Video monitoring information integration management method for multi-scene security

Abstract

The present application belongs to the technical field of video monitoring processing, and particularly relates to a video monitoring information integrated management method for multi-scene security, comprising: integratedly acquiring video images in a multi-scene security operation area, and constructing multi-dimensional basic risk indexes including distance risk indexes, posture risk indexes and speed risk indexes; constructing an event-driven cooperative amplification model, which is used to generate a cooperative amplifier; calculating a basic risk score based on the multi-dimensional basic risk indexes, and multiplying the basic risk score by the cooperative amplifier to obtain a final dynamic risk score; and integratedly managing and grading early warning of the video monitoring information according to the final dynamic risk score. The present application realizes accurate identification and amplification of cooperative risks, and significantly improves the timeliness and accuracy of security information management.

Description

Technical Field

[0001] This invention relates to the field of video surveillance processing technology. More specifically, this invention relates to a method for integrated management of video surveillance information for multi-scenario security. Background Technology

[0002] With the development of smart cities and IoT technologies, video surveillance systems play a crucial role in security across various scenarios, such as public safety monitoring in smart communities and production safety management in industrial parks. These systems generate massive amounts of video data, and effectively integrating and managing this video surveillance information is key to improving security efficiency.

[0003] Currently, mainstream security platforms tend to integrate multi-source sensing data and use artificial intelligence technology for analysis. In terms of risk assessment, a common technical solution is to use target detection and human key point detection technology to monitor and analyze the behavior of workers or targets in real time, calculate multi-dimensional risk indicators, and finally use a linear weighted model for comprehensive evaluation.

[0004] However, this risk fusion method, which uses fixed weights for linear superposition, has significant technical flaws. The problem lies in its tendency to mistakenly treat individual risk factors as independent, failing to capture the synergistic amplification effect of multiple concurrent hazards in the real world. For example, when two events—being too close to the target and a dangerous posture—occur simultaneously, the combined risk far exceeds the sum of their individual risks. Existing linear models are slow to react to such situations, resulting in limited increases in calculated risk scores. This often leads to missed detections at critical moments, significantly compromising the reliability of the entire security information management system. Summary of the Invention

[0005] To address the technical problem that existing technologies cannot effectively assess the collaborative risks of multiple concurrent hazard factors, this invention provides a video surveillance information integration and management method for multi-scenario security. The method includes: integrating and acquiring video images within multiple security operation areas, and constructing multi-dimensional basic risk indicators, including at least distance risk indicators, posture risk indicators, and speed risk indicators; constructing an event-driven collaborative amplification model, which generates a collaborative amplifier, wherein the product of the distance risk indicator and the posture risk indicator is used as a hazard event score, and the risk is non-linearly amplified based on the magnitude of the hazard event score to generate the collaborative amplifier; calculating a basic risk score based on the multi-dimensional basic risk indicators, and multiplying the basic risk score by the collaborative amplifier to obtain a final dynamic risk score; and performing integrated management and hierarchical early warning of video surveillance information based on the final dynamic risk score, thereby dynamically assessing the collaborative risks of multiple concurrent hazard factors.

[0006] This invention constructs an event-driven collaborative amplification model that can capture collaborative events involving multiple key risk factors, such as close proximity and dangerous posture, occurring simultaneously. The amplifier is activated only when such collaborative events occur, drastically and non-linearly amplifying the underlying risk, resulting in a decisive abrupt change in the final risk score. This enables the system to accurately filter and manage truly high-risk events from massive amounts of video surveillance information, ensuring timely and accurate early warnings of collaborative risks and significantly improving the effectiveness of information integration management.

[0007] Preferably, the process of obtaining the multi-dimensional basic risk indicators includes: calculating the shortest distance from key points on the human body to the boundary box of the hazard source at time t. and through relational expressions Calculate distance risk indicators ,in Use a constant distance as a reference; count the number of key human figures entering the boundary box of the hazard source at time t. and through relational expressions Calculate attitude risk index ,in The total number of key points on the human body; calculate the rate of change of the shortest distance over time, and use the relevant formula. Calculation speed risk index ,in For time intervals, For reference speed.

