A double prevention system for enterprise safety production

Abstract

The present invention discloses a dual prevention system for enterprise safety production, relates to the technical field of dual prevention, and comprises a control center, wherein the control center is connected with a database, a standard behavior model generation module, a real-time behavior data acquisition module, a real-time behavior data processing module and a risk processing module; the database is used to store risk management information; the standard behavior model generation module is used to generate standard behavior data corresponding to employees and equipment respectively according to the risk management information, and generate a standard behavior model according to the standard behavior data; the real-time behavior data acquisition module is used to obtain the real-time behavior data of employees and equipment respectively in real time; the real-time behavior data processing module is used to determine whether there is abnormal behavior in the real-time behavior data through the standard behavior model, and if there is abnormal behavior, generate corresponding risk warning according to the abnormal behavior; the risk processing module is used to generate corresponding risk resolution strategy according to the risk warning.

Description

Technical Field

[0001] The invention relates to the technical field of double prevention, in particular to a double prevention system for enterprise safety production. Background Art

[0002] The dual prevention mechanism is "safety risk classification control" and "hidden danger investigation and management". The first level of "prevention" is to control risks well and prevent hidden dangers from appearing in risk control measures. The second level of "prevention" is to promptly discover and manage hidden dangers in risk control measures to prevent accidents from happening.

[0003] In modern enterprise safety production, the construction of a double prevention system is an important means to improve the level of safety management. How to improve the efficiency and accuracy of anomaly detection in safety production through data collection and data analysis, so as to protect the lives of employees and the property of the enterprise, is an issue we need to consider at present. Summary of the invention

[0004] In order to solve the above problems, the object of the present invention is to provide a double prevention system for enterprise safety production.

[0005] The object of the present invention can be achieved by the following technical solutions: A dual prevention system for enterprise safety production, comprising a control center, wherein the control center is communicatively connected with a database, a standard behavior model generation module, a real-time behavior data acquisition module, a real-time behavior data processing module and a risk processing module;

[0006] The database is used to store employee information, equipment information and risk management information, and the risk management information includes risk management behavior information and risk management behavior level;

[0007] The standard behavior model generation module is used to generate standard behavior data corresponding to employees and equipment respectively according to risk management and control information, and generate a standard behavior model according to the standard behavior data;

[0008] The real-time behavior data collection module is used to obtain the real-time behavior data of employees and equipment respectively in real time;

[0009] The real-time behavior data processing module is used to determine whether there is abnormal behavior in the real-time behavior data through a standard behavior model. If there is no abnormal behavior, no operation is required. If there is abnormal behavior, a corresponding risk warning is generated according to the abnormal behavior.

[0010] The risk processing module is used to generate corresponding risk resolution strategies according to risk warnings;

[0011] Further, the database includes employee units, equipment units and risk units;

[0012] The employee unit is used to store employee information, which includes employee number, employee name, ID number, contact information and operation proficiency value; wherein the employee number is generated according to the ID number, and the employee number is unique;

[0013] The device unit is used to store device information, and the device information includes device number, device name, device model, user number and associated number; wherein the device number is unique, the user number is the employee number corresponding to the employee who uses the device, and the associated number is the device number corresponding to the device that has an associated relationship with the device;

[0014] The risk unit is used to store risk control information, which includes risk control behavior information and risk control behavior level; wherein the risk control behavior information is text information describing the risk control behavior, and the risk control behavior level is the risk level corresponding to the risk control behavior information; the risk level includes white risk, blue risk, yellow risk, orange risk and red risk, and the risk level is ranked as follows: white risk Blue Risk Yellow Risk Orange Risk Red risk, that is, white risk is the lowest risk level, and red risk is the highest risk level.

[0015] Furthermore, the process of generating standard behavior data corresponding to employees and equipment respectively according to risk control information includes:

[0016] Setting a bag-of-words dataset, wherein the bag-of-words dataset includes an object bag-of-words, a behavior bag-of-words, an intention bag-of-words, and a region bag-of-words;

[0017] The object word bag is composed of a number of object keywords, which are the subject and object in the risk control behavior text information, and the subject and object are both nouns;

[0018] The behavior word bag is composed of a number of behavior keywords, which are predicates in the risk management behavior text information;

[0019] The intention word bag is composed of a number of intention keywords, which are the willing verbs in the risk control behavior text information;

[0020] The regional word bag is composed of a number of regional keywords, which are place adverbials in the risk control behavior text information;

[0021] Generate standard behavior data corresponding to employees and equipment based on the bag-of-words dataset and risk management behavior information;

[0022] The standard behavior data includes subject data, object data and standard data;

[0023] If standard behavior data corresponding to an employee is generated according to risk control information, the subject data of the standard behavior data is the employee, and object keywords in the risk control behavior information included in the risk control information are filtered out according to the object word bag, and object data is generated according to the object keywords. Behavior keywords, intention keywords and regional keywords in the risk control behavior information in the risk control information are filtered out according to the behavior word bag, intention word bag and regional word bag, and standard data is generated according to the filtered behavior keywords, intention keywords and regional keywords;

