Energy facility real-time safety monitoring and risk early warning system based on video AI analysis

Abstract

The invention relates to the technical field of intelligent monitoring, in particular to an energy facility real-time safety monitoring and risk early warning system based on video AI analysis, which comprises a video acquisition module, an image processing module, an AI analysis module, a risk assessment module and a real-time monitoring alarm module. Wherein the video acquisition module is used for acquiring image data in an energy facility in real time; the image processing module is used for preprocessing the image data acquired by the video acquisition module; the AI analysis module is used for identifying abnormal conditions in the image; the risk assessment module is used for performing risk assessment on the abnormal condition identified by the AI analysis module; the real-time monitoring alarm module is used for sending an early warning signal to related personnel; according to the invention, through integration of the high-definition camera, image processing, AI analysis and risk assessment technologies, efficient identification, accurate risk assessment and timely early warning of abnormal conditions of energy facilities are realized, and the safety monitoring efficiency and the management capability are significantly improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent monitoring technology, and in particular to a real-time safety monitoring and risk warning system for energy facilities based on video AI analysis. Background Art

[0002] As the scale of energy facilities continues to expand, especially in the fields of oil, natural gas, and electricity, the safety management of facilities has become a key factor in ensuring their normal operation and preventing safety accidents. Traditional energy facility safety monitoring methods mainly rely on manual inspections and single surveillance camera systems. These methods not only have problems such as low personnel efficiency and limited monitoring range, but also cannot detect and issue early warnings for abnormal events such as sudden fires, smoke leaks, and illegal intrusions, making it difficult to achieve real-time monitoring and timely response to equipment safety.

[0003] In addition, most existing monitoring systems rely on manual analysis of surveillance videos, which has problems such as judgment delays and false alarms. They lack effective support for intelligence and automation and are difficult to adapt to the increasingly complex and sophisticated security needs of modern energy facilities. Existing security monitoring systems usually lack intelligent analysis of video images and comprehensive assessment of abnormal situations. Especially in the detection of special events such as fire smoke and illegal intrusion, there are often problems of poor real-time performance, missed detections and false alarms. Summary of the Invention

[0004] Based on the above objectives, the present invention provides a real-time safety monitoring and risk warning system for energy facilities based on video AI analysis.

[0005] The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis includes a video acquisition module, an image processing module, an AI analysis module, a risk assessment module, and a real-time monitoring and alarm module. Video acquisition module: includes multiple high-definition cameras for real-time image data collection within energy facilities; Image processing module: used to pre-process the image data collected by the video acquisition module, including denoising, enhancement and cropping, to improve image quality; AI analysis module: Uses a single-shot detection algorithm to intelligently analyze pre-processed images and identify anomalies in the images, including fire smoke and illegal intrusion; Risk Assessment Module: This module is used to assess the risks of abnormal situations identified by the AI ​​analysis module, generate risk levels, and provide corresponding early warning information based on the type, severity, and urgency of the risk. Real-time monitoring alarm module: used to receive early warning information from the risk assessment module and send early warning signals to relevant personnel.

[0006] Optionally, the video acquisition module includes a camera deployment unit, a video segmentation unit, and an image data transmission unit; Camera Deployment Unit: Used to securely install multiple high-definition cameras at designated locations within energy facilities, including facility entrances, control rooms, areas surrounding critical equipment, and high-risk areas. Brackets or wall mounting ensures camera stability and viewing angle coverage. Video segmentation unit: used to segment the continuous video stream captured by the camera according to the frame rate of 30 frames per second to form single-frame image data; Image data transmission unit: used to transmit the segmented image data to the image processing module in real time via a wired or wireless network.

