Personnel safety intelligent monitoring system

Abstract

The invention relates to the technical field of safety monitoring, in particular to a personnel safety intelligent monitoring system which comprises a millimeter wave monitoring unit and a real-time monitoring module of an image acquisition unit. The label setting module is used for calculating an abnormal characterization value and setting a potential abnormal label; the feature extraction module is used for estimating an image acquisition unit associated with the target and an associated time domain segment; the frame extraction analysis module is used for determining a video segment needing to be extracted; the feature analysis module is used for evaluating the feature prominence of the target in each time domain sub-segment; and the result analysis module is used for determining a time domain sub-segment needing to be analyzed based on an evaluation result of the feature analysis module for feature saliency, determining an action type of a target in the time domain sub-segment, and judging whether the target is abnormal or not according to a historical sample corresponding to the action type. According to the method, the potential anomaly analysis is carried out on the target, the target with characteristic highlight is selected, whether the anomaly exists or not is judged, and the anomaly monitoring efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of security monitoring, and particularly to an intelligent personnel security monitoring system. Background Art

[0002] With the application and development of modern technologies, big data, cloud computing, the Internet of Things, etc. have been gradually introduced into police stations' case handling areas, law enforcement management centers, and detention houses, providing strong technical support for the security monitoring of detainees, helping to improve the safety management level and intelligence degree of detention facilities. The wide application of high-definition cameras and face recognition technologies can achieve all-weather and omnidirectional monitoring and automatically identify abnormal behaviors. Some detention houses are equipped with electronic bracelets or vital sign monitoring mattresses for detainees to monitor vital signs in real time and detect health problems in a timely manner. At the same time, the application of 60GHz millimeter-wave radar technology in detention house monitoring is being gradually promoted, and its characteristics of high precision and all-weather operation give it unique advantages in perimeter protection and indoor behavior monitoring.

[0003] Chinese Patent Publication No.: CN118966588A discloses a smart prison personnel management system, which includes a security monitoring module, a health management module, a dynamic scheduling module, a behavior analysis and evaluation module, an emergency handling module, and a data management module that are communicatively connected in sequence. The security monitoring module is responsible for monitoring the security status of the prison and collecting video data. The health management module monitors and records the health of personnel. The dynamic scheduling module uses algorithms to dynamically adjust personnel allocation and schedules according to behaviors and health conditions. The behavior analysis and evaluation module uses machine learning to evaluate and predict personnel behaviors. The emergency handling module combines intelligent analysis to predict risks and responds quickly in case of emergencies. The data management module processes and stores data, supporting data backup, analysis, and report generation. Through the close combination of multiple modules, the invention improves the security management efficiency of the prison, enhances the emergency handling ability, and realizes the efficient management of the prison environment and personnel.

[0004] Chinese patent publication number: CN111341069A, discloses a smart construction site personnel safety real-time monitoring system, the system includes: at least one smart helmet, access control equipment, electronic fence, reference nodes dispersedly set on the ground of some floors of the construction site, and a background monitoring system; the access control equipment includes linearly arranged antennas and RFID readers; the electronic fence is composed of beacons of various dangerous sources set on the construction site; the reference nodes dispersedly set on the ground of some floors of the construction site are used to provide reference air pressure; the smart helmet is used to locate the height of the wearer, detect the motion posture and abnormal alarm; the background monitoring system is used to display the attendance status, the distribution of the electronic fence, monitor the height and posture of the wearer, notify the management personnel when receiving abnormal alarm information, and issue safety-related instructions to the smart helmet. The embodiment of the invention can effectively reduce the safety hazards of construction sites, improve the safety of the construction site, and improve the management efficiency.

[0005] However, there are still the following problems in the prior art: In the prior art, the monitoring of abnormal persons in detention centers is mostly carried out through a combination of wristbands and surveillance, and the determination of abnormal persons usually involves indiscriminate analysis of targets in the area, which generates a large amount of data and reduces monitoring efficiency and accuracy. Summary of the invention

[0006] To this end, the present invention provides an intelligent personnel safety monitoring system to solve the problem that in the prior art, the monitoring of abnormal personnel in detention centers is mostly carried out through a combination of wristbands and monitoring, and the determination of abnormal personnel usually involves indiscriminate analysis of targets in the area, which generates a large amount of data and reduces monitoring efficiency and accuracy.