[0024] If the standard behavior data corresponding to the device is generated according to the risk control information, the subject data of the standard behavior data is the device, and the object keywords in the risk control behavior information included in the risk control information are filtered out according to the object word bag, and the object data is generated according to the object keywords, and the behavior keywords, intention keywords and regional keywords in the risk control behavior information in the risk control information are filtered out according to the behavior word bag, intention word bag and regional word bag respectively, and the standard data is generated according to the filtered behavior keywords, intention keywords and regional keywords;

[0025] Filter out object keywords in the risk control behavior information according to the object word bag, and generate object data according to the remaining object keywords except the object keywords corresponding to the subject data in the object keywords; if there is only one object keyword in the object keywords filtered out according to the object word bag and the object keyword is consistent with the object keyword corresponding to the subject data, the object data is NULL;

[0026] Acquire image data corresponding to the object data, generate a corresponding overall model according to the image data, and generate an object recognition model according to the overall model corresponding to the object data, employee information, and equipment information through a convolutional neural network;

[0027] If the risk control behavior level corresponding to the risk control behavior information is white risk, the standard behavior data generated according to the risk control behavior information and the bag-of-words dataset is marked as standard regular behavior data;

[0028] If the risk control behavior level corresponding to the risk control behavior information is not white risk, the standard behavior data generated according to the risk control behavior information and the bag-of-words dataset will be marked as standard abnormal behavior data.

[0029] Furthermore, the process of generating a standard behavior model according to the standard behavior data includes:

[0030] The standard behavior model includes a standard normal behavior sub-model and a standard abnormal behavior sub-model;

[0031] Generate a standard normal behavior sub-model based on standard normal behavior data through deep learning, and generate a standard abnormal behavior sub-model based on standard abnormal behavior data through deep learning;

[0032] The standard behavior model input data is subject number, valid object data and real-time video data, and output data is generated according to the subject number, valid object data and real-time video data, and the output data includes 0 and 1;

[0033] The subject number is the number corresponding to the subject data. If the subject data is an employee, the number is the employee number; if the subject data is a device, the number is the device number.

[0034] Furthermore, the process of generating output data according to the subject number, valid object data and real-time video data is as follows:

[0035] The standard conventional behavior sub-model input data is subject number, valid object data and real-time video data, and generates output data according to the input data, wherein the output data includes T and F;

[0036] The input data of the standard abnormal behavior sub-model is the subject number, valid object data and real-time video data, and output data is generated according to the input data, and the output data includes T and F;

[0037] When the standard behavior model receives the subject number, valid object data and real-time video data, it sends the subject number, valid object data and real-time video data to the standard normal behavior sub-model and the standard abnormal behavior sub-model respectively;

[0038] Obtain output data of a standard normal behavior sub-model and a standard abnormal behavior sub-model;

[0039] If the output data of the standard normal behavior sub-model is T, and the output data of the standard abnormal behavior sub-model is F, then the output data of the standard behavior model is 0;

[0040] If the output data of the standard normal behavior sub-model is F, and the output data of the standard abnormal behavior sub-model is T, then the output data of the standard behavior model is 1;

[0041] Otherwise, the subject number will be marked as a suspicious number, and a third-level warning information will be generated according to the suspicious number. The third-level warning information will be sent to the risk processing module, and the risk processing module will make corresponding processing.

[0042] Furthermore, the process of obtaining real-time behavior data of employees and equipment includes:

[0043] The enterprise is divided into several monitoring areas, each of which is provided with a camera and an infrared thermal imager; the camera is used to obtain ordinary video of the monitoring area, and the infrared thermal imager is used to obtain thermal video of the monitoring area;

[0044] Set the monitoring cycle and regularly obtain normal video and thermal video according to the monitoring cycle;

[0045] It should be noted that there is a one-to-one correspondence between ordinary videos and thermal imaging videos;

[0046] Acquire image data of each employee and each device, generate a corresponding overall model based on the image data, and generate a subject recognition model based on the overall model, employee information, and device information corresponding to each employee and each device through a convolutional neural network;

[0047] The subject recognition model is used to identify objects including employees and equipment;

[0048] Detecting whether there is an object in the ordinary video by using the subject recognition model, when the subject recognition model detects that there is an object in the ordinary video, marking the object as a target object, obtaining a video clip in which the target object exists in the ordinary video, and marking the video clip as a first-level surveillance video;

[0049] Get the object number of the target object; if the target object is an employee, the object number is the employee number; if the target object is a device, the object number is the device number;

[0050] Perform object enhancement processing on the primary surveillance video to generate real-time video data;

[0051] Generate real-time behavioral data based on object ID and real-time video data.

