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned protective equipment wearing status monitoring method based on artificial intelligence is implemented.

[0030] By adopting the above technical solution, the above-mentioned artificial intelligence-based protective equipment wearing status monitoring method is generated into a computer program and stored in a memory to be loaded and executed by a processor, so that a terminal device is made according to the memory and the processor for easy use.

[0031] The present invention has the following technical effects:

[0032] Based on the above technical scheme, when determining the monitoring results of the wearing status of protective equipment, the present invention determines the body area by obtaining the key data points in the matrix, and combines the cosine similarity between the feature vectors of the body area and the base library samples, thereby effectively improving the efficiency of the monitoring results of determining the wearing status of the protective equipment; in addition, the present invention also calibrates the wearing probability by obtaining the regional effectiveness of the body area, which can effectively reduce the influence of occlusion in photoelectric data acquisition and the accuracy of key data points on the accuracy of the obtained wearing probability, thereby effectively improving the accuracy of the monitoring results obtained based on the calibrated wearing probability. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.

[0034] Figure 1 A flowchart of a method for monitoring the wearing status of protective equipment based on artificial intelligence provided by an embodiment of the present invention;

[0035] Figure 2 A schematic diagram of key points of a human body provided by an embodiment of the present invention;

[0036] Figure 3 A schematic diagram of a divided body region provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0038] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0039] It should be noted that protective equipment is the first line of defense to protect individuals from injuries and health risks in the work environment. It plays a vital role in ensuring the safety and health of workers. Monitoring the wearing status of protective equipment can monitor in real time whether the equipment is used correctly, effectively prevent safety hazards caused by improper wearing of protective equipment, strengthen the implementation of safety regulations, and ensure the compliance of work processes. Automated monitoring can effectively improve the efficiency of safety management, reduce human errors, and provide strong support for building an accident-free work environment.

[0040] There are few studies on the monitoring of the wearing status of protective equipment in the prior art. Based on this, the embodiment of the present invention discloses a method for monitoring the wearing status of protective equipment based on artificial intelligence. The technical solution provided by the embodiment of the present invention can effectively improve the efficiency of determining the monitoring results of the wearing status of protective equipment by combining the cosine similarity between the feature vectors of the body region and the base library samples. In addition, the embodiment of the present invention also calibrates the wearing probability in combination with the effectiveness of the body region, and finally obtains the monitoring results of the wearing status of the protective equipment in the body region based on the calibrated wearing probability, which can effectively improve the accuracy of the obtained wearing status monitoring results. For details, please refer to Figure 1 As shown, Figure 1 A flow chart of a method for monitoring the wearing status of protective equipment based on artificial intelligence is provided in an embodiment of the present invention, and the method specifically includes the following steps.

[0041] S1: Obtain each body region and the feature vector corresponding to the body region in the photoelectric data matrix of the protective equipment obtained in advance; obtain the probability that each key data point in the body region belongs to the corresponding human key point category.

[0042] For example, in an embodiment of the present invention, the coordinates of key data points and the probability that the key data points belong to the corresponding human key point category can be obtained through the YOLOv8 model, and the corresponding region matrix is ​​extracted using the coordinates of the key data points in the SVM model to obtain each body region. Then, the region matrix corresponding to each body region is processed through the RESNET-50 model to obtain the feature vector corresponding to each body region.

[0043] The following is a detailed description:

[0044] It should be noted that YOLOv8 (You Only Look Once version 8, referred to as YOLOv8) is an advanced target detection algorithm. After the photoelectric data is input into the YOLOv8 model, the model can quickly process the matrix and identify the target object in the positioning area.

[0045] Specifically, in an embodiment of the present invention, before obtaining the coordinates of key data points through the YOLOv8 model, it also includes: in order to effectively capture all important photoelectric changes when workers are active in the work area, a photoelectric sensor system that can comprehensively and effectively detect various positions in the work area at different angles can be deployed in the worker's work area; photoelectric data is collected in real time by the sensor, and after the photoelectric data is stored in a matrix form, denoising preprocessing is performed to obtain a photoelectric data matrix.

[0046] The denoising preprocessing may be implemented by Gaussian filtering, which may be specifically configured according to actual needs, and the embodiment of the present invention does not impose too many limitations thereon.

