Multi-source induction early warning method and system based on dangerous scene of production environment

Abstract

The invention relates to a multi-source induction early warning method and system based on a production environment dangerous scene, and belongs to the technical field of monitoring and early warning. The method comprises the following steps: obtaining a historical monitoring picture of a core production node, and obtaining a historical monitoring high-definition picture of the core production node through an EffiSAM-ESRGAN algorithm; constructing a core production node monitoring set according to the core production node historical monitoring high-definition pictures, and constructing a core production node violation identification model through an SE-FocalNASNet algorithm according to the core production node monitoring set; monitoring a core production node monitoring video in real time, obtaining a core production node personnel behavior label through the core production node violation identification model, and sending an early warning prompt to a terminal according to the core production node personnel behavior label; the production field multi-source sensing data is monitored in real time, and when the production field multi-source sensing data exceeds the preset production field multi-source sensing data range, the production environment early warning prompt is sent to the terminal, so that the omnibearing and intelligent monitoring and early warning of the dangerous scene of the production environment are realized, and the production safety level is improved.

Description

Technical Field

[0001] The present invention belongs to the field of monitoring and early warning technology, and in particular relates to a multi-source sensing early warning method and system based on dangerous scenarios in a production environment. Background Art

[0002] The chemical production process is full of potential risks, including the leakage of flammable and explosive gases, abnormalities of high-temperature and high-pressure equipment, etc. Ensuring production safety is the primary task of the enterprise. With the expansion of the production scale and the increase in process complexity of chemical enterprises, the safety risks in the production process are increasing. Therefore, how to effectively monitor and warn of these dangerous situations is crucial to protecting the lives of employees and the continuity of production.

[0003] At present, most monitoring and early warning systems usually rely on specific types of sensors and equipment to identify potential hazards, but often only focus on a specific risk scenario, which limits their ability to comprehensively and timely detect various potential hazards in chemical production. When multiple risk factors exist at the same time, the existing system may not be able to provide accurate alarms, which will affect the overall risk identification and response, leaving safety hazards. In addition, many traditional monitoring and early warning systems rely on rules and thresholds for judgment and lack intelligent analysis capabilities, resulting in low accuracy in identifying dangerous scenes in the production environment, prone to false alarms and missed alarms, increasing the workload of managers and even causing safety accidents. More importantly, traditional early warning systems usually have only a single warning method, which is difficult to effectively attract the attention of staff in an emergency, which may result in a failure to respond in time when danger occurs.

[0004] Therefore, it is urgent to develop an intelligent and integrated multi-source sensing warning system to detect dangerous scenes in the production environment and issue timely warnings. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention proposes a multi-source sensing early warning method and system based on dangerous scenarios in a production environment.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A multi-source sensing early warning method based on dangerous scenes in a production environment, comprising:

[0008] Obtain historical monitoring pictures of core production nodes, and obtain historical monitoring high-definition pictures of core production nodes through the EffiSAM-ESRGAN algorithm based on the historical monitoring pictures of core production nodes;

[0009] Building a core production node monitoring set based on the core production node historical monitoring high-definition pictures, and building a core production node violation identification model based on the core production node monitoring set using the SE-FocalNASNet algorithm;

[0010] Real-time monitoring of core production node surveillance videos, obtaining core production node personnel behavior labels through the core production node violation identification model based on the core production node surveillance videos, and sending early warning prompts to the terminal based on the core production node personnel behavior labels;

[0011] Real-time monitoring of multi-source sensor data in the production field, wherein the multi-source sensor data in the production field includes gaseous molecule diffusion data, temperature sensing data, and smoke sensing data. When the multi-source sensor data in the production field exceeds the preset multi-source sensor data range of the production field, a production environment early warning prompt is sent to the terminal.