[0052] Furthermore, the process of object strengthening is as follows:

[0053] Divide the first-level surveillance video into a plurality of images in units of frames, and mark the images as first-level surveillance images; obtain a thermal imaging video corresponding to the first-level surveillance video, and divide the thermal imaging video into a plurality of images in units of frames, and mark the images as first-level thermal imaging images;

[0054] It should be noted that there is a one-to-one correspondence between the first-level monitoring image and the first-level thermal image;

[0055] Obtain an overall model corresponding to the target object, and determine whether the target object in the first-level monitoring image is obscured based on the overall model. If not, no operation is required. Otherwise, obtain a first-level thermal image corresponding to the first-level monitoring image, and determine whether the target object in the first-level thermal image is obscured based on the overall model. If not, use the first-level thermal image to replace the first-level monitoring image. Otherwise, generate a first-level warning information based on the object number, and send the first-level warning information to the risk processing module.

[0056] Furthermore, the standard behavior model is used to determine whether there is abnormal behavior in the real-time behavior data. If there is no abnormal behavior, no operation is required. If there is abnormal behavior, the process of generating a corresponding risk warning according to the abnormal behavior includes:

[0057] Obtaining an object number in the real-time behavior data, and marking the object number as a subject number;

[0058] Acquire real-time video data, divide the real-time video data into a plurality of images in units of frames, and mark the images as images to be detected;

[0059] Acquire a subject object in the image to be detected by using a subject recognition model; the subject object is an object whose object number is consistent with the subject number;

[0060] Obtaining pixel points corresponding to the main object, and marking the pixel points as main pixel points;

[0061] Acquire object data corresponding to the subject data when the subject object is subject data, and acquire a quasi-object object according to the object data through an object recognition model;

[0062] Acquire pixel points corresponding to the quasi-object object, and mark the pixel points as quasi-object pixel points;

[0063] Establish a two-dimensional coordinate system and map the image to be detected into the two-dimensional coordinate system so that each pixel has its corresponding coordinates;

[0064] Obtain the horizontal difference value and the vertical difference value of the main object, and obtain the main center coordinates of the main object according to the horizontal difference value and the vertical difference value of the main object;

[0065] The horizontal difference is the difference between the maximum horizontal coordinate value and the minimum horizontal coordinate value in the corresponding coordinates of the main pixel point; the vertical difference is the difference between the maximum vertical coordinate value and the minimum vertical coordinate value in the corresponding coordinates of the main pixel point; the main center coordinate = (horizontal difference, vertical difference);

[0066] According to the principle of obtaining the corresponding principal center coordinates of the subject object, the corresponding guest center coordinates of the quasi-object object are obtained;

[0067] Set the correlation threshold;

[0068] Generate a distance vector according to the main center coordinates and the guest center coordinates, obtain the modulus of the distance vector, and if the modulus is less than the associated threshold, mark the distance vector as a valid vector, otherwise, mark the distance vector as an invalid vector;

[0069] Mark the quasi-object object corresponding to the valid vector as a valid object object;

[0070] Acquire object data corresponding to a valid object object, and mark the object data as valid object data;

[0071] When the subject object is an employee and the valid object data is a device, or the subject object is a device and the valid object data is an employee, the user number corresponding to the device and the employee number corresponding to the employee are obtained;

[0072] If the user number corresponding to the device and the employee number corresponding to the employee are consistent, no operation is required; otherwise, the subject number is marked as a mismatch number, and a level 4 warning information is generated according to the mismatch number, and the level 4 warning information is submitted to the risk processing module;

[0073] The subject number, valid object data and real-time video data in the real-time behavior data are input into the standard behavior model. If the output data of the standard behavior model is 0, there is no abnormal behavior in the real-time behavior data and no operation is required. If the output data is 1, there is abnormal behavior in the real-time behavior data, and the secondary warning information is generated according to the subject number and sent to the risk processing module.

[0074] Furthermore, the process of generating corresponding risk resolution strategies based on risk warnings includes:

[0075] When a first-level warning message is received, the object number is sent to the user, informing the user that the behavior of the object cannot be accurately monitored, and the user is asked to take corresponding measures;

[0076] When receiving the second-level warning information, if the subject number is an employee number, the employee number is sent to the user to inform the user that the employee has abnormal behavior and the user takes corresponding measures. If the subject number is a device number, the device number is sent to the user to inform the user that the device has abnormal behavior and the user takes corresponding measures. The associated number corresponding to the device number is obtained, and the devices corresponding to the device number and the associated number are powered off;

[0077] When receiving a level 3 warning message, the suspicious number will be sent to the user, and the user will be informed that the behavior of the object corresponding to the suspicious number cannot be identified, and the user will take corresponding measures;

[0078] When a level 4 warning message is received, the mismatch number is sent to the user, informing the user that the object corresponding to the mismatch number is mismatched, and the user takes corresponding action.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] 1. The present invention obtains real-time video data of employees and equipment respectively, performs abnormality judgment on the real-time video data, realizes abnormality judgment from both employees and equipment perspectives, and improves the abnormality detection effect;

[0081] 2. The present invention obtains ordinary surveillance videos through cameras, identifies ordinary surveillance videos, obtains video clips with surveillance targets, and performs object enhancement processing on the video clips, that is, detects whether the surveillance objects in the video clips are blocked. If blocked, the corresponding infrared images are obtained to replace the blocked images in the video clips, thereby improving the behavior recognition capability of the surveillance targets through infrared images.