[0007] Optionally, the image processing module includes a denoising unit, an image enhancement unit, and an image cropping unit; Denoising unit: uses the median filter algorithm to denoise the image data. The median filter algorithm removes random noise by calculating the median of the neighboring pixels around each pixel in the image, thereby effectively preserving the edge information of the image. Image enhancement unit: used to perform contrast enhancement and brightness adjustment on the denoised image data. It uses the histogram equalization method to adjust the grayscale distribution of the image to enhance the image's detail expression and recognition. Image cropping unit: used to crop the enhanced image according to the preset monitoring area coordinates, use the region selection algorithm to remove irrelevant background parts in the image, and focus on the target monitoring area to improve the processing efficiency of the subsequent AI analysis module.

[0008] Optionally, the image cropping unit includes: Target area detection: Use the Canny edge detection algorithm to calculate the change in image gradient to identify the outline of the object in the image; Region candidate frame generation: Based on target region detection, a contour extraction algorithm is used to generate a candidate frame containing the monitoring area; Region cropping: Based on the generated candidate frame coordinates, the image is cropped using a rectangular frame to obtain the target monitoring area; Background removal: After obtaining the target area, the Otsu binarization algorithm is used to remove the background part.

[0009] Optionally, the AI ​​analysis module includes a feature extraction unit, a target recognition unit, and an anomaly recognition unit; Feature extraction unit: used to extract features from preprocessed image data, including edges, textures, and shapes, to support subsequent target recognition; Target recognition unit: processes the extracted features based on the single-shot detection algorithm and searches for potential anomalies in the image using sliding window technology; Abnormal identification unit: Based on the abnormal situation results of the target identification unit's search, the identified fire smoke and illegal intrusion are classified and the corresponding abnormal category labels are generated.

[0010] Optionally, the feature extraction unit includes: Edge extraction: Use the Sobel operator to detect the edge of the preprocessed image, and detect the image edge by calculating the gradient value of each pixel in the image; Texture feature extraction: Gray level co-occurrence matrix method is used to extract texture features in images; Shape feature extraction: Hough transform is used to extract shape features in the image, including straight lines and circles.

[0011] Optionally, the target recognition unit includes: Sliding window: Scan the pre-processed image through the sliding window technology to determine the potential abnormal area in the image. Set the window size to , the step size is , then the process expression of the window sliding on the image is: ,in, is the image area covered by the sliding window, and is the position of the upper left corner of the window, step size Used to control the stride of window movement; Feature extraction and candidate box generation: When the sliding window traverses the image, feature matching is performed on each window based on the features extracted by the feature extraction unit, and candidate boxes are generated; Candidate box transfer: The generated candidate box is used as the potential location of the abnormal area and is passed to the subsequent abnormality recognition unit for classification to identify the specific abnormality type.

[0012] Optionally, the abnormality identification unit includes: Feature classification: According to the candidate box generated by the target recognition unit, the feature vector of the candidate area is extracted, and the pre-trained support vector machine model is used to classify the feature vector. Let the feature vector be , then the classification function expression of the support vector machine model is: ,in, is the classification result, with a value of +1 or , is the weight vector of the model, is the bias term; Probability assessment: Perform probability assessment on the classification results and calculate the probability value of the candidate area belonging to fire smoke or illegal intrusion. The calculation formula is: ,in, represents the probability that the candidate area belongs to fire smoke or illegal intrusion, is the weight vector of the logistic regression model, is the bias term; Abnormal category label generation: Generate corresponding abnormal category labels based on probability evaluation results , the specific formula is: ; in, and are the thresholds for fire smoke and trespass, respectively.

[0013] Optionally, the risk assessment module includes a risk level determination unit, a risk classification unit, and an early warning information generation unit; wherein: Risk level determination unit: used to calculate the risk score of each abnormal event based on the abnormal situation identified by the AI ​​analysis module. The formula is: ,in, represents the risk score, Indicates the type weight of abnormal events, Indicates the severity weight of the abnormal event, Indicates the urgency weight of the abnormal event, is the corresponding weight coefficient, and ; Risk classification unit: used to classify abnormal events into different risk levels based on risk scores. The risk classification unit uses a preset risk level threshold for classification. The specific classification formula is: , in, Indicates the risk level, and is the preset risk level threshold; Warning information generation unit: used to generate warning information based on risk level Generate specific warning content. The specific generation rules include: when When the risk is low, generate early warning information for continuous monitoring; when When the risk is medium, an early warning message is generated to pay attention; when When the risk is high, generate immediate response warning information.