[0008] Preferably, the synergistic amplification model satisfies the following relationship: ;in, For the coordinated amplifier at time t, Here, C represents the score for dangerous events, and C represents the maximum magnification factor. The steepness coefficient of the activation function. This is the event activation critical point.

[0009] This allows the co-amplifier to maintain a stable value around 1 when no dangerous event occurs, without amplifying the signal and avoiding false alarms. However, at the moment a dangerous event occurs, its value can rapidly jump to the preset maximum amplification factor, achieving a response mode from stable to rapid increase in risk assessment.

[0010] Preferably, the scores of hazardous events satisfy the following relationship: ;in, As a distance risk indicator, This is an indicator of attitude risk.

[0011] By defining the hazard score by multiplying distance risk and attitude risk, it is ensured that the event score will be significantly greater than zero only when both risk indicators have high values, thus guaranteeing the relevance and accuracy of the amplification mechanism.

[0012] Preferably, the basic risk score satisfies the following relationship: ;in, It is the basic risk score. , , These are the preset base weights.

[0013] Preferably, the final dynamic risk score satisfies the following relationship: ;in, Based on the basic risk score, It is a synergistic amplifier.

[0014] Preferably, the human body key points and hazard source bounding boxes are obtained by processing video images using a target detection model and a human pose estimation model.

[0015] This invention utilizes a computational synergistic amplifier to exponentially amplify risks when both proximity and dangerous posture conditions are met simultaneously. This significantly improves the accuracy of identifying high-risk events and effectively avoids missed detections.

[0016] Furthermore, when personnel approach the hazard source normally or engage in large-scale physical activities at a safe distance, the amplifier remains inactive, and the risk assessment value remains consistent with the standard basic risk value, preventing false alarms. This resolves the logical flaws in existing models that may arise from oversensitivity to a single indicator, thereby improving the system's usability and the trust of on-site personnel in the system. Attached Figure Description

[0017] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0018] Figure 1 This is a flowchart illustrating a video surveillance information integration and management method for multi-scenario security according to the present invention;

[0019] Figure 2 This is a schematic diagram illustrating the response of posture risk indicators to dangerous interaction scenarios;

[0020] Figure 3 This is a schematic diagram illustrating the basic risk score and the synergistic amplifier response to dangerous interaction scenarios;

[0021] Figure 4 This is a schematic diagram illustrating the interaction scenarios of prior art and the present invention in response to dangerous situations. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0024] This invention discloses a method for integrated management of video surveillance information for multi-scenario security, referring to... Figure 1 This includes steps S1-S4:

[0025] S1. Integrate and acquire video images within the security operation area of ​​multiple scenarios, and construct multi-dimensional basic risk indicators, including at least distance risk indicators, posture risk indicators, and speed risk indicators.

[0026] In an optional embodiment, a real-time video stream can be acquired by a monitoring camera deployed at the work site. For each frame of the video stream, a target detection model, such as the YOLO series model and a human pose estimation model, such as a key point detection model, can be used to identify and locate the coordinates of all human key points of the person in the image and the bounding box of the hazard source. The hazard source can be large machinery, high-temperature equipment or high-pressure devices in the production workshop, or electronic fences, restricted areas, etc. in community security.

[0027] Furthermore, based on the obtained positioning and coordinate information, multi-dimensional basic risk indicators, including risk indicators, attitude risk indicators, and velocity risk indicators at the current time t, can be calculated.

[0028] Specifically, distance risk indicators Satisfying the relation:

[0029]

[0030] Among them, the distance risk index is used to measure the proximity of workers to hazards. This represents the shortest distance from a key point on the human body to the boundary box of the hazard source at time t. This serves as a reference distance constant, used to adjust the sensitivity to distance-related risks. When personnel are very far from the hazard source... It's very big. Approaching 0; when personnel come into contact with the hazard source, , It equals 1.

[0031] Posture risk indicators Satisfying the relation:

[0032]

[0033] in, Let be the number of key human figures entering the boundary box of the hazard source at time t. This represents the total number of human body keypoints detected by the human posture estimation model. The posture risk index is used to measure the degree to which a worker's body parts intrude into a hazardous area. When no part of the worker's body enters the hazardous area, the risk level is considered low. , .