[0082] 3. The present invention obtains standard behavior data according to risk management behavior information, marks the monitoring target as a subject object, obtains the object data corresponding to the subject object as the subject data of the standard behavior data, obtains the quasi-object object according to the object data, sets an association threshold, obtains the distance vector modulus between the subject object and the quasi-object object, and when the modulus is less than the association threshold, marks the quasi-object object as a valid object object, identifies whether there is abnormal behavior between the subject object and the valid object object, reduces irrelevant identification objects, and greatly improves the efficiency of abnormal identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0084] like Figure 1 As shown, a dual prevention system for enterprise safety production includes a control center, wherein the control center is connected to a database, a standard behavior model generation module, a real-time behavior data acquisition module, a real-time behavior data processing module and a risk processing module;

[0085] The database is used to store employee information, equipment information and risk management information, and the risk management information includes risk management behavior information and risk management behavior level;

[0086] The standard behavior model generation module is used to generate standard behavior data corresponding to employees and equipment respectively according to risk management and control information, and generate a standard behavior model according to the standard behavior data;

[0087] The real-time behavior data collection module is used to obtain the real-time behavior data of employees and equipment respectively in real time;

[0088] The real-time behavior data processing module is used to determine whether there is abnormal behavior in the real-time behavior data through a standard behavior model. If there is no abnormal behavior, no operation is required. If there is abnormal behavior, a corresponding risk warning is generated according to the abnormal behavior.

[0089] The risk processing module is used to generate corresponding risk resolution strategies according to risk warnings;

[0090] It should be further explained that, in the specific implementation process, the database includes employee units, equipment units and risk units;

[0091] The employee unit is used to store employee information, which includes employee number, employee name, ID number, contact information and operation proficiency value; wherein the employee number is generated according to the ID number, and the employee number is unique;

[0092] The device unit is used to store device information, and the device information includes device number, device name, device model, user number and associated number; wherein the device number is unique, the user number is the employee number corresponding to the employee who uses the device, and the associated number is the device number corresponding to the device that has an associated relationship with the device;

[0093] The risk unit is used to store risk control information, which includes risk control behavior information and risk control behavior level; wherein the risk control behavior information is text information describing the risk control behavior, and the risk control behavior level is the risk level corresponding to the risk control behavior information; the risk level includes white risk, blue risk, yellow risk, orange risk and red risk, and the risk level is ranked as follows: white risk Blue Risk Yellow Risk Orange Risk Red risk, that is, white risk is the lowest risk level, and red risk is the highest risk level;

[0094] It should be further explained that, in the specific implementation process, the process of generating standard behavior data corresponding to employees and equipment based on risk management information includes:

[0095] Setting a bag-of-words dataset, wherein the bag-of-words dataset includes an object bag-of-words, a behavior bag-of-words, an intention bag-of-words, and a region bag-of-words;

[0096] The object word bag is composed of a number of object keywords, which are the subject and object in the risk control behavior text information, and the subject and object are both nouns;

[0097] The behavior word bag is composed of a number of behavior keywords, which are predicates in the risk management behavior text information;

[0098] The intention word bag is composed of a number of intention keywords, which are the willing verbs in the risk control behavior text information;

[0099] The regional word bag is composed of a number of regional keywords, which are place adverbials in the risk control behavior text information;

[0100] Generate standard behavior data corresponding to employees and equipment based on the bag-of-words dataset and risk management behavior information;

[0101] The standard behavior data includes subject data, object data and standard data;

[0102] If standard behavior data corresponding to an employee is generated according to risk control information, the subject data of the standard behavior data is the employee, and object keywords in the risk control behavior information included in the risk control information are filtered out according to the object word bag, and object data is generated according to the object keywords. Behavior keywords, intention keywords and regional keywords in the risk control behavior information in the risk control information are filtered out according to the behavior word bag, intention word bag and regional word bag, and standard data is generated according to the filtered behavior keywords, intention keywords and regional keywords;

[0103] If the standard behavior data corresponding to the device is generated according to the risk control information, the subject data of the standard behavior data is the device, and the object keywords in the risk control behavior information included in the risk control information are filtered out according to the object word bag, and the object data is generated according to the object keywords, and the behavior keywords, intention keywords and regional keywords in the risk control behavior information in the risk control information are filtered out according to the behavior word bag, intention word bag and regional word bag respectively, and the standard data is generated according to the filtered behavior keywords, intention keywords and regional keywords;

[0104] Filter out object keywords in the risk control behavior information according to the object word bag, and generate object data according to the remaining object keywords except the object keywords corresponding to the subject data in the object keywords; if there is only one object keyword in the object keywords filtered out according to the object word bag and the object keyword is consistent with the object keyword corresponding to the subject data, the object data is NULL;