[0047] After basic preparations have been made by collecting data through sensors, subsequent data processing can be performed based on the collected photoelectric data matrix.

[0048] It should be noted that the coordinates of the key points of the human body can be used to determine the various body regions of the human body from the photoelectric data stored in the form of a matrix. In the embodiment of the present invention, there are 17 key points of the human body, including 1 nose, 2 eyes, 2 ears, 2 shoulders, 2 elbows, 2 wrists, 2 hips, 2 knee joints and 2 ankles. Figure 2 As shown, Figure 2 A schematic diagram of a key point of a human body provided by an embodiment of the present invention, wherein each white dot represents a key point of a human body, Figure 2 As can be seen, 0 represents the nose, 1 and 2 represent the eyes, 3 and 4 represent the ears, 5 and 6 represent the shoulders, 7 and 8 represent the elbows, 9 and 10 represent the wrists, 11 and 12 represent the hips, 13 and 14 represent the knees, and 15 and 16 represent the ankles.

[0049] Before obtaining the coordinates of key data points through the YOLOv8 model, the YOLOv8 model needs to be pre-trained so that the key points can be accurately extracted later.

[0050] By way of example, in an embodiment of the present invention, when obtaining the probability that each key data point in a body region belongs to a corresponding human key point category, a historically collected matrix can be used as a sample set, wherein the sample set includes multiple matrix samples; after marking the human key points in the matrix samples with the coordinates of the key data points, the YOLOv8 model is input for training to obtain a trained YOLOv8 model; the trained YOLOv8 model extracts key data points and the probability that the key data points belong to the corresponding human key point category from the photoelectric data stored in matrix form.

[0051] Specifically, the human body key points in the historically collected matrix samples are annotated by manually using the coordinates of the key data points to obtain the annotated historically collected matrix samples. The YOLOv8 model is trained using the annotated historically collected matrix samples to obtain a trained YOLOv8 model that can be used to determine the key data points. After the currently collected matrix is ​​input into the trained YOLOv8 model, the coordinates of the key data points in the matrix and the probability that the key data points belong to the corresponding human body key point category are obtained.

[0052] Among them, the specific steps of obtaining the coordinates of the key data points in the matrix and the probability that the key data points belong to the corresponding human key point category through the YOLOv8 model can be obtained through the existing technology, and the embodiments of the present invention are not described in detail here.

[0053] It should be further explained that when the coordinates of the key data points in the matrix are obtained through the YOLOv8 model, they are determined by the probability that each identified key data point belongs to the corresponding human key point category. For example, there are two key data points collected near the human key point eye, one with a probability of 0.6 and the other with a probability of 0.5. At this time, the key data point corresponding to the larger probability of 0.6 is the current human key point eye, and the coordinates of the key data point are used as the coordinates of the eye.

[0054] However, not all human key points can be accurately matched to the corresponding key data point coordinates. During the model processing, the probability that some key data points belong to the corresponding human key point category may be low, but they will be mistaken for human key points. For example, there are two key data points collected near the human key point eyes, one with a probability of 0.1 and the other with a probability of 0.2. At this time, the key data point corresponding to the larger probability 0.2 is the current human key point eye. Obviously, the coordinates of the key data point may not be the true coordinates of the human key point eye, and the subsequent monitoring results obtained by data processing based on this are less accurate. Therefore, when obtaining the coordinates of the key data points through the YOLOv8 model and obtaining the human body area, it is necessary not only to obtain the coordinates of the key data points, but also to obtain the probability that the key data points belong to the corresponding human key point category, so as to facilitate subsequent processing.

[0055] After obtaining the coordinates of the key data points and the probability that the key data points belong to the corresponding human key point category based on the above steps, each body region can be obtained through the coordinates of the key data points.

[0056] By way of example, in an embodiment of the present invention, when obtaining each body region in a pre-acquired photoelectric data matrix of protective equipment, the categories of human key points required to determine each body region can be preset, and key point groups of the body regions can be formed to obtain multiple key point groups of the body regions; the coordinates of the key data points are used in the SVM model to extract the corresponding regional matrix, and if the key data points in the regional matrix meet the categories of all human key points in any key point group, the regional matrix is ​​determined to be the body region corresponding to the key point group.