[0034] Speed ​​risk indicators Satisfying the relation:

[0035]

[0036] in, The time interval between two frames. As a reference speed, the speed risk index is used to measure the approach speed of workers towards a hazard source, by calculating the shortest safe distance. The rate of change over time is obtained and normalized using a reference velocity. The function ensures that the speed risk is positive only when approaching a hazard source.

[0037] like Figure 2 The diagram illustrates the response of the posture risk index to a dangerous interactive scenario. In time steps 0-50, the worker approaches the hazard source; in time steps 51-70, the worker stays near the hazard source. In particular, in time steps 60-70, the worker suddenly adopts a brief dangerous posture, causing a pulse in the posture risk. Finally, in time steps 71-100, the worker moves away, and the posture risk index returns to its initial level.

[0038] In this way, by analyzing video images from multiple scenarios in real time, we can accurately obtain information on multiple basic dimensions reflecting the on-site safety situation, providing reliable data input for subsequent risk assessment.

[0039] S2. Construct an event-driven collaborative amplification model to generate a collaborative amplifier. The model uses the product of the distance risk index and the attitude risk index as the danger event score, and performs nonlinear amplification of the risk based on the magnitude of the danger event score to generate the collaborative amplifier.

[0040] In an alternative embodiment, to address the problem that existing technologies cannot respond to collaborative risks, an event-driven collaborative amplification model can be constructed to generate a collaborative amplifier that is only activated when a dangerous event occurs, such as when a person makes a dangerous gesture at close range, and produces a huge amplification effect.

[0041] Specifically, the product of the distance risk index and the attitude risk index can be used as the hazard event score. The scores of dangerous events satisfy the following relation:

[0042]

[0043] in, As a distance risk indicator, This serves as a posture risk indicator. This ensures that only when... and When it has a high value, Only then will the value be significantly greater than zero, thus enabling the accurate identification of collaborative risk events.

[0044] Furthermore, a soft switch controlled by the fraction of dangerous events can be constructed using the Sigmoid function to generate the final synergistic amplifier. The synergistic amplifier model satisfies the following relationship:

[0045]

[0046] in, For the coordinated amplifier at time t, C represents the percentage of hazardous events, and C is the maximum magnification factor used to determine the magnification intensity. This is the steepness coefficient of the activation function; the larger this value, the more pronounced the switching effect. This is the event activation threshold, used to set the threshold that triggers amplification. For example, the maximum amplification factor C is set to 15, which is the event activation threshold. The value is 0.1, which is the steepness coefficient. The value is 50.

[0047] Under normal circumstances, such as when a person is approaching a hazard but not in a dangerous posture, the hazard event score is strictly 0. Since 0 is less than the event activation threshold, the Sigmoid function output is close to 0, and the amplifier... The value will remain around 1, having virtually no amplifying effect; however, once a person makes a dangerous gesture at close range, the danger event score will instantly exceed 0 and surpass the event activation threshold, causing the Sigmoid function output to rapidly approach 1, thus amplifying the signal. The value jumped from 1 to This produces a dramatic amplification effect.

[0048] In this way, by constructing a nonlinear synergistic amplifier, it is possible to ensure that risk assessment only produces a dramatic response when it is most needed, thus achieving precise capture of synergistic risks.

[0049] S3. Calculate the basic risk score based on multi-dimensional basic risk indicators, and multiply the basic risk score by the synergistic amplifier to obtain the final dynamic risk score.

[0050] In an optional embodiment, a basic risk score can be calculated based on the obtained multi-dimensional basic risk indicators, and the basic risk score satisfies the following relationship:

[0051]

[0052] in, It is the basic risk score. , , These are preset base weights. For example, , , The values ​​are 0.4, 0.4, and 0.2.

[0053] Furthermore, by combining the obtained basic risk score with the synergistic amplifier, a final dynamic risk score that accurately reflects the synergistic risk is obtained. The final dynamic risk score satisfies the following relationship:

[0054]

[0055] in, Based on the basic risk score, As a synergistic amplifier, through The saturation treatment of the function ensures that the upper limit of the final dynamic risk score is 1.

[0056] like Figure 3 The diagram illustrates the basic risk score and the collaborative amplifier's response to a dangerous interaction scenario. At time steps 60-70, the amplifier is instantly activated the moment the dangerous posture appears, its value drastically increasing to its peak value, while the basic risk score... Although the orange dotted line also rises, the increase is gradual and the response is slow.