[0105] Acquire image data corresponding to the object data, generate a corresponding overall model according to the image data, and generate an object recognition model according to the overall model corresponding to the object data, employee information, and equipment information through a convolutional neural network;

[0106] If the risk control behavior level corresponding to the risk control behavior information is white risk, the standard behavior data generated according to the risk control behavior information and the bag-of-words dataset is marked as standard regular behavior data;

[0107] If the risk control behavior level corresponding to the risk control behavior information is not white risk, the standard behavior data generated according to the risk control behavior information and the bag-of-words dataset is marked as standard abnormal behavior data;

[0108] It should be further explained that, in a specific implementation process, the process of generating a standard behavior model according to the standard behavior data includes:

[0109] The standard behavior model includes a standard normal behavior sub-model and a standard abnormal behavior sub-model;

[0110] Generate a standard normal behavior sub-model based on standard normal behavior data through deep learning, and generate a standard abnormal behavior sub-model based on standard abnormal behavior data through deep learning;

[0111] The standard behavior model input data is subject number, valid object data and real-time video data, and output data is generated according to the subject number, valid object data and real-time video data, and the output data includes 0 and 1;

[0112] The subject number is the number corresponding to the subject data. If the subject data is an employee, the number is the employee number; if the subject data is a device, the number is the device number;

[0113] The process of generating output data according to the subject number, valid object data and real-time video data is as follows:

[0114] The standard conventional behavior sub-model input data is subject number, valid object data and real-time video data, and generates output data according to the input data, wherein the output data includes T and F;

[0115] The input data of the standard abnormal behavior sub-model is the subject number, valid object data and real-time video data, and output data is generated according to the input data, and the output data includes T and F;

[0116] When the standard behavior model receives the subject number, valid object data and real-time video data, it sends the subject number, valid object data and real-time video data to the standard normal behavior sub-model and the standard abnormal behavior sub-model respectively;

[0117] Obtain output data of a standard normal behavior sub-model and a standard abnormal behavior sub-model;

[0118] If the output data of the standard normal behavior sub-model is T, and the output data of the standard abnormal behavior sub-model is F, then the output data of the standard behavior model is 0;

[0119] If the output data of the standard normal behavior sub-model is F, and the output data of the standard abnormal behavior sub-model is T, then the output data of the standard behavior model is 1;

[0120] Otherwise, the subject number is marked as a suspicious number, and a third-level warning information is generated according to the suspicious number, and the third-level warning information is sent to the risk processing module, and the risk processing module makes corresponding processing;

[0121] It should be further explained that, in the specific implementation process, the process of obtaining the real-time behavior data of employees and devices in real time includes:

[0122] The enterprise is divided into several monitoring areas, each of which is provided with a camera and an infrared thermal imager; the camera is used to obtain ordinary video of the monitoring area, and the infrared thermal imager is used to obtain thermal video of the monitoring area;

[0123] Set the monitoring cycle and regularly obtain normal video and thermal video according to the monitoring cycle;

[0124] It should be noted that there is a one-to-one correspondence between ordinary videos and thermal imaging videos;

[0125] Acquire image data of each employee and each device, generate a corresponding overall model based on the image data, and generate a subject recognition model based on the overall model, employee information, and device information corresponding to each employee and each device through a convolutional neural network;

[0126] The subject recognition model is used to identify objects including employees and equipment;

[0127] Detecting whether there is an object in the ordinary video by using the subject recognition model, when the subject recognition model detects that there is an object in the ordinary video, marking the object as a target object, obtaining a video clip in which the target object exists in the ordinary video, and marking the video clip as a first-level surveillance video;

[0128] Get the object number of the target object; if the target object is an employee, the object number is the employee number; if the target object is a device, the object number is the device number;

[0129] Perform object enhancement processing on the primary surveillance video to generate real-time video data;

[0130] Generate real-time behavior data based on object ID and real-time video data;

[0131] Among them, the process of object strengthening is:

[0132] Divide the first-level surveillance video into a plurality of images in units of frames, and mark the images as first-level surveillance images; obtain a thermal imaging video corresponding to the first-level surveillance video, and divide the thermal imaging video into a plurality of images in units of frames, and mark the images as first-level thermal imaging images;

[0133] It should be noted that there is a one-to-one correspondence between the first-level monitoring image and the first-level thermal image;

[0134] Obtain an overall model corresponding to the target object, and determine whether the target object in the first-level monitoring image is blocked according to the overall model. If not, no operation is required. Otherwise, obtain a first-level thermal image corresponding to the first-level monitoring image, and determine whether the target object in the first-level thermal image is blocked according to the overall model. If not, replace the first-level monitoring image with the first-level thermal image. Otherwise, generate first-level warning information according to the object number, and send the first-level warning information to the risk processing module;