[0012] Preferably, the obtaining of the historical monitoring pictures of the core production nodes and obtaining the historical monitoring high-definition pictures of the core production nodes by using the EffiSAM-ESRGAN algorithm according to the historical monitoring pictures of the core production nodes include:

[0013] Deploy industrial-grade high-definition cameras at core production nodes to obtain historical monitoring images of core production nodes, including but not limited to operation consoles, equipment control areas, and hazardous materials handling areas;

[0014] The EffiSAM-ESRGAN algorithm includes a generator and a discriminator. The generator receives the core production node historical monitoring picture through an input layer, replaces the basic network in ESRGAN with EfficientNet to extract preliminary features to improve the performance and efficiency of the model, and then introduces a spatial attention mechanism to enable the network to focus on key features in the image, thereby better retaining details during super-resolution reconstruction. The feature map is further processed through three residual dense blocks, and feature reuse and direct transfer of gradient flow are achieved through jump connections to ensure the stability and efficiency of training. Finally, the size of the feature map is enlarged to the size of a preset high-resolution core production node monitoring image through an upsampling layer, and a high-definition picture of the core production node historical monitoring is generated through an output layer;

[0015] The discriminator is the discriminator structure in the original ESRGAN. During the training process, the generator and the discriminator are updated alternately through adversarial training. At the same time, on the basis of the original perceptual loss and adversarial loss, the structural similarity SSIM loss is added to better maintain the structural information of the image. At the same time, the gradient loss is introduced to encourage the generated image to have clearer edges.

[0016] The loss function of the EffiSAM-ESRGAN algorithm is expressed as:

[0017]

[0018] Among them, a is the perceptual loss weight coefficient, b is the adversarial loss weight coefficient, c is the structural similarity SSIM loss weight coefficient, d is the gradient loss weight coefficient, G(x) is the historical monitoring HD picture of the core production node generated by the generator, y is the real historical monitoring HD picture of the core production node, T i is the feature extraction function of the i-th layer of the network, S i is the weight of the i-th layer, D is the discriminator, G is the generator, x is the input core production node historical monitoring image, E x is the average loss of the input core production node historical monitoring picture set, SSIM(G(x), y) is the structural similarity index between the generated core production node historical monitoring HD pictures and the real core production node historical monitoring HD pictures, ΔG(x) and Δy are the gradients of the generated core production node historical monitoring HD pictures and the real core production node historical monitoring HD pictures, respectively.

[0019] Preferably, the step of constructing a core production node monitoring set based on the core production node historical monitoring high-definition pictures, and constructing a core production node violation identification model based on the core production node monitoring set by using the SE-FocalNASNet algorithm includes:

[0020] Performing data enhancement on the historical monitoring high-definition pictures of the core production nodes to obtain a core production node monitoring set;

[0021] Obtaining a core production node monitoring training set and a core production node monitoring test set by randomly dividing the core production node monitoring set;

[0022] The core production node monitoring training set is trained using the SE-FocalNASNet algorithm, and further testing and evaluation is performed on the core production node monitoring test set to obtain a core production node violation identification model;

[0023] The core production node violation identification model is used to identify the core production node personnel behavior labels, which include normal behavior labels of core production node personnel and abnormal behavior labels of core production node personnel. The abnormal behavior labels of core production node personnel include off-duty violation behavior labels and other violation behavior labels.

[0024] Preferably, the SE-FocalNASNet algorithm includes integrating the SE attention mechanism on the basis of the NASNet architecture to enhance the model's attention to key areas. At the same time, focal loss is used to balance the positive and negative samples in the core production node monitoring set, and an L2 regularization term is added to prevent overfitting. During the training process, the optimizer uses the Adam optimizer, and the learning rate is initialized to 0.001;

[0025] The loss function of the SE-FocalNASNet algorithm is expressed as:

[0026]

[0027] Among them, α is the positive and negative sample balance coefficient, which is set to 0.25 and is used to adjust the contribution of positive and negative samples to the loss function. i is the model weight parameter, p t is the predicted probability of the core production node monitoring set, β is the focal loss weight coefficient, which is set to 2, and λ is the regularization coefficient, which is used to control the strength of the L2 regularization term.