[0014] Optionally, the real-time monitoring alarm module includes a display unit and a notification unit; wherein: Display unit: used to display the warning information generated by the risk assessment module on the display screen of the monitoring center in real time, ensuring that relevant personnel can intuitively view the risk level and warning content; Notification unit: used to send warning information to relevant personnel through various communication methods, including SMS, email or mobile application notification.

[0015] Beneficial effects of the present invention: This invention uses a high-definition camera to capture images in real time and adopts denoising, enhancement and cropping technologies to improve image quality, ensuring that the AI ​​analysis module can accurately identify abnormal situations such as fire smoke and illegal intrusion, significantly improving the monitoring response speed and detection accuracy.

[0016] The present invention can comprehensively analyze the type, severity and urgency of anomalies through the risk assessment module to generate accurate risk levels, and send early warning information to relevant personnel in a timely manner through the real-time monitoring alarm module, which effectively reduces false alarms and missed alarms and enhances the safety management capabilities of energy facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 Schematic diagram of a security monitoring and risk warning system according to an embodiment of the present invention; Figure 2 Schematic diagram of the AI ​​analysis module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0020] It should be noted that references in the specification to "one embodiment," "an embodiment," "exemplary embodiments," "some embodiments," etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not necessarily every embodiment will include such specific features, structures, or characteristics. Furthermore, when specific features, structures, or characteristics are described in conjunction with an embodiment, it is within the knowledge of persons skilled in the relevant art to implement such features, structures, or characteristics in conjunction with other embodiments (whether or not explicitly described).

[0021] In general, terms can be understood, at least in part, from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending at least in part on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0022] like Figure 1-Figure 2 As shown in the figure, the real-time safety monitoring and risk warning system for energy facilities based on video AI analysis includes a video acquisition module, an image processing module, an AI analysis module, a risk assessment module, and a real-time monitoring and alarm module; among which: Video acquisition module: includes multiple high-definition cameras for real-time image data collection within energy facilities; Image processing module: used to pre-process the image data collected by the video acquisition module, including denoising, enhancement and cropping, to improve image quality; AI Analysis Module: This module uses the Single Shot Detection (SSD) algorithm to intelligently analyze pre-processed images and identify anomalies in the images, including fire smoke and illegal intrusion. Risk Assessment Module: This module is used to assess the risks of abnormal situations identified by the AI ​​analysis module, generate risk levels, and provide corresponding early warning information based on the type, severity, and urgency of the risk. Real-time monitoring alarm module: used to receive early warning information from the risk assessment module and send early warning signals to relevant personnel.

[0023] The video acquisition module includes a camera deployment unit, a video segmentation unit, and an image data transmission unit; Camera Deployment Unit: Used to securely install multiple high-definition cameras at designated locations within energy facilities, including facility entrances, control rooms, areas surrounding critical equipment, and high-risk areas. Brackets or wall mounting ensures camera stability and viewing angle coverage. Video segmentation unit: used to segment the continuous video stream captured by the camera at a frame rate of 30 frames per second to form single-frame image data, ensuring the timing and integrity of the image data; Image data transmission unit: used to transmit the segmented image data to the image processing module in real time via a wired or wireless network to ensure the efficiency and stability of data transmission; through the set camera deployment unit, video segmentation unit and image data transmission unit, high-definition image data can be obtained from key areas within the energy facility in real time and stably, providing a basis for subsequent analysis.