[0135] It should be further explained that, in the specific implementation process, the standard behavior model is used to determine whether there is abnormal behavior in the real-time behavior data. If there is no abnormal behavior, no operation is required. If there is abnormal behavior, the process of generating corresponding risk warnings based on the abnormal behavior includes:

[0136] Obtaining an object number in the real-time behavior data, and marking the object number as a subject number;

[0137] Acquire real-time video data, divide the real-time video data into a plurality of images in units of frames, and mark the images as images to be detected;

[0138] Acquire a subject object in the image to be detected by using a subject recognition model; the subject object is an object whose object number is consistent with the subject number;

[0139] Obtaining pixel points corresponding to the main object, and marking the pixel points as main pixel points;

[0140] Acquire object data corresponding to the subject data when the subject object is subject data, and acquire a quasi-object object according to the object data through an object recognition model;

[0141] Acquire pixel points corresponding to the quasi-object object, and mark the pixel points as quasi-object pixel points;

[0142] Establish a two-dimensional coordinate system and map the image to be detected into the two-dimensional coordinate system so that each pixel has its corresponding coordinates;

[0143] Obtain the horizontal difference value and the vertical difference value of the main object, and obtain the main center coordinates of the main object according to the horizontal difference value and the vertical difference value of the main object;

[0144] The horizontal difference is the difference between the maximum horizontal coordinate value and the minimum horizontal coordinate value in the corresponding coordinates of the main pixel point; the vertical difference is the difference between the maximum vertical coordinate value and the minimum vertical coordinate value in the corresponding coordinates of the main pixel point; the main center coordinate = (horizontal difference, vertical difference);

[0145] According to the principle of obtaining the corresponding principal center coordinates of the subject object, the corresponding guest center coordinates of the quasi-object object are obtained;

[0146] Set the correlation threshold;

[0147] Generate a distance vector according to the main center coordinates and the guest center coordinates, obtain the modulus of the distance vector, and if the modulus is less than the associated threshold, mark the distance vector as a valid vector, otherwise, mark the distance vector as an invalid vector;

[0148] Mark the quasi-object object corresponding to the valid vector as a valid object object;

[0149] Acquire object data corresponding to a valid object object, and mark the object data as valid object data;

[0150] When the subject object is an employee and the valid object data is a device, or the subject object is a device and the valid object data is an employee, the user number corresponding to the device and the employee number corresponding to the employee are obtained;

[0151] If the user number corresponding to the device and the employee number corresponding to the employee are consistent, no operation is required; otherwise, the subject number is marked as a mismatch number, and a level 4 warning information is generated according to the mismatch number, and the level 4 warning information is submitted to the risk processing module;

[0152] Input the subject number, valid object data and real-time video data in the real-time behavior data into the standard behavior model. If the output data of the standard behavior model is 0, there is no abnormal behavior in the real-time behavior data and no operation is required. If the output data is 1, there is abnormal behavior in the real-time behavior data, and the secondary warning information is generated according to the subject number, and the secondary warning information is sent to the risk processing module;

[0153] It should be further explained that, in the specific implementation process, the process of generating corresponding risk resolution strategies based on risk warnings includes:

[0154] When a first-level warning message is received, the object number is sent to the user, informing the user that the behavior of the object cannot be accurately monitored, and the user is asked to take corresponding measures;

[0155] When receiving the second-level warning information, if the subject number is an employee number, the employee number is sent to the user to inform the user that the employee has abnormal behavior and the user takes corresponding measures. If the subject number is a device number, the device number is sent to the user to inform the user that the device has abnormal behavior and the user takes corresponding measures. The associated number corresponding to the device number is obtained, and the devices corresponding to the device number and the associated number are powered off;

[0156] When receiving a level 3 warning message, the suspicious number will be sent to the user, and the user will be informed that the behavior of the object corresponding to the suspicious number cannot be identified, and the user will take corresponding measures;

[0157] When receiving the fourth-level warning information, the mismatch number is sent to the user, informing the user that the object corresponding to the mismatch number is mismatched, and the user takes corresponding measures;