[0057] It should be noted that when monitoring the wearing status of the user's protective equipment, it is necessary to set the body part to be monitored according to the actual situation. In the embodiment of the present invention, the monitoring object can be a construction worker on the construction site. For details, see Figure 3 As shown, Figure 3 A schematic diagram of a body region obtained by division provided by an embodiment of the present invention, combined with Figure 3 It can be seen that after extracting the corresponding regional matrix using the coordinates of key data points in the SVM (Support Vector Machine, SVM for short) model, the body regions obtained are the head, torso, hands and feet. The targets that need to be monitored for the head are the safety helmets and masks on the workers' heads, the targets that need to be monitored for the torso are the safety vests, the targets that need to be monitored for the hands are the safety gloves, and the targets that need to be monitored for the feet are the safety shoes.

[0058] When using the coordinates of key data points to extract the corresponding region matrix, there may be a combination of multiple human key point categories that can realize the division of the region matrix. For example: when dividing the region matrix of the human head, it can be divided by the nose and eyes, by the ears, or by the shoulders, that is, the nose, left eye and right eye are a key point group of the head region matrix, the left ear and right ear are another key point group of the head region matrix, and the left shoulder and right shoulder are another key point group of the head region matrix. When determining the facial region matrix, if the key data point coordinates of all human key points in any key point group are detected at the same time, the division of the facial region matrix can be realized.

[0059] Among them, the key points of the human body for dividing the body areas can be set according to actual needs, and the embodiments of the present invention will not be described in detail here.

[0060] In order to accurately divide each body region through the SVM model, the SVM model needs to be trained before using the coordinates of key data points in the SVM model to extract the corresponding region matrix to obtain each body region.

[0061] Specifically, in an embodiment of the present invention, when training an SVM model, the matrix collected historically can be manually divided to obtain the regional matrices representing the body regions, and each regional matrix can be labeled with the corresponding key point group and the coordinates corresponding to the key point group to obtain a labeled regional matrix. The labeled regional matrix is ​​used to complete the regional matrix division training of the SVM model to obtain a trained SVM model. The coordinates of the currently extracted key data points are input into the trained SVM model to obtain the boundaries of each body region, and the photoelectric data corresponding to each boundary is obtained from the matrix to obtain the regional matrix of the body region.

[0062] The step of obtaining each body region by dividing the body region through the SVM model can be implemented by the existing technology, and the embodiment of the present invention will not be described in detail here.

[0063] In this way, the embodiment of the present invention can realize the division of the regional matrix through the coordinates of the key data points. When the coordinates of the key data points of the new matrix are input later, the division of the regional matrix corresponding to each body area can be accurately realized without retraining the model.

[0064] After extracting the coordinates of the key data points, the probability that the key data points belong to the corresponding human key point category, and the region matrix corresponding to each body region based on the above steps, continue to perform the following steps.

[0065] It should be noted that after obtaining the regional matrix corresponding to each body region based on the above steps, the feature vector corresponding to each body region can be obtained by processing the regional matrix corresponding to each body region. RESNET-50 (Residual Network with 50 layers, RESNET-50 for short) is a model with powerful feature extraction and classification capabilities. By processing the regional matrix corresponding to the body region with the RESNET-50 model, the feature vector corresponding to the body region can be accurately obtained.

[0066] Before obtaining the feature vectors corresponding to the body regions through the RESNET-50 model, the RESNET-50 model needs to be pre-trained.

[0067] Specifically, each region matrix corresponding to the historically collected body regions is manually labeled, wherein the region matrix of the historically collected wearable device is labeled as 1, and the region matrix of the historically collected non-wearable device is labeled as 0, to obtain the region matrix of the historically collected labeled area. The RESNET-50 model is trained using the region matrix of the historically collected labeled area to obtain the trained RESNET-50 model. The currently collected region matrix is ​​input into the trained RESNET-50 model, and the feature vector of the region matrix is ​​extracted in the convolution layer and the custom layer of the trained RESNET-50 model, which is the feature vector corresponding to the current body region.