[0028] Preferably, the off-duty violation behavior label includes monitoring the core production node 24 hours a day during production, and when there is no one in the core production node, it is an off-duty violation behavior label;

[0029] The other violation label is when other violation behaviors other than leaving the post occur. The other violation behaviors include but are not limited to playing with mobile phones, listening to music, and playing around.

[0030] Preferably, the real-time monitoring of the core production node surveillance video, obtaining the core production node personnel behavior label through the core production node violation identification model according to the core production node surveillance video, and sending an early warning prompt to the terminal according to the core production node personnel behavior label includes:

[0031] The core production nodes are monitored by industrial-grade high-definition cameras deployed at the core production nodes;

[0032] According to the core production node monitoring video, video frames are intercepted to obtain a core production node monitoring picture;

[0033] Inputting the core production node monitoring image into the core production node violation identification model to obtain the core production node personnel behavior label;

[0034] When the core production node personnel behavior label is an off-duty violation label or other violation label, a core production node personnel violation warning prompt is sent to the terminal.

[0035] Preferably, the real-time monitoring of multi-source sensor data of the production field, the multi-source sensor data of the production field including gaseous molecule diffusion data, temperature sensing data, and smoke sensing data, when the multi-source sensor data of the production field exceeds a preset range of the multi-source sensor data of the production field, sending a production environment early warning prompt to the terminal includes:

[0036] Install environmental sensors in key areas, including toxic and hazardous gas leakage sensors, temperature detectors, and smoke detectors;

[0037] Monitoring multi-source sensor data of the production site through the environmental sensor;

[0038] Preset the multi-source sensor data range of the production site, and the multi-source sensor data range of the production site is set by the production enterprise itself;

[0039] When the multi-source sensor data of the production site exceeds the preset multi-source sensor data range of the production site, a production environment early warning prompt is automatically sent to the control room. At the same time, sound and light alarms are triggered on site, and warning lights and sirens are triggered in the monitoring room. The video monitoring screen automatically pops up a warning and a prompt is broadcast on site.

[0040] A multi-source sensing early warning system based on dangerous production environment scenarios, applied to the above-mentioned multi-source sensing early warning method based on dangerous production environment scenarios, including a data acquisition and processing module, a violation identification model building module, a violation identification early warning module, and an environmental monitoring early warning module;

[0041] The data acquisition and processing module is used to obtain the historical monitoring pictures of the core production nodes, and obtain the historical monitoring high-definition pictures of the core production nodes through the EffiSAM-ESRGAN algorithm according to the historical monitoring pictures of the core production nodes;

[0042] The violation identification model building module is used to build a core production node monitoring set based on the core production node historical monitoring high-definition pictures, and build a core production node violation identification model through the SE-FocalNASNet algorithm based on the core production node monitoring set;

[0043] The violation identification and early warning module is used to monitor the core production node monitoring video in real time, obtain the core production node personnel behavior label based on the core production node monitoring video through the core production node violation identification model, and send an early warning prompt to the terminal based on the core production node personnel behavior label;

[0044] The environmental monitoring and early warning module is used to monitor the multi-source sensor data of the production site in real time. The multi-source sensor data of the production site includes gaseous molecule diffusion data, temperature sensing data, and smoke sensing data. When the multi-source sensor data of the production site exceeds the preset multi-source sensor data range of the production site, a production environment early warning prompt is sent to the terminal.

[0045] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the multi-source sensing warning method based on dangerous scenes in a production environment is implemented.

[0046] A storage medium containing computer executable instructions, which, when executed by a computer processor, are used to execute the above-mentioned multi-source sensing early warning method based on dangerous scenarios in a production environment.