[0158] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A dual prevention system for enterprise safety production, comprising a control center, characterized in that: The control center is communicatively connected to a database, a standard behavior model generation module, a real-time behavior data acquisition module, a real-time behavior data processing module and a risk processing module; The database is used to store employee information, equipment information and risk management information, and the risk management information includes risk management behavior information and risk management behavior level; The standard behavior model generation module is used to generate standard behavior data corresponding to employees and equipment respectively according to risk management and control information, and generate a standard behavior model according to the standard behavior data; The real-time behavior data collection module is used to obtain the real-time behavior data of employees and equipment respectively in real time; The real-time behavior data processing module is used to determine whether there is abnormal behavior in the real-time behavior data through a standard behavior model. If there is no abnormal behavior, no operation is required. If there is abnormal behavior, a corresponding risk warning is generated according to the abnormal behavior. The risk processing module is used to generate corresponding risk resolution strategies according to risk warnings; The process of generating standard behavior data corresponding to employees and equipment based on risk management information includes: Setting a bag-of-words dataset, wherein the bag-of-words dataset includes an object bag-of-words, a behavior bag-of-words, an intention bag-of-words, and a region bag-of-words; The object word bag is composed of several object keywords; the behavior word bag is composed of several behavior keywords; the intention word bag is composed of several intention keywords; the region word bag is composed of several region keywords; Generate standard behavior data corresponding to employees and equipment respectively according to the bag-of-words data set and risk control behavior information; the standard behavior data includes subject data, object data and standard data; If standard behavior data corresponding to an employee is generated according to risk control information, the subject data of the standard behavior data is the employee, and object keywords in the risk control behavior information included in the risk control information are filtered out according to the object word bag, and object data is generated according to the object keywords. Behavior keywords, intention keywords and regional keywords in the risk control behavior information in the risk control information are filtered out according to the behavior word bag, intention word bag and regional word bag, and standard data is generated according to the filtered behavior keywords, intention keywords and regional keywords; Based on the principle of generating standard behavior data corresponding to employees according to risk control information, standard behavior data corresponding to the equipment is obtained; Filter out object keywords in the risk control behavior information according to the object word bag, and generate object data according to the remaining object keywords except the object keywords corresponding to the subject data in the object keywords; if there is only one object keyword in the object keywords filtered out according to the object word bag and the object keyword is consistent with the object keyword corresponding to the subject data, the object data is NULL; Acquire image data corresponding to the object data, generate a corresponding overall model according to the image data, and generate an object recognition model according to the overall model corresponding to the object data, employee information, and equipment information through a convolutional neural network; If the risk control behavior level corresponding to the risk control behavior information is white risk, the standard behavior data generated according to the risk control behavior information and the bag-of-words dataset is marked as standard regular behavior data; otherwise, the standard behavior data generated according to the risk control behavior information and the bag-of-words dataset is marked as standard abnormal behavior data; The process of generating a standard behavior model according to the standard behavior data includes: The standard behavior model includes a standard normal behavior sub-model and a standard abnormal behavior sub-model; Generate a standard normal behavior sub-model based on standard normal behavior data through deep learning, and generate a standard abnormal behavior sub-model based on standard abnormal behavior data through deep learning; The standard behavior model input data is subject number, valid object data and real-time video data, and output data is generated according to the subject number, valid object data and real-time video data, and the output data includes 0 and 1; The subject number is the number corresponding to the subject data. If the subject data is an employee, the number is the employee number; if the subject data is a device, the number is the device number; The process of obtaining real-time behavioral data of employees and equipment includes: The enterprise is divided into several monitoring areas, each of which is provided with a camera and an infrared thermal imager; the camera is used to obtain ordinary video of the monitoring area, and the infrared thermal imager is used to obtain thermal video of the monitoring area; Set the monitoring cycle and regularly obtain ordinary videos and thermal imaging videos according to the monitoring cycle; ordinary videos and thermal imaging videos are in a one-to-one correspondence; Acquire image data of each employee and each device, generate a corresponding overall model based on the image data, and generate a subject recognition model based on the overall model, employee information, and device information corresponding to each employee and each device through a convolutional neural network; The subject recognition model is used to identify objects including employees and equipment; Detecting whether there is an object in the ordinary video by using the subject recognition model, when the subject recognition model detects that there is an object in the ordinary video, marking the object as a target object, obtaining a video clip in which the target object exists in the ordinary video, and marking the video clip as a first-level surveillance video; Get the object number of the target object; if the target object is an employee, the object number is the employee number; if the target object is a device, the object number is the device number; Perform object enhancement processing on the primary surveillance video to generate real-time video data; Generate real-time behavior data based on object ID and real-time video data; The standard behavior model is used to determine whether there is abnormal behavior in the real-time behavior data. If there is no abnormal behavior, no operation is required. If there is abnormal behavior, the process of generating corresponding risk warnings based on the abnormal behavior includes: Obtaining an object number in the real-time behavior data, and marking the object number as a subject number; Acquire real-time video data, divide the real-time video data into a number of images in units of frames, and mark the images as images to be detected; acquire a subject object in the image to be detected by using a subject recognition model; the subject object is an object whose object number is consistent with the subject number; Obtaining pixel points corresponding to the main object, and marking the pixel points as main pixel points; Acquire object data corresponding to the subject data when the subject object is subject data, and acquire a quasi-object object according to the object data through an object recognition model; Acquire pixel points corresponding to the quasi-object object, and mark the pixel points as quasi-object pixel points; Establish a two-dimensional coordinate system and map the image to be detected into the two-dimensional coordinate system so that each pixel has its corresponding coordinates; Obtain the horizontal difference value and the vertical difference value of the main object, and obtain the principal center coordinates of the main object according to the horizontal difference value and the vertical difference value of the main object; obtain the object center coordinates corresponding to the quasi-object object according to the principle of obtaining the corresponding principal center coordinates of the main object; Set an association threshold; generate a distance vector according to the main center coordinates and the guest center coordinates, obtain the modulus of the distance vector, and if the modulus is less than the association threshold, mark the distance vector as a valid vector, otherwise, mark the distance vector as an invalid vector; Mark the quasi-object object corresponding to the valid vector as a valid object object; Obtain object data corresponding to a valid object object, and mark the object data as valid object data; when the subject object is an employee and the valid object data is a device, or when the subject object is a device and the valid object data is an employee, obtain the user number corresponding to the device and the employee number corresponding to the employee; if the user number corresponding to the device and the employee number corresponding to the employee are inconsistent, mark the subject number as a mismatch number, generate a four-level warning information according to the mismatch number, and submit the four-level warning information to the risk processing module; The subject number, valid object data and real-time video data in the real-time behavior data are input into the standard behavior model. If the output data of the standard behavior model is 0, there is no abnormal behavior in the real-time behavior data and no operation is required. If the output data is 1, there is abnormal behavior in the real-time behavior data, and the secondary warning information is generated according to the subject number and sent to the risk processing module.