[0007] To achieve the above object, the present invention provides a personnel safety intelligent monitoring system, which includes: A real-time monitoring module, including a plurality of millimeter wave monitoring units arranged in each area for monitoring target vital sign data and a plurality of image acquisition units for continuously acquiring regional images; A label setting module, which is connected to the real-time monitoring module, is used to extract the motion characteristics of each target in the regional image, calculate the abnormal characterization value in combination with the corresponding vital sign data, and set a potential abnormal label for each target; A feature extraction module connected to the label setting module is used to extract the spatiotemporal trajectory features of the target in response to the target being set with a potential abnormal label, and to estimate the image acquisition unit and the associated time domain segment associated with the target; A frame extraction analysis module connected to the feature extraction module for retrieving the regional image stored in the image acquisition unit based on the estimated result of the feature extraction module to determine the video segment to be extracted; A feature analysis module, which is connected to the frame extraction analysis module, is used to identify the joint point map of the target with an abnormal label set in the video segment, so as to identify the dynamic features of the target and evaluate the feature saliency of the target in each time-domain sub-segment; A result analysis module, which is connected to the feature analysis module, is used to determine the time-domain sub-segment to be analyzed based on the evaluation result of the feature saliency by the feature analysis module, determine the action type of the target in the time-domain sub-segment, and determine whether the target is abnormal according to the historical samples of the corresponding action type; Among them, the motion feature is the sum average of the target's lateral displacement and the target's longitudinal displacement, and the spatio-temporal trajectory feature includes the target's moving speed and moving direction.

[0008] Further, the label setting module calculates the abnormal characterization value, including, It is used to determine that the ratio of the absolute value of the difference between the motion feature and the reference motion feature to the reference motion feature is the motion feature influence factor; It is used to determine the physical sign data, including the breathing rate and the heart rate; It is used to determine that the ratio of the absolute value of the difference between the breathing rate and the reference breathing rate to the reference breathing rate is the breathing rate influence factor; It is used to determine that the ratio of the absolute value of the difference between the heart rate and the reference heart rate to the reference heart rate is the heart rate influence factor; It is used to determine that the weighted sum value of the motion feature influence factor, the breathing rate influence factor and the heart rate influence factor is the abnormal characterization value.

[0009] Further, the label setting module sets potential abnormal labels for each target, where, If the abnormal characterization value is outside the preset abnormal threshold range, a potential abnormal label is set for the target; If the abnormal characterization value is within the preset abnormal threshold range, no potential abnormal label is set for the target.

[0010] Further, the feature extraction module estimates the image acquisition unit and the associated time-domain segment associated with the target, including, It is used to determine the path area of the target based on the spatio-temporal trajectory feature; It is used to determine the image acquisition unit within the path area; It is used to establish the association relationship between the image acquisition unit and the target; It is used to predict the estimated arrival time to reach the path area based on the moving speed, and determine the associated time-domain segment based on the estimated arrival time.

[0011] Further, the frame extraction analysis module determines the video segment to be extracted, including, To extract the video segment collected by the image acquisition unit associated with the target within the associated time domain segment.

[0012] Further, the joint points of the joint point map of the feature analysis module include head joint points, hand joint points, foot joint points, and chest joint points.

[0013] Further, the feature analysis module identifies the dynamic features of the target, including, To determine the displacement amount of each joint point in the joint point map; To set the total displacement amount as the dynamic feature.

[0014] Further, the feature analysis module evaluates the feature saliency of the target in each time domain sub-segment, where, If the dynamic feature is greater than a predetermined dynamic feature, it is determined that the target has feature saliency in the time domain sub-segment; If the dynamic feature is less than or equal to the predetermined dynamic feature, it is determined that the target does not have feature saliency in the time domain sub-segment.

[0015] Further, the result analysis module determines the time domain sub-segment to be analyzed based on the evaluation result of the feature saliency by the feature analysis module, including, To determine the time domain sub-segment where the target has feature saliency; To determine the time domain sub-segment as the time domain sub-segment to be analyzed.