[0047] The beneficial effects of the present invention are:

[0048] (1) The EffiSAM-ESRGAN algorithm is used to obtain historical high-definition monitoring images of core production nodes. The spatial attention module is introduced to better preserve image details. The SSIM loss can better maintain the structural information of the image, and the gradient loss encourages the generated image to have clearer edges, thereby improving the clarity of the reconstructed image. Replacing the basic network in ESRGAN with EfficientNet can speed up training and improve reconstruction quality, and can generate clearer and more realistic high-definition images, making subsequent violation identification more accurate.

[0049] (2) A core production node violation identification model is constructed through the SE-FocalNASNet algorithm. The SE attention mechanism is integrated into NASNet to enhance the model's attention to key areas. The focal loss is used to balance the core production node monitoring set. At the same time, a regularization term is added to prevent overfitting, which can more accurately identify violations.

[0050] (3) Real-time monitoring of core production node surveillance videos, and real-time generation of personnel behavior labels through violation identification models. At the same time, real-time monitoring of multi-source sensor data in the production field has achieved comprehensive monitoring of the production environment. It can not only warn of personnel violations, but also warn of production environment anomalies, reducing the frequency and cost of manual inspections, improving safety management efficiency, and being able to timely discover and warn of violations and production environment anomalies, effectively reducing the occurrence of accidents and ensuring the safety of personnel and equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0052] Figure 1 The present invention is a flowchart of a multi-source sensing early warning method based on dangerous scenes in a production environment. DETAILED DESCRIPTION

[0053] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0054] See also Figure 1 , a multi-source sensing early warning method based on dangerous scenes in production environment, comprising:

[0055] S1: Obtain historical monitoring pictures of core production nodes, and obtain historical monitoring high-definition pictures of core production nodes through the EffiSAM-ESRGAN algorithm according to the historical monitoring pictures of core production nodes;

[0056] S2: construct a core production node monitoring set based on the historical monitoring high-definition pictures of the core production nodes, and construct a core production node violation identification model based on the core production node monitoring set using the SE-FocalNASNet algorithm;

[0057] S3: Real-time monitoring of the core production node surveillance video, obtaining the core production node personnel behavior label through the core production node violation identification model according to the core production node surveillance video, and sending an early warning prompt to the terminal according to the core production node personnel behavior label;

[0058] S4: Real-time monitoring of multi-source sensor data of the production site, wherein the multi-source sensor data of the production site includes gaseous molecule diffusion data, temperature sensing data, and smoke sensing data. When the multi-source sensor data of the production site exceeds the preset range of the multi-source sensor data of the production site, a production environment early warning prompt is sent to the terminal.

[0059] In this embodiment, the acquisition of the core production node historical monitoring picture and the acquisition of the core production node historical monitoring high-definition picture through the EffiSAM-ESRGAN algorithm are specifically implemented by the following steps:

[0060] Deploy industrial-grade high-definition cameras at core production nodes to obtain historical monitoring images of core production nodes, including but not limited to operation consoles, equipment control areas, and hazardous materials handling areas;

[0061] The EffiSAM-ESRGAN algorithm includes a generator and a discriminator. The generator receives the core production node historical monitoring picture through an input layer, replaces the basic network in ESRGAN with EfficientNet to extract preliminary features to improve the performance and efficiency of the model, and then introduces a spatial attention mechanism to enable the network to focus on key features in the image, thereby better retaining details during super-resolution reconstruction. The feature map is further processed through three residual dense blocks, and feature reuse and direct transfer of gradient flow are achieved through jump connections to ensure the stability and efficiency of training. Finally, the size of the feature map is enlarged to the size of a preset high-resolution core production node monitoring image through an upsampling layer, and a high-definition picture of the core production node historical monitoring is generated through an output layer;

[0062] The discriminator is the discriminator structure in the original ESRGAN. During the training process, the generator and the discriminator are updated alternately through adversarial training. At the same time, on the basis of the original perceptual loss and adversarial loss, the structural similarity SSIM loss is added to better maintain the structural information of the image. At the same time, the gradient loss is introduced to encourage the generated image to have clearer edges.