2. A dual prevention system for enterprise safety production according to claim 1, characterized in that: The database includes an employee unit, an equipment unit, and a risk unit; The employee unit is used to store employee information, which includes employee number, employee name, ID number, contact information and operation proficiency value; the employee number is unique; The device unit is used to store device information, and the device information includes device number, device name, device model, user number and associated number; wherein the device number is unique, the user number is the employee number corresponding to the employee who uses the device, and the associated number is the device number corresponding to the device that has an associated relationship with the device; The risk unit is used to store risk management information, which includes risk management behavior information and risk management behavior level; wherein the risk management behavior information is text information describing the risk management behavior, and the risk management behavior level is the risk level corresponding to the risk management behavior information; the risk level includes white risk, blue risk, yellow risk, orange risk and red risk, wherein white risk is the lowest risk level and red risk is the highest risk level.

3. A dual prevention system for enterprise safety production according to claim 2, characterized in that: The process of generating output data based on subject number, valid object data and real-time video data is: The standard conventional behavior sub-model input data is subject number, valid object data and real-time video data, and generates output data according to the input data, wherein the output data includes T and F; The input data of the standard abnormal behavior sub-model is the subject number, valid object data and real-time video data, and output data is generated according to the input data, and the output data includes T and F; When the standard behavior model receives the subject number, valid object data and real-time video data, it sends the subject number, valid object data and real-time video data to the standard normal behavior sub-model and the standard abnormal behavior sub-model respectively; Obtain output data of a standard normal behavior sub-model and a standard abnormal behavior sub-model; If the output data of the standard normal behavior sub-model is T and the output data of the standard abnormal behavior sub-model is F, then the output data of the standard behavior model is 0; if the output data of the standard normal behavior sub-model is F and the output data of the standard abnormal behavior sub-model is T, then the output data of the standard behavior model is 1; otherwise, the subject number is marked as the suspicious point number, and the third-level warning information is generated according to the suspicious point number, and the third-level warning information is sent to the risk processing module, and the risk processing module makes corresponding processing.

4. A dual prevention system for enterprise safety production according to claim 3, characterized in that: The process of object enhancement is as follows: Divide the first-level surveillance video into a plurality of images in units of frames, and mark the images as first-level surveillance images; obtain a thermal imaging video corresponding to the first-level surveillance video, and divide the thermal imaging video into a plurality of images in units of frames, and mark the images as first-level thermal imaging images; There is a one-to-one correspondence between the first-level monitoring image and the first-level thermal image; Obtain an overall model corresponding to the target object, and determine whether the target object in the first-level monitoring image is obscured based on the overall model. If not, no operation is required. Otherwise, obtain a first-level thermal image corresponding to the first-level monitoring image, and determine whether the target object in the first-level thermal image is obscured based on the overall model. If not, use the first-level thermal image to replace the first-level monitoring image. Otherwise, generate a first-level warning information based on the object number, and send the first-level warning information to the risk processing module.

5. A dual prevention system for enterprise safety production according to claim 4, characterized in that: The process of generating corresponding risk resolution strategies based on risk warnings includes: When a first-level warning message is received, the object number is sent to the user, informing the user that the behavior of the object cannot be accurately monitored, and the user is asked to take corresponding measures; When receiving the second-level warning information, if the subject number is an employee number, the employee number is sent to the user to inform the user that the employee has abnormal behavior and the user takes corresponding measures. If the subject number is a device number, the device number is sent to the user to inform the user that the device has abnormal behavior and the user takes corresponding measures. The associated number corresponding to the device number is obtained, and the devices corresponding to the device number and the associated number are powered off; When receiving a level 3 warning message, the suspicious number will be sent to the user, and the user will be informed that the behavior of the object corresponding to the suspicious number cannot be identified, and the user will take corresponding measures; When a level 4 warning message is received, the mismatch number is sent to the user, informing the user that the object corresponding to the mismatch number is mismatched, and the user takes corresponding action.

Images