[0024] The image processing module includes a denoising unit, an image enhancement unit, and an image cropping unit; Denoising unit: uses the median filter algorithm to denoise the image data. The median filter algorithm removes random noise by calculating the median of the neighboring pixels around each pixel in the image, thereby effectively preserving the edge information of the image and improving the image clarity. Image enhancement unit: used to perform contrast enhancement and brightness adjustment on the denoised image data. It uses the histogram equalization method to adjust the grayscale distribution of the image to make the low-contrast areas in the image more obvious, thereby enhancing the image's detail expression and recognition. Image cropping unit: used to crop the enhanced image according to the preset monitoring area coordinates, use the region selection algorithm to remove irrelevant background parts in the image, focus on the target monitoring area, and improve the processing efficiency and accuracy of the subsequent AI analysis module; through the set denoising unit, image enhancement unit and image cropping unit, the quality of image data collected by the video acquisition module can be efficiently improved, which helps to quickly and accurately identify abnormal situations such as fire smoke and illegal intrusion.

[0025] The image cropping unit includes: Target area detection: The Canny edge detection algorithm is used to calculate the change in image gradient to identify the outline of the object in the image; the calculation formula of the Canny edge detection algorithm is: ,in, is the gradient value of the image, and The images are and The first-order derivative in the direction represents the edge strength of the image; by calculating the image gradient, the edge and contour area of ​​the object are determined; Region candidate frame generation: Based on the target area detection, a contour extraction algorithm (such as the findContours function in OpenCV) is used to generate a candidate frame containing the monitoring area; Region cropping: Based on the generated candidate frame coordinates, the image is cropped using a rectangular frame to obtain the target monitoring area; the cropping formula is: ,in, is the cropped image, is the original image, and The displacement value of the cropping frame determines the coordinates of the upper left corner and lower right corner of the cropping area; Background removal: After obtaining the target area, the Otsu binarization algorithm is used to remove the background part. The formula is: ,in, is the between-class variance, and is the number of pixels of the two classes, and is the mean of the two classes, The system uses the total number of pixels in the image, calculates the inter-class variance, and selects the optimal threshold to distinguish the foreground and background, thereby effectively removing irrelevant background parts. By using edge detection, contour extraction, and region cropping techniques in the image cropping unit, the system can accurately identify and extract key monitoring areas in energy facilities. At the same time, the Otsu binarization algorithm is used to effectively remove background information, ensuring that the cropped image is concentrated on the area related to the monitoring target. This process improves image processing efficiency.

[0026] The AI ​​analysis module includes a feature extraction unit, a target recognition unit, and an anomaly recognition unit; Feature extraction unit: used to extract features from preprocessed image data, including edges, textures, and shapes, to support subsequent target recognition; Target recognition unit: processes the extracted features based on the single-shot detection algorithm and searches for potential anomalies in the image using sliding window technology; Abnormal identification unit: Based on the abnormal situation results of the target recognition unit's search, the identified fire smoke and illegal intrusion are classified, the corresponding abnormal category labels are generated, and the identification results are passed to the risk assessment module; by setting up the feature extraction unit, target recognition unit and abnormal identification unit, it can efficiently and accurately analyze the pre-processed image data and identify abnormal situations such as fire smoke and illegal intrusion in real time.