[0063] The loss function of the EffiSAM-ESRGAN algorithm is expressed as:

[0064]

[0065] Among them, a is the perceptual loss weight coefficient, b is the adversarial loss weight coefficient, c is the structural similarity SSIM loss weight coefficient, d is the gradient loss weight coefficient, G(x) is the historical monitoring HD picture of the core production node generated by the generator, y is the real historical monitoring HD picture of the core production node, T i is the feature extraction function of the i-th layer of the network, S i is the weight of the i-th layer, D is the discriminator, G is the generator, x is the input core production node historical monitoring image, E x is the average loss of the input core production node historical monitoring picture set, SSIM(G(x), y) is the structural similarity index between the generated core production node historical monitoring HD pictures and the real core production node historical monitoring HD pictures, ΔG(x) and Δy are the gradients of the generated core production node historical monitoring HD pictures and the real core production node historical monitoring HD pictures, respectively.

[0066] In this embodiment, the core production node monitoring set is constructed according to the core production node historical monitoring high-definition pictures, and the core production node violation identification model is constructed according to the core production node monitoring set by using the SE-FocalNASNet algorithm, which is specifically implemented by the following steps:

[0067] Performing data enhancement on the historical monitoring high-definition pictures of the core production nodes to obtain a core production node monitoring set;

[0068] Obtaining a core production node monitoring training set and a core production node monitoring test set by randomly dividing the core production node monitoring set;

[0069] The core production node monitoring training set is trained using the SE-FocalNASNet algorithm, and further testing and evaluation is performed on the core production node monitoring test set to obtain a core production node violation identification model;

[0070] The SE-FocalNASNet algorithm includes integrating the SE attention mechanism on the basis of the NASNet architecture to enhance the model's attention to key areas. At the same time, the focal loss is used to balance the positive and negative samples in the core production node monitoring set, and the L2 regularization term is added to prevent overfitting. During the training process, the optimizer uses the Adam optimizer, and the learning rate is initialized to 0.001;

[0071] The loss function of the SE-FocalNASNet algorithm is expressed as:

[0072]

[0073] Among them, α is the positive and negative sample balance coefficient, which is set to 0.25 and is used to adjust the contribution of positive and negative samples to the loss function. i is the model weight parameter, p t is the predicted probability of the core production node monitoring set, β is the focal loss weight coefficient, which is set to 2, and λ is the regularization coefficient, which is used to control the strength of the L2 regularization term.

[0074] It should be noted that the predicted probability of the core production node monitoring set includes the probability that the model predicts a positive sample for the core production node monitoring data; and the probability that the model predicts a negative sample for the core production node monitoring data.

[0075] The core production node violation identification model is used to identify the core production node personnel behavior labels, which include normal behavior labels of core production node personnel and abnormal behavior labels of core production node personnel. The abnormal behavior labels of core production node personnel include off-duty violation behavior labels and other violation behavior labels.

[0076] The off-duty violation label includes 24-hour uninterrupted monitoring of the core production node during production. When there is no one in the core production node, it is an off-duty violation label;

[0077] The other violation label is when other violation behaviors other than leaving the post occur. The other violation behaviors include but are not limited to playing with mobile phones, listening to music, and playing around.

[0078] In this embodiment, the real-time monitoring of the core production node surveillance video, obtaining the core production node personnel behavior label through the core production node violation identification model according to the core production node surveillance video, and sending the early warning prompt to the terminal according to the core production node personnel behavior label are specifically implemented by the following steps:

[0079] The core production nodes are monitored by industrial-grade high-definition cameras deployed at the core production nodes;

[0080] According to the core production node monitoring video, video frames are intercepted to obtain a core production node monitoring picture;

[0081] Inputting the core production node monitoring image into the core production node violation identification model to obtain the core production node personnel behavior label;

[0082] When the core production node personnel behavior label is an off-duty violation label or other violation label, a core production node personnel violation warning prompt is sent to the terminal.