[0068] Among them, the specific steps of training the RESNET-50 model and obtaining the feature vector corresponding to the body area through the RESNET-50 model can be achieved through existing technologies, and the embodiments of the present invention will not be described in detail here.

[0069] S2: Determine the regional effectiveness of the body region.

[0070] It should be noted that, based on the above steps, the YOLOv8 model can obtain key data points, but not all key data points are human key points, and the accuracy of the wearing probability of the body area obtained based on this is low. Based on this, it is necessary to analyze the accuracy of the body area obtained based on the key data points. The higher the probability that the key data point belongs to the corresponding human key point category, the higher the accuracy of the obtained body area, and the higher the corresponding regional effectiveness; the regional effectiveness of the body area is positively correlated with the probability that each key data point in the body area belongs to the corresponding human key point category.

[0071] For example, in the embodiment of the present invention, the regional effectiveness of the body region is determined, and the specific relationship can be as follows:

[0072] ;

[0073] In the formula, Indicates The regional effectiveness of each body area, Indicates The number of key point groups in each body region, Indicates body region The number of human key points in a key point group, Indicates body region The first key point in the group The probability that a key data point belongs to the corresponding human key point category, Indicates the multiplication symbol.

[0074] In the above formula, Indicates body region Key point group The contribution of the regional effectiveness of each body region is The higher the probability that the key data points in the key point group belong to the corresponding human key point category, the The greater the regional effectiveness, the greater the area of ​​the body.

[0075] Take the above formula for example: a key point group in the head and body region is nose, left eye and right eye, and among the key data points corresponding to nose, left eye and right eye, there is a key data point with a higher probability of being in the nose category, then the regional effectiveness of the head and body region is also higher.

[0076] After obtaining the regional effectiveness and wearing probability of the body region based on the above steps, the wearing probability obtained based on the body region can be calibrated based on the regional effectiveness of the body region, that is, continue to perform the following steps.

[0077] S3: Obtain preset samples to construct a base library, obtain the cosine similarity of the feature vectors of the body region and the samples in the base library, and determine the wearing probability of the body region.

[0078] It should be noted that in actual applications, there may be occlusion or data missing during the photoelectric data collection process, which may result in the identification of human body areas as non-human areas, or the identification of normally worn body areas as areas with low wearing probability, etc. The accuracy of the wearing probability of the obtained body area is low, and it is impossible to accurately obtain the monitoring results of the wearing status based on this. The regional effectiveness of the body area can characterize the accuracy of the obtained body area. The higher the regional effectiveness of the body area, the higher the possibility that the regional matrix extracted by the key data points corresponding to the body area is a human body area. Therefore, the regional effectiveness based on the body area can calibrate the wearing probability obtained based on the body area.

[0079] Based on this, the embodiment of the present invention presets a base database, regularly adds a whitelist of normal wear of body regions to the base database, and compares the similarity between the identified body region and the whitelist in the base database. The higher the similarity between the current body region and the whitelist in the base database, the higher the wearing probability of the body region. The wearing probability of the body region is obtained through the similarity between the body region and the whitelist in the base database, and the wearing probability is positively correlated with the cosine similarity of the feature vectors of the body region and the samples in the base database.

[0080] For example, in the embodiment of the present invention, the area matrix of normal wear in each body area can be used as a sample to construct a base library.

[0081] For example, in the embodiment of the present invention, a wearing probability threshold may be preset, and the body region with a wearing probability less than the wearing probability threshold may be compared with the whitelist in the base database for similarity to obtain the wearing probability of the body region.

[0082] Among them, the wearing probability threshold can be set to 0.6, which can be set according to actual needs.

[0083] In this way, by screening body areas based on preset wearing probability thresholds, the amount of data processing can be effectively reduced, thereby improving data processing efficiency.

[0084] For example, in the embodiment of the present invention, the wearing probability of a body region is determined, and the specific relationship can be referred to as follows:

[0085] ;

[0086] In the formula, Indicates The wearing probability of each body region, Indicates the number of samples in the base library, Indicates The feature vector of each body region, Indicates the first The feature vector of the samples, represents the cosine similarity function, Indicates the multiplication symbol, represents the linear normalization function.