[0016] Further, the result analysis module determines whether the target is abnormal, including, To extract the regional image collected by the image acquisition unit in the time domain sub-segment; To compare the regional image with the historical sample image to determine the similarity for the target; If the similarity is greater than or equal to the similarity threshold, it is determined that the target is not abnormal; If the similarity is less than the similarity threshold, it is determined that the target is abnormal.

[0017] Compared with the prior art, the present invention provides a real-time monitoring module including a millimeter-wave monitoring unit and an image acquisition unit; a tag setting module that calculates an abnormal characterization value and sets potential abnormal tags; a feature extraction module that estimates the image acquisition unit associated with the target and the associated time domain segment; a frame extraction analysis module that determines the video segment to be extracted; a feature analysis module that evaluates the feature saliency of the target in each time domain sub-segment; and a result analysis module that determines the time domain sub-segment to be analyzed based on the evaluation result of the feature saliency by the feature analysis module, determines the action type of the target in the time domain sub-segment, and determines whether the target is abnormal according to the historical samples corresponding to the action type. By performing potential abnormal analysis on the target, selecting the target with feature saliency, and determining whether there is an abnormality, the present invention improves the efficiency and accuracy of abnormal monitoring.

[0018] In particular, by calculating the abnormal characterization value and setting potential abnormal tags for each target to classify the targets, and analyzing the targets with potential abnormal tags. In actual situations, when monitoring targets, it is usually chosen to perform non-discriminatory analysis on all targets in the area, which will lead to the accumulation of a large amount of data, reducing the efficiency of target abnormal analysis. It not only causes a large waste of computing power but also reduces the analysis accuracy of abnormal targets. In particular, some targets may have small action amplitudes and will be ignored during abnormal analysis, resulting in incomplete and inaccurate abnormal analysis of the targets. Based on this, the present invention calculates the abnormal characterization value at the initial stage of abnormal analysis of the target, sets potential abnormal tags for some targets, and only considers the targets with potential abnormal tags during analysis, reducing the waste of computing power and improving the efficiency and accuracy of abnormal monitoring.

[0019] In particular, the present invention extracts the spatio-temporal trajectory features of the targets with potential abnormal tags. By estimating the image acquisition unit associated with the target and the associated time domain segment, it provides a data basis for subsequent evaluation of the feature saliency of the target. In the prior art, when performing data analysis on the target, in most cases, it is chosen to obtain all the data in the area and then select the data related to the target, wasting computing power and reducing the analysis speed. Based on this, the present invention determines the action state of the target through the spatio-temporal trajectory features of the target to determine several image acquisition units associated with the target and the time to reach several image acquisition units, accurately locates the associated time domain segment of the target, extracts the video segment within the time domain segment to identify the dynamic features of the target, reducing the waste of computing power and improving the efficiency and accuracy of abnormal monitoring.

[0020] In particular, the present invention identifies the dynamic features of the target, evaluates the feature salience of the target in each time-domain sub-segment to determine the time-domain sub-segment that needs to be analyzed, providing a data basis for determining whether the target is abnormal. In actual situations, not every video frame in a video segment needs to be analyzed. The target may remain stationary in a certain time-domain sub-segment. Analyzing this time-domain sub-segment will not only cause waste of computing power and inability to judge the abnormal result, but also reduce the analysis speed. Based on this, the present invention determines the time-domain sub-segment that needs to be analyzed by evaluating the feature salience of the target in each time-domain sub-segment, reducing the waste of computing power and improving the efficiency and accuracy of anomaly monitoring.