[0083] Specifically, the violation warning prompt of the core production node personnel is first sent to the management personnel in charge of production safety of the enterprise. If the violation warning prompt information of the core production node is not processed within the preset time, it will be automatically forwarded to the responsible leader of the enterprise and the duty room of the district and municipal safety supervision department.

[0084] In this embodiment, the real-time monitoring of the multi-source sensor data of the production field, the multi-source sensor data of the production field including gaseous molecular diffusion data, temperature sensing data, and smoke sensing data, when the multi-source sensor data of the production field exceeds the preset multi-source sensor data range of the production field, sending the production environment early warning prompt to the terminal is specifically implemented by the following steps:

[0085] Install environmental sensors in key areas, including toxic and hazardous gas leakage sensors, temperature detectors, and smoke detectors;

[0086] Monitoring multi-source sensor data of the production site through the environmental sensor;

[0087] Preset the multi-source sensor data range of the production site, and the multi-source sensor data range of the production site is set by the production enterprise itself;

[0088] When the multi-source sensor data of the production site exceeds the preset multi-source sensor data range of the production site, a production environment early warning prompt is automatically sent to the control room. At the same time, sound and light alarms are triggered on site, and warning lights and sirens are triggered in the monitoring room. The video monitoring screen automatically pops up a warning and a prompt is broadcast on site.

[0089] A multi-source sensing warning system based on dangerous scenes in production environment includes a data acquisition and processing module, a violation identification model building module, a violation identification warning module, and an environmental monitoring warning module;

[0090] The data acquisition and processing module is used to obtain the historical monitoring pictures of the core production nodes, and obtain the historical monitoring high-definition pictures of the core production nodes through the EffiSAM-ESRGAN algorithm according to the historical monitoring pictures of the core production nodes;

[0091] The violation identification model building module is used to build a core production node monitoring set based on the core production node historical monitoring high-definition pictures, and build a core production node violation identification model through the SE-FocalNASNet algorithm based on the core production node monitoring set;

[0092] The violation identification and early warning module is used to monitor the core production node monitoring video in real time, obtain the core production node personnel behavior label based on the core production node monitoring video through the core production node violation identification model, and send an early warning prompt to the terminal based on the core production node personnel behavior label;

[0093] The environmental monitoring and early warning module is used to monitor the multi-source sensor data of the production site in real time. The multi-source sensor data of the production site includes gaseous molecule diffusion data, temperature sensing data, and smoke sensing data. When the multi-source sensor data of the production site exceeds the preset multi-source sensor data range of the production site, a production environment early warning prompt is sent to the terminal.

[0094] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device.

[0095] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0096] The program code included on the computer readable medium can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, RF, etc., or any suitable combination of the above. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or their combinations, and the programming language includes object-oriented programming languages-such as Java, Smalltalk, C++, and also includes conventional procedural programming languages-such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0097] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A multi-source sensing early warning method based on dangerous scenes in production environment, characterized in that: include: Obtain historical monitoring pictures of core production nodes, and obtain historical monitoring high-definition pictures of core production nodes through the EffiSAM-ESRGAN algorithm based on the historical monitoring pictures of core production nodes; Building a core production node monitoring set based on the core production node historical monitoring high-definition pictures, and building a core production node violation identification model based on the core production node monitoring set using the SE-FocalNASNet algorithm; Real-time monitoring of core production node surveillance videos, obtaining core production node personnel behavior labels through the core production node violation identification model based on the core production node surveillance videos, and sending early warning prompts to the terminal based on the core production node personnel behavior labels; Real-time monitoring of multi-source sensor data in the production field, wherein the multi-source sensor data in the production field includes gaseous molecule diffusion data, temperature sensing data, and smoke sensing data. When the multi-source sensor data in the production field exceeds the preset multi-source sensor data range of the production field, a production environment early warning prompt is sent to the terminal.