[0087] In the above formula, Indicates The cumulative sum of the differences between the feature vector of the body region and the feature vectors of all samples in the base library. The larger the value, the The higher the probability that the wearing state of the body area is not worn, the smaller the value is, the higher the probability that the wearing state of the body area is not worn. The higher the probability that the wearing state of each body area is worn.

[0088] After respectively obtaining the regional effectiveness and the wearing probability of the body region based on the above steps, the wearing probability can be calibrated based on the regional effectiveness of the body region, so as to accurately obtain the calibrated wearing probability.

[0089] S4: Calculate the wearing probability after body region calibration, and obtain the wearing status of the protective equipment by comparing the wearing probability after body region calibration with a preset threshold.

[0090] It should be noted that after obtaining the regional effectiveness and wearing probability of the body region respectively based on the above steps, the wearing probability of the body region can be calibrated based on the regional effectiveness of the body region, so as to accurately obtain the calibrated wearing probability of the body region.

[0091] For example, in the embodiment of the present invention, the wearing probability after body region calibration is determined, and specifically, the following relationship can be referred to:

[0092] ;

[0093] In the formula, Indicates The wearing probability after calibration of body regions, Indicates The wearing probability of each body region, Indicates The regional effectiveness of each body area.

[0094] In the above formula, represents the wearing probability after calibration according to the regional effectiveness, The higher the regional effectiveness of each body region, the higher the credibility of the wearing probability of the body region obtained based on it. When calculating the calibrated wearing probability, The wearing probability of the first body area is given a higher weight; The lower the regional validity of the body region, the lower the credibility of the wearing probability of the body region obtained based on it, and it is impossible to determine whether it is a real body region. When calculating the wearing probability after calibration, The wearing probability of the first body region is directly set to 0.5. The higher the wearing probability of each body area, the The more similar the feature vectors of the body regions are to the body regions in the base database, the closer the wearing probability of the obtained body regions after calibration is to the wearing probability 1.

[0095] After obtaining the wearing probability of the body region after calibration based on the above steps, the wearing status of the protective equipment can be obtained based on the wearing probability of the body region after calibration.

[0096] By way of example, in an embodiment of the present invention, the wearing status of the protective equipment is obtained by comparing the wearing probability of the body area after calibration with a preset threshold, including: a preset threshold; if the wearing probability of the body area after calibration is greater than the threshold, the wearing status of the body area is obtained as worn; if the wearing probability of the body area after calibration is not greater than the threshold, the wearing status of the body area is obtained as not worn.

[0097] The threshold value may be set according to actual needs, and the embodiment of the present invention does not impose too many restrictions on this.

[0098] In order to ensure the safety of work, when the wearing state of the body area is obtained as not wearing, the not wearing state can be processed in time.

[0099] For example, in an embodiment of the present invention, the wearing state of the body region is obtained as not worn, and then the method further includes: marking the body region corresponding to the not worn wearing state and issuing an external warning.

[0100] Specifically, the wearing status of the body regions belonging to the same human body is counted, and different levels of warnings are issued according to the number of wearing status of the body regions of the same human body that are not worn.

[0101] Among them, the warning method can be a sound warning, a light warning, etc., which can be set according to actual needs, and the embodiment of the present invention does not impose too many restrictions here.

[0102] It can be seen that in the embodiment of the present invention, when determining the monitoring result of the wearing status of the protective equipment, each body area in the pre-acquired photoelectric data matrix of the protective equipment, the characteristic vector corresponding to the body area and the probability that each key data point in the body area belongs to the corresponding human key point category can be obtained; the regional effectiveness of the body area is determined, and the regional effectiveness is positively correlated with the probability that each key data point in the body area belongs to the corresponding human key point category; the preset samples are obtained to build a base library, the cosine similarity of the characteristic vectors of the body area and the samples in the base library is obtained, and the wearing probability of the body area is determined, and the wearing probability is positively correlated with the cosine similarity; the wearing probability of the body area after calibration is calculated; and the wearing status of the protective equipment is obtained by comparing the wearing probability of the body area after calibration with the preset threshold.