[0027] The feature extraction unit includes: Edge extraction: The Sobel operator is used to detect the edge of the preprocessed image. The image edge is detected by calculating the gradient value of each pixel in the image. The calculation formula is: ; ; in, and Respectively represent the Hedi The gradient of the direction, Represents the original image at pixel position The grayscale value at the edge is obtained by calculation: ,in, For the location The edge strength at represents the edge information of the image at that point; Texture feature extraction: The gray level co-occurrence matrix (GLCM) method is used to extract texture features in the image; the co-occurrence matrix between different gray values ​​in the image is calculated, and the formula is: ,in, is the grayscale value in the image and The probability of symbiosis between is the distance between pixels, is the direction angle, Represents the image at pixel location The gray value at is a unit pulse function, which is used to determine whether the gray value matches; by calculating Matrix, can obtain the characteristic parameters of texture, such as contrast, correlation, etc., which are used to describe the texture characteristics of the image; Shape feature extraction: Hough transform is used to extract shape features in the image, including straight lines and circles. The calculation formula of Hough transform is: ,in, is the polar coordinate form of the line, is the angle of the straight line, and is the coordinate of the pixel position in the image; through the calculation of Hough transform, it can identify the straight line or circular shape in the image and extract the geometric shape features in the image; by using the Sobel operator for edge detection, the GLCM method to extract texture features and the Hough transform to identify shape features in the feature extraction unit, the system can comprehensively extract key information in the image, including edges, textures and shapes. These features can effectively characterize the appearance characteristics of the target object and provide accurate data support for subsequent target recognition and anomaly detection.

[0028] The target recognition unit includes: Sliding window: Scan the pre-processed image through the sliding window technology to determine the potential abnormal area in the image. Set the window size to , the step size is , then the process expression of the window sliding on the image is: ,in , ,in, is the image area covered by the sliding window, and is the position of the upper left corner of the window, step size Used to control the stride of window movement; Feature extraction and candidate box generation: When the sliding window traverses the image, feature matching is performed on each window based on the features extracted by the feature extraction unit (such as edges, textures, and shapes), and candidate boxes are generated; Candidate box transfer: The generated candidate boxes are used as potential locations of abnormal areas and are passed to the subsequent abnormality recognition unit for classification to identify specific abnormality types (such as fire smoke or illegal intrusion). Through the above steps, potential abnormal areas can be efficiently located in the image and passed as candidate boxes to the subsequent abnormality recognition unit for accurate classification and judgment.

[0029] The anomaly recognition unit includes: Feature classification: According to the candidate box generated by the target recognition unit, the feature vector of the candidate area is extracted, and the pre-trained support vector machine model is used to classify the feature vector. Let the feature vector be , then the classification function expression of the support vector machine model is: ,in, is the classification result, with a value of +1 or , is the weight vector of the model, is the bias term; through this formula, it can be determined whether the candidate area belongs to fire smoke or illegal intrusion; Probability assessment: Perform probability assessment on the classification results and calculate the probability value of the candidate area belonging to fire smoke or illegal intrusion. The calculation formula is: ,in, represents the probability that the candidate area belongs to fire smoke or illegal intrusion, is the weight vector of the logistic regression model, is a bias term; this step ensures the confidence assessment of the classification results and provides a basis for subsequent decision-making; Abnormal category label generation: Generate corresponding abnormal category labels based on probability evaluation results , the specific formula is: ; in, and The thresholds for fire smoke and illegal intrusion are set respectively to ensure accurate distinction between different anomaly categories; by setting feature classification, probability assessment and anomaly category label generation steps in the anomaly recognition unit, the candidate areas generated by the target recognition unit can be efficiently and accurately classified and confirmed, and anomalies such as fire smoke and illegal intrusion can be accurately identified.

[0030] The risk assessment module includes a risk level determination unit, a risk classification unit, and an early warning information generation unit; among which: Risk level determination unit: used to calculate the risk score of each abnormal event based on the abnormal situation identified by the AI ​​analysis module. The formula is: ,in, represents the risk score, Indicates the type weight of abnormal events, Indicates the severity weight of the abnormal event, Indicates the urgency weight of the abnormal event, is the corresponding weight coefficient, and Through this formula, the risk level determination unit comprehensively considers the type, severity and urgency of the abnormal event and calculates the risk score of each abnormal event; Risk classification unit: used to classify abnormal events into different risk levels based on risk scores. The risk classification unit uses a preset risk level threshold for classification. The specific classification formula is: , in, Indicates the risk level, and is the preset risk level threshold. Through this formula, the risk classification unit will score the risk of each abnormal event Classify as low risk, medium risk or high risk to ensure accurate distinction between different risk levels; Warning information generation unit: used to generate warning information based on risk level Generate specific warning content. The specific generation rules include: when When the risk is low, generate early warning information for continuous monitoring; when When the risk is medium, an early warning message is generated to pay attention; when When the risk is high, an immediate response warning information is generated; through the above-mentioned risk level determination unit, risk classification unit and warning information generation unit, the abnormal situations identified by the AI ​​analysis module can be systematically comprehensively evaluated and classified to generate accurate risk levels and corresponding warning information.