2. The multi-source sensing early warning method based on dangerous scenes in production environment according to claim 1 is characterized in that: The obtaining of the core production node historical monitoring picture, and obtaining the core production node historical monitoring high-definition picture through the EffiSAM-ESRGAN algorithm according to the core production node historical monitoring picture, comprises: Deploy industrial-grade high-definition cameras at core production nodes to obtain historical monitoring images of core production nodes, including but not limited to operation consoles, equipment control areas, and hazardous materials handling areas; The EffiSAM-ESRGAN algorithm includes a generator and a discriminator. The generator receives the core production node historical monitoring picture through an input layer, replaces the basic network in ESRGAN with EfficientNet to extract preliminary features to improve the performance and efficiency of the model, and then introduces a spatial attention mechanism to enable the network to focus on key features in the image, thereby better retaining details during super-resolution reconstruction. The feature map is further processed through three residual dense blocks, and feature reuse and direct transfer of gradient flow are achieved through jump connections to ensure the stability and efficiency of training. Finally, the size of the feature map is enlarged to the size of a preset high-resolution core production node monitoring image through an upsampling layer, and a high-definition picture of the core production node historical monitoring is generated through an output layer; The discriminator is the discriminator structure in the original ESRGAN. During the training process, the generator and the discriminator are updated alternately through adversarial training. At the same time, on the basis of the original perceptual loss and adversarial loss, the structural similarity SSIM loss is added to better maintain the structural information of the image. At the same time, the gradient loss is introduced to encourage the generated image to have clearer edges. The loss function of the EffiSAM-ESRGAN algorithm is expressed as: Among them, a is the perceptual loss weight coefficient, b is the adversarial loss weight coefficient, c is the structural similarity SSIM loss weight coefficient, d is the gradient loss weight coefficient, G(x) is the historical monitoring HD picture of the core production node generated by the generator, y is the real historical monitoring HD picture of the core production node, T i is the feature extraction function of the i-th layer of the network, S i is the weight of the i-th layer, D is the discriminator, G is the generator, x is the input core production node historical monitoring image, E x is the average loss of the input core production node historical monitoring picture set, SSIM(G(x), y) is the structural similarity index between the generated core production node historical monitoring HD pictures and the real core production node historical monitoring HD pictures, ΔG(x) and Δy are the gradients of the generated core production node historical monitoring HD pictures and the real core production node historical monitoring HD pictures, respectively.

3. The multi-source sensing early warning method based on dangerous scenes in production environment according to claim 1 is characterized in that: The constructing of a core production node monitoring set according to the core production node historical monitoring high-definition pictures, and constructing a core production node violation identification model according to the core production node monitoring set by using the SE-FocalNASNet algorithm include: Performing data enhancement on the historical monitoring high-definition pictures of the core production nodes to obtain a core production node monitoring set; Obtaining a core production node monitoring training set and a core production node monitoring test set by randomly dividing the core production node monitoring set; The core production node monitoring training set is trained using the SE-FocalNASNet algorithm, and further testing and evaluation is performed on the core production node monitoring test set to obtain a core production node violation identification model; The core production node violation identification model is used to identify the core production node personnel behavior labels, which include normal behavior labels of core production node personnel and abnormal behavior labels of core production node personnel. The abnormal behavior labels of core production node personnel include off-duty violation behavior labels and other violation behavior labels.

4. The multi-source sensing early warning method based on dangerous scenes in production environment according to claim 3 is characterized in that: The SE-FocalNASNet algorithm includes integrating the SE attention mechanism on the basis of the NASNet architecture to enhance the model's attention to key areas. At the same time, the focal loss is used to balance the positive and negative samples in the core production node monitoring set, and the L2 regularization term is added to prevent overfitting. During the training process, the optimizer uses the Adam optimizer, and the learning rate is initialized to 0.001; The loss function of the SE-FocalNASNet algorithm is expressed as: Among them, α is the positive and negative sample balance coefficient, which is set to 0.25 and is used to adjust the contribution of positive and negative samples to the loss function. i is the model weight parameter, p t is the predicted probability of the core production node monitoring set, β is the focal loss weight coefficient, which is set to 2, and λ is the regularization coefficient, which is used to control the strength of the L2 regularization term.