[0057] Thus, by multiplying the basic risk by the event amplifier, a final risk score that truly reflects the degree of collaborative danger on site is obtained, providing a basis for decision-making in integrated management and reliable early warning.

[0058] S4. Based on the final dynamic risk score, integrate and manage video surveillance information and issue graded early warnings, thereby dynamically assessing the collaborative risks when multiple risk factors occur concurrently.

[0059] In an optional embodiment, the obtained final dynamic risk score can be compared with a preset risk threshold to trigger a warning of the corresponding level. For example, the risk threshold can be set to 0.8. When the value exceeds 0.8, the system immediately activates the audible and visual alarm device, sending a high-risk alert to on-site personnel and the remote monitoring center.

[0060] like Figure 4 The diagram illustrates the response of the prior art and the present invention to dangerous interactive scenarios. As can be seen, the risk score of the prior art is always far below the danger threshold of 0.8 throughout the process, resulting in false negatives. However, the final risk score of the present invention, at time step 60, surges to a saturation value of 1.0 almost instantaneously due to the amplifier, and during this period, it stably outputs the highest level of danger signal, thus successfully providing an early warning.

[0061] In this way, by setting thresholds for the final scores that reflect collaborative risks, intelligent and differentiated integrated management of video surveillance information is achieved, enabling timely early warnings and effectively preventing security incidents caused by missed judgments.

[0062] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise expressly and specifically defined.

[0063] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method for integrated management of video surveillance information for multi-scenario security, characterized in that, include: The system integrates and acquires video images from multiple security operation areas and constructs multi-dimensional basic risk indicators, including at least distance risk indicators, posture risk indicators, and speed risk indicators. An event-driven collaborative amplification model is constructed to generate a collaborative amplifier, wherein the product of the distance risk index and the attitude risk index is used as the danger event score, and the risk is nonlinearly amplified according to the magnitude of the danger event score to generate the collaborative amplifier. The basic risk score is calculated based on the multi-dimensional basic risk indicators, and the basic risk score is multiplied by the synergistic amplifier to obtain the final dynamic risk score. Based on the final dynamic risk score, video surveillance information is integrated, managed, and tiered for early warning, thereby dynamically assessing the collaborative risks when multiple risk factors occur concurrently.

2. The video surveillance information integration and management method for multi-scenario security as described in claim 1, characterized in that, The process of obtaining the multi-dimensional basic risk indicators includes: Calculate the shortest distance from key points on the human body to the boundary box of the hazard source at time t. and through relational expressions Calculate distance risk indicators ,in Used as a reference distance constant; Count the number of key human figures entering the boundary box of the hazard source at time t. and through relational expressions Calculate attitude risk index ,in The total number of key points in the human body; Calculate the rate of change of the shortest distance over time, and use the relationship... Calculation speed risk index ,in For time intervals, For reference speed.

3. The video surveillance information integration and management method for multi-scenario security as described in claim 1, characterized in that, The synergistic amplification model satisfies the following relationship: ； in, For the coordinated amplifier at time t, Here, C represents the score for dangerous events, and C represents the maximum magnification factor. The steepness coefficient of the activation function. This is the event activation critical point.

4. The video surveillance information integration and management method for multi-scenario security as described in claim 3, characterized in that, The scores of dangerous events satisfy the following relationship: ； in, The score represents the number of dangerous events. As a distance risk indicator, This is an indicator of attitude risk.

5. A video surveillance information integration and management method for multi-scenario security as described in claim 1 or 2, characterized in that, The basic risk score satisfies the following relationship: ； in, It is the basic risk score. As a distance risk indicator, As an indicator of attitude risk, As a speed risk indicator, , , These are the preset base weights.

6. The video surveillance information integration and management method for multi-scenario security as described in claim 1, characterized in that, The final dynamic risk score satisfies the following relationship: ； in, Based on the basic risk score, It is a synergistic amplifier.

7. The video surveillance information integration and management method for multi-scenario security as described in claim 2, characterized in that, The human body key points and hazard source bounding boxes are obtained by processing video images using a target detection model and a human pose estimation model.

Images