[0021] In particular, the present invention determines the action type of the target within the time-domain sub-segment and determines whether the target is abnormal based on the historical samples corresponding to the action type. When the target is in a dynamic behavior, if there is an abnormality, it will surely be different from the historical samples. Based on this, the present invention determines the abnormality of the target by comparing the difference between the action type of the target within the time-domain sub-segment and the historical samples, improving the efficiency and accuracy of anomaly monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a schematic structural diagram of the personnel safety intelligent monitoring system according to an embodiment of the invention; Figure 2 is a logic block diagram for setting potential anomaly labels for each target according to an embodiment of the invention; Figure 3 is a logic block diagram for evaluating the feature salience of the target in each time-domain sub-segment according to an embodiment of the invention; Figure 4 is a logic block diagram for determining whether the target is abnormal according to an embodiment of the invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] In order to make the objectives and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0024] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0025] It should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the term "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0026] Please refer to Figure 1 , Figure 1 which is a schematic structural diagram of the personnel safety intelligent monitoring system according to the embodiment of the invention. The personnel safety intelligent monitoring system according to the embodiment of the present invention includes: A real-time monitoring module, including several millimeter-wave monitoring units arranged in each area for monitoring target physical sign data and several image acquisition units for continuously acquiring area images; A tag setting module, which is connected to the real-time monitoring module, is used to extract the motion features of each target in the area image, calculate the abnormal characterization value in combination with the corresponding physical sign data, and set potential abnormal tags for each target; A feature extraction module, which is connected to the tag setting module, is used to respond to the target being set with a potential abnormal tag, extract the spatio-temporal trajectory features of the target, and estimate the image acquisition unit and the associated time domain segment associated with the target; A frame extraction analysis module, which is connected to the feature extraction module, is used to retrieve the stored area images of the image acquisition unit based on the estimation result of the feature extraction module and determine the required video segment to be extracted; A feature analysis module, which is connected to the frame extraction analysis module, is used to identify the joint point map corresponding to the target with an abnormal tag set in the video segment to identify the dynamic features of the target and evaluate the feature saliency of the target in each time domain sub-segment; A result analysis module, which is connected to the feature analysis module, is used to determine the required time domain sub-segment to be analyzed based on the evaluation result of the feature saliency by the feature analysis module, determine the action type of the target in the time domain sub-segment, and determine whether the target is abnormal according to the historical samples of the corresponding action type; Wherein, the motion feature is the sum mean of the target's lateral displacement and the target's longitudinal displacement, and the spatio-temporal trajectory features include the moving speed and moving direction of the target.

[0027] Specifically, the specific device of the millimeter-wave monitoring unit is not limited. For example, in this embodiment, a millimeter-wave radar is used to monitor the target physical sign data. Of course, those skilled in the art can also use other devices as long as they can ensure the acquisition of the required data. This is the prior art and will not be elaborated here.

[0028] Specifically, the specific device of the image acquisition unit is not limited. For example, it can be a camera or other various types of imaging devices, as long as it can ensure obtaining the regional image. Those skilled in the art can determine it according to the application scenario and requirements, which will not be elaborated here.

[0029] It can be understood that for the target without the potential anomaly label, the existing monitoring state is continued to be maintained.

[0030] Specifically, the specific structures of the label setting module, the feature extraction module, the frame extraction analysis module, and the feature analysis module are not limited, and they can all be composed of logic components. The logic components include field programmable components, computers, or microprocessors in a computer.

[0031] Specifically, the label setting module calculates the anomaly characterization value including determining the ratio of the absolute value of the difference between the motion feature and the reference motion feature to the reference motion feature as the motion feature influence factor; determining the physical sign data, including the breathing rate and the heart rate; determining the ratio of the absolute value of the difference between the breathing rate and the reference breathing rate to the reference breathing rate as the breathing rate influence factor; determining the ratio of the absolute value of the difference between the heart rate and the reference heart rate to the reference heart rate as the heart rate influence factor; determining the weighted sum value of the motion feature influence factor, the breathing rate influence factor, and the heart rate influence factor as the anomaly characterization value.

[0032] Specifically, the reference motion feature is calculated in advance. A number of historical motion features of the target are obtained in advance, and the average value of the historical motion features is determined as the reference motion feature.

[0033] Specifically, the reference breathing rate is calculated in advance. A number of historical breathing rates of the target are obtained in advance, and the average value of the historical breathing rates is determined as the reference breathing rate.

[0034] Specifically, the reference heart rate is calculated in advance. A number of historical heart rates of the target are obtained in advance, and the average value of the historical heart rates is determined as the reference heart rate.