5. The multi-source sensing early warning method based on dangerous scenes in production environment according to claim 3 is characterized in that: The off-duty violation label includes 24-hour uninterrupted monitoring of the core production node during production. When there is no one in the core production node, it is an off-duty violation label; The other violation label is when other violation behaviors other than leaving the post occur. The other violation behaviors include but are not limited to playing with mobile phones, listening to music, and playing around.

6. The multi-source sensing early warning method based on dangerous scenes in production environment according to claim 1 is characterized in that: The real-time monitoring of the core production node surveillance video, obtaining the core production node personnel behavior label through the core production node violation identification model according to the core production node surveillance video, and sending the early warning prompt to the terminal according to the core production node personnel behavior label includes: The core production nodes are monitored by industrial-grade high-definition cameras deployed at the core production nodes; According to the core production node monitoring video, video frames are intercepted to obtain a core production node monitoring picture; Inputting the core production node monitoring image into the core production node violation identification model to obtain the core production node personnel behavior label; When the core production node personnel behavior label is an off-duty violation label or other violation label, a core production node personnel violation warning prompt is sent to the terminal.

7. The multi-source sensing early warning method based on dangerous scenes in production environment according to claim 1 is characterized in that: The real-time monitoring of the multi-source sensor data of the production field, wherein the multi-source sensor data of the production field includes gaseous molecular diffusion data, temperature sensing data, and smoke sensing data, and when the multi-source sensor data of the production field exceeds the preset multi-source sensor data range of the production field, sending a production environment early warning prompt to the terminal includes: Install environmental sensors in key areas, including toxic and hazardous gas leakage sensors, temperature detectors, and smoke detectors; Monitoring multi-source sensor data of the production site through the environmental sensor; Preset the multi-source sensor data range of the production site, and the multi-source sensor data range of the production site is set by the production enterprise itself; When the multi-source sensor data of the production site exceeds the preset multi-source sensor data range of the production site, a production environment early warning prompt is automatically sent to the control room. At the same time, sound and light alarms are triggered on site, and warning lights and sirens are triggered in the monitoring room. The video monitoring screen automatically pops up a warning and a prompt is broadcast on site.

8. A multi-source sensing warning system based on dangerous scenes in a production environment, applied to the multi-source sensing warning method based on dangerous scenes in a production environment as claimed in any one of claims 1 to 7, characterized in that: It includes data collection and processing module, violation identification model building module, violation identification warning module, and environmental monitoring warning module; The data acquisition and processing module is used to obtain the historical monitoring pictures of the core production nodes, and obtain the historical monitoring high-definition pictures of the core production nodes through the EffiSAM-ESRGAN algorithm according to the historical monitoring pictures of the core production nodes; The violation identification model building module is used to build a core production node monitoring set based on the core production node historical monitoring high-definition pictures, and build a core production node violation identification model through the SE-FocalNASNet algorithm based on the core production node monitoring set; The violation identification and early warning module is used to monitor the core production node monitoring video in real time, obtain the core production node personnel behavior label based on the core production node monitoring video through the core production node violation identification model, and send an early warning prompt to the terminal based on the core production node personnel behavior label; The environmental monitoring and early warning module is used to monitor the multi-source sensor data of the production site in real time. The multi-source sensor data of the production site includes gaseous molecule diffusion data, temperature sensing data, and smoke sensing data. When the multi-source sensor data of the production site exceeds the preset multi-source sensor data range of the production site, a production environment early warning prompt is sent to the terminal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the multi-source sensing warning method based on dangerous scenarios in a production environment as described in any one of claims 1-7 is implemented.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions, when executed by a computer processor, are used to execute the multi-source sensing warning method based on dangerous scenarios in a production environment as described in any one of claims 1 to 7.

Images