[0031] The real-time monitoring alarm module includes a display unit and a notification unit; wherein: Display unit: used to display the warning information generated by the risk assessment module on the display screen of the monitoring center in real time, ensuring that relevant personnel can intuitively view the risk level and warning content; Notification unit: used to send early warning information to relevant personnel through various communication methods, including SMS, email or mobile application notifications; through the above-mentioned display unit and notification unit, the early warning information generated by the risk assessment module can be effectively and intuitively displayed to relevant personnel, and notifications can be sent in a timely manner through various communication methods to ensure that personnel can respond quickly.

[0032] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0033] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. Real-time safety monitoring and risk warning system for energy facilities based on video AI analysis, characterized by: It includes video acquisition module, image processing module, AI analysis module, risk assessment module and real-time monitoring and alarm module; among them: Video acquisition module: includes multiple high-definition cameras for real-time image data collection within energy facilities; Image processing module: used to pre-process the image data collected by the video acquisition module, including denoising, enhancement and cropping, to improve image quality; AI analysis module: Uses a single-shot detection algorithm to intelligently analyze pre-processed images and identify anomalies in the images, including fire smoke and illegal intrusion; Risk Assessment Module: This module is used to assess the risks of abnormal situations identified by the AI ​​analysis module, generate risk levels, and provide corresponding early warning information based on the type, severity, and urgency of the risk. Real-time monitoring alarm module: used to receive early warning information from the risk assessment module and send early warning signals to relevant personnel.

2. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 1 is characterized in that: The video acquisition module includes a camera deployment unit, a video segmentation unit and an image data transmission unit; Camera Deployment Unit: Used to securely install multiple high-definition cameras at designated locations within energy facilities, including facility entrances, control rooms, areas surrounding critical equipment, and high-risk areas. Brackets or wall mounting ensures camera stability and viewing angle coverage. Video segmentation unit: used to segment the continuous video stream captured by the camera according to the frame rate of 30 frames per second to form single-frame image data; Image data transmission unit: used to transmit the segmented image data to the image processing module in real time via a wired or wireless network.

3. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 1 is characterized in that: The image processing module includes a denoising unit, an image enhancement unit and an image cropping unit; Denoising unit: uses the median filter algorithm to denoise the image data. The median filter algorithm removes random noise by calculating the median of the neighboring pixels around each pixel in the image, thereby effectively preserving the edge information of the image. Image enhancement unit: used to perform contrast enhancement and brightness adjustment on the denoised image data. It uses the histogram equalization method to adjust the grayscale distribution of the image to enhance the image's detail expression and recognition. Image cropping unit: used to crop the enhanced image according to the preset monitoring area coordinates, use the region selection algorithm to remove irrelevant background parts in the image, and focus on the target monitoring area to improve the processing efficiency of the subsequent AI analysis module.

4. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 3 is characterized in that: The image cropping unit includes: Target area detection: Use the Canny edge detection algorithm to calculate the change in image gradient to identify the outline of the object in the image; Region candidate frame generation: Based on target region detection, a contour extraction algorithm is used to generate a candidate frame containing the monitoring area; Region cropping: Based on the generated candidate frame coordinates, the image is cropped using a rectangular frame to obtain the target monitoring area; Background removal: After obtaining the target area, the Otsu binarization algorithm is used to remove the background part.

5. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 1 is characterized in that: The AI ​​analysis module includes a feature extraction unit, a target recognition unit, and an anomaly recognition unit; Feature extraction unit: used to extract features from preprocessed image data, including edges, textures, and shapes, to support subsequent target recognition; Target recognition unit: processes the extracted features based on the single-shot detection algorithm and searches for potential anomalies in the image using sliding window technology; Abnormal identification unit: Based on the abnormal situation results of the target identification unit's search, the identified fire smoke and illegal intrusion are classified and the corresponding abnormal category labels are generated.

6. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 5 is characterized in that: The feature extraction unit includes: Edge extraction: Use the Sobel operator to detect the edge of the preprocessed image, and detect the image edge by calculating the gradient value of each pixel in the image; Texture feature extraction: Gray level co-occurrence matrix method is used to extract texture features in images; Shape feature extraction: Hough transform is used to extract shape features in the image, including straight lines and circles.

7. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 6 is characterized in that: The target recognition unit includes: Sliding window: Scan the pre-processed image through the sliding window technology to determine the potential abnormal area in the image. Set the window size to , the step size is , then the process expression of the window sliding on the image is: ,in, is the image area covered by the sliding window, and is the position of the upper left corner of the window, step size Used to control the stride of window movement; Feature extraction and candidate box generation: When the sliding window traverses the image, feature matching is performed on each window based on the features extracted by the feature extraction unit, and candidate boxes are generated; Candidate box transfer: The generated candidate box is used as the potential location of the abnormal area and is passed to the subsequent abnormality recognition unit for classification to identify the specific abnormality type.

8. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 7 is characterized in that: The abnormality identification unit includes: Feature classification: According to the candidate box generated by the target recognition unit, the feature vector of the candidate area is extracted, and the pre-trained support vector machine model is used to classify the feature vector. Let the feature vector be , then the classification function expression of the support vector machine model is: ,in, is the classification result, with a value of +1 or , is the weight vector of the model, is the bias term; Probability assessment: Perform probability assessment on the classification results and calculate the probability value of the candidate area belonging to fire smoke or illegal intrusion. The calculation formula is: ,in, represents the probability that the candidate area belongs to fire smoke or illegal intrusion, is the weight vector of the logistic regression model, is the bias term; Abnormal category label generation: Generate corresponding abnormal category labels based on probability evaluation results , the specific formula is: ； in, and are the thresholds for fire smoke and trespass, respectively.

9. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 1 is characterized in that: The risk assessment module includes a risk level determination unit, a risk classification unit, and an early warning information generation unit; wherein: Risk level determination unit: used to calculate the risk score of each abnormal event based on the abnormal situation identified by the AI ​​analysis module. The formula is: ,in, represents the risk score, Indicates the type weight of abnormal events, Indicates the severity weight of the abnormal event, Indicates the urgency weight of the abnormal event, is the corresponding weight coefficient, and ; Risk classification unit: used to classify abnormal events into different risk levels based on risk scores. The risk classification unit uses a preset risk level threshold for classification. The specific classification formula is: ， in, Indicates the risk level, and is the preset risk level threshold; Warning information generation unit: used to generate warning information based on risk level Generate specific warning content. The specific generation rules include: when When the risk is low, generate early warning information for continuous monitoring; when When the risk is medium, an early warning message is generated to pay attention; when When the risk is high, generate immediate response warning information.

10. The real-time safety monitoring and risk warning system for energy facilities based on video AI analysis according to claim 1 is characterized in that: The real-time monitoring alarm module includes a display unit and a notification unit; wherein: Display unit: used to display the warning information generated by the risk assessment module on the display screen of the monitoring center in real time, ensuring that relevant personnel can intuitively view the risk level and warning content; Notification unit: used to send warning information to relevant personnel through various communication methods, including SMS, email or mobile application notification.

Images