[0035] It can be understood that when calculating the reference motion feature, the reference breathing rate, and the reference heart rate, data cleaning needs to be performed on the data first to ensure the usability of the data.

[0036] Specifically, the sum of the weight coefficients of the motion feature influence factor, the breathing rate influence factor, and the heart rate influence factor is 1. The weight coefficient of the motion feature influence factor is 0.37, the weight coefficient of the breathing rate influence factor is 0.31, and the weight coefficient of the heart rate influence factor is 0.32.

[0037] Specifically, by calculating the abnormal characterization value, potential abnormal labels are set for each target to classify the targets, and the targets with potential abnormal labels are analyzed. In actual situations, when monitoring targets, it is usually chosen to conduct non-discriminatory analysis on all targets in the area, which will lead to the accumulation of a large amount of data, reducing the efficiency of target abnormal analysis. This not only causes a large waste of computing power but also reduces the analysis accuracy of abnormal targets. In particular, some targets may have small movement amplitudes and will be ignored during abnormal analysis, resulting in incomplete and inaccurate abnormal analysis of the targets. Based on this, the present invention calculates the abnormal characterization value at the initial stage of abnormal analysis of the targets, sets potential abnormal labels for some targets, and only considers the targets with potential abnormal labels during analysis, reducing the waste of computing power and improving the abnormal monitoring efficiency and accuracy.

[0038] Please refer to Figure 2 , Figure 2 which is the logic block diagram for setting potential abnormal labels for each target in the embodiment of the invention. Specifically, the label setting module sets potential abnormal labels for each target, where if the abnormal characterization value is outside the preset abnormal threshold interval, a potential abnormal label is set for the target; if the abnormal characterization value is within the preset abnormal threshold interval, no potential abnormal label is set for the target.

[0039] Specifically, the abnormal threshold interval is a closed interval. The upper limit of the abnormal threshold interval is selected within the interval [1.15, 1.45], and the lower limit of the abnormal threshold interval is selected within the interval [0.65, 0.85].

[0040] Specifically, the feature extraction module estimates the image acquisition unit and the associated time domain segment associated with the target, including to determine the path area of the target based on the spatio-temporal trajectory feature. It can be understood that the path area of the target can be determined according to the moving direction; to determine the image acquisition unit within the path area; to establish the association relationship between the image acquisition unit and the target; to predict the estimated arrival time to reach the path area based on the moving speed, and determine the associated time domain segment based on the estimated arrival time; wherein, taking the estimated arrival time as the midpoint, a time period with a predetermined time length is constructed, and the time period is determined as the associated time domain segment.

[0041] It can be understood that the distance between the target and the path area can be determined, and the ratio of the solved distance to the moving speed is determined as the moving time required, and the estimated arrival time can be determined by combining the current time.

[0042] Specifically, the present invention extracts the spatio-temporal trajectory features of the target with potential abnormal tags. By estimating the image acquisition unit associated with the target and the associated time domain segment, it provides a data basis for subsequent evaluation of the feature saliency of the target. In the prior art, when analyzing the data of the target, in most cases, all the data in the acquisition area is selected and then the data related to the target is selected, which wastes computing power and reduces the analysis speed. Based on this, the present invention determines the action state of the target through the spatio-temporal trajectory features of the target, so as to determine several image acquisition units associated with the target and the time to reach several image acquisition units, accurately locate the associated time domain segment of the target, extract the video segment within the time domain segment, and identify the dynamic features of the target, reducing the waste of computing power and improving the efficiency and accuracy of anomaly monitoring.

[0043] Specifically, the frame extraction analysis module determines the required video segment to be extracted, including extracting the video segment collected by the image acquisition unit associated with the target within the associated time domain segment.

[0044] Specifically, the joint points of the joint point graph of the feature analysis module include head joint points, hand joint points, foot joint points, and chest joint points.

[0045] It can be understood that those skilled in the art can also increase or decrease the joint points according to the actual situation, as long as it can provide a data basis for identifying the dynamic features, which will not be elaborated here.

[0046] Specifically, the feature analysis module identifies the dynamic features of the target, including determining the displacement amount of each joint point in the joint point graph; setting the total displacement amount as the dynamic feature.