[0103] In this way, the embodiment of the present invention determines the body area by obtaining the key data points in the matrix, and combines the cosine similarity between the feature vectors of the body area and the base library samples, so as to effectively improve the efficiency of the monitoring results of determining the wearing status of the protective equipment; in addition, the embodiment of the present invention also takes into account the occlusion in the photoelectric data acquisition and the accuracy of the key data points. The impact will have on the accuracy of the obtained wearing probability. Therefore, by obtaining the regional effectiveness of the body area to calibrate the wearing probability, the accuracy of the monitoring results obtained based on the calibrated wearing probability can be effectively improved.

[0104] An embodiment of the present invention also discloses an artificial intelligence-based protective equipment wearing status monitoring system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an artificial intelligence-based protective equipment wearing status monitoring method according to the present invention is implemented.

[0105] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0106] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM, a dynamic random access memory DRAM, a static random access memory SRAM, an enhanced dynamic random access memory EDRAM, a high bandwidth memory HBM, a hybrid memory cube HMC, etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device.

[0107] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0108] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A protective equipment wearing status monitoring method based on artificial intelligence, characterized in that: include: Acquire each body region in the pre-acquired optoelectronic data matrix of the protective equipment and a feature vector corresponding to the body region; Obtaining the probability that each key data point in the body region belongs to the corresponding human key point category; The step of obtaining each body region in the pre-acquired optoelectronic data matrix of the protective equipment also includes: The photoelectric data is collected in real time by a sensor, and after the photoelectric data is stored in a matrix form, a denoising preprocessing is performed to obtain a photoelectric data matrix; The obtaining of the probability that each key data point in the body region belongs to the corresponding human key point category includes: Using the historically collected matrix as a sample set, wherein the sample set includes a plurality of matrix samples; After marking the human key points in the matrix sample with the coordinates of the key data points, the YOLOv8 model is input for training to obtain a trained YOLOv8 model; the trained YOLOv8 model extracts the key data points and the probability that the key data points belong to the corresponding human key point category from the photoelectric data stored in the matrix form; The step of obtaining each body region in the photoelectric data matrix of the protective equipment obtained in advance comprises: Presetting and determining the categories of human key points required for each of the body regions, and forming key point groups for the body regions, so as to obtain a plurality of key point groups for the body regions; Extracting a corresponding region matrix using the coordinates of the key data points in the SVM model, and determining that the region matrix is ​​a body region corresponding to the key point group if the key data points in the region matrix satisfy the categories of all human key points in any key point group; Determine the regional effectiveness of the body region, satisfying the relationship: ； In the formula, Indicates The regional effectiveness of each body area, Indicates The number of key point groups in each body region, Indicates body region The number of human key points in a key point group, Indicates body region The first key point in the group The probability that a key data point belongs to the corresponding human key point category, Indicates the symbol for continuous multiplication; Obtain preset samples to construct a base library, obtain the cosine similarity of the feature vectors of the body region and the samples in the base library, determine the wearing probability of the body region, and satisfy the relationship: ； In the formula, Indicates The wearing probability of each body region, Indicates the number of samples in the base library, Indicates The feature vector of each body region, Indicates the first The feature vector of the samples, represents the cosine similarity function, represents the linear normalization function; Calculate the calibrated wearing probability of the body region: ； In the formula, Indicates The wearing probability of each body area after calibration is calculated; and the wearing status of the protective equipment is obtained by comparing the wearing probability of the body area after calibration with a preset threshold.

2. The method for monitoring the wearing status of protective equipment based on artificial intelligence according to claim 1 is characterized in that: The comparing result of the wearing probability after the calibration of the body region with the preset threshold value to obtain the wearing state of the protective equipment includes: Preset threshold value; If the wearing probability of the body region after calibration is greater than the threshold, obtaining the wearing state of the body region as wearing; If the wearing probability of the body region after calibration is not greater than the threshold, the wearing state of the body region is obtained as not worn.

3. The method for monitoring the wearing status of protective equipment based on artificial intelligence according to claim 2 is characterized in that: The step of obtaining that the wearing state of the body region is not worn further comprises: After marking the body area corresponding to the not-worn wearing state, an early warning is issued.

4. The protective equipment wearing status monitoring system based on artificial intelligence is characterized by: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for monitoring the wearing status of protective equipment based on artificial intelligence according to any one of claims 1 to 3 is implemented.

Images