[0047] Specifically, the displacement amount is the sum of the horizontal displacement amount and the vertical displacement amount of the joint point, which will not be elaborated here.

[0048] Please refer to Figure 3 , Figure 3 which is the logic block diagram for evaluating the feature saliency of the evaluation target in each time domain sub-segment in the embodiment of the invention. Specifically, the feature analysis module evaluates the feature saliency of the target in each time domain sub-segment, where if the dynamic feature is greater than the predetermined dynamic feature, it is determined that the target has feature saliency in the time domain sub-segment; if the dynamic feature is less than or equal to the predetermined dynamic feature, it is determined that the target does not have feature saliency in the time domain sub-segment.

[0049] Specifically, the predetermined dynamic feature is pre-calculated. A number of historical dynamic features of the target are obtained in advance, and the average value of the historical dynamic features is determined as the predetermined dynamic feature.

[0050] Specifically, the present invention identifies the dynamic features of the target, evaluates the feature saliency of the target in each time-domain sub-segment to determine the time-domain sub-segment that needs to be analyzed, providing a data basis for determining whether the target is abnormal. In actual situations, not every video frame of the video segment needs to be analyzed. The target may remain stationary in a certain time-domain sub-segment. Analyzing this time-domain sub-segment will not only cause waste of computing power and unable to judge the abnormal result, but also reduce the analysis speed. Based on this, the present invention determines the time-domain sub-segment that needs to be analyzed by evaluating the feature saliency of the target in each time-domain sub-segment, reducing the waste of computing power and improving the efficiency and accuracy of abnormal monitoring.

[0051] Specifically, the result analysis module determines the time-domain sub-segment that needs to be analyzed based on the evaluation result of the feature analysis module for feature saliency, including the time-domain sub-segment for determining the presence of feature saliency of the target; for determining the time-domain sub-segment as the time-domain sub-segment that needs to be analyzed.

[0052] It can be understood that if the target does not have a feature saliency value, the target may remain stationary in this time-domain sub-segment. Preferably, to ensure the effectiveness of the calculation and improve the calculation efficiency, this time-domain sub-segment is not considered, and the time-domain sub-segment with feature saliency is directly analyzed, which will not be elaborated here.

[0053] Please refer to Figure 4 , Figure 4 which is the logic block diagram for determining whether the target is abnormal in the embodiment of the invention.

[0054] Specifically, the result analysis module determines whether the target is abnormal, including for extracting the regional image collected by the image acquisition unit in the time-domain sub-segment; for comparing the regional image with the historical sample image to determine the similarity for the target; If the similarity is greater than or equal to the similarity threshold, it is determined that the target is not abnormal; If the similarity is less than the similarity threshold, it is determined that the target is abnormal.

[0055] Specifically, the calculation method of the similarity is not limited. In implementation, the method of calculating cosine similarity is adopted. The feature vectors of the corresponding action features of the target in the regional image and the historical sample can be extracted, and the cosine similarity is calculated through the feature vectors. The cosine similarity is defined as the similarity. Of course, other methods can also be adopted, which will not be elaborated here.

[0056] It can be understood that the closer the cosine similarity is to 1, the more similar the two actions are; It can be understood that the historical sample images are pre-recorded. To ensure the reliability of comparison, historical sample images of multiple action types can be pre-recorded for comparison. For example, historical sample images of walking type and running type can be recorded. The action type of the target can be determined based on the extracted regional images, and the corresponding historical sample images of the action type can be selected accordingly, which will not be elaborated here.

[0057] Specifically, the similarity threshold is set in advance. A number of historical sample images of a target during normal actions can be pre-recorded, the average similarity between the corresponding historical sample images of a single target is solved, the average of the average similarities corresponding to each target is solved, and 0.95 times of the average is set as the similarity threshold.

[0058] Specifically, the present invention determines the action type of the target within the time domain sub-segment, and determines whether the target is abnormal based on the historical samples corresponding to the action type. When the target is in a dynamic behavior, if there is an abnormality, it will surely be different from the historical samples. Based on this, the present invention determines the abnormality of the target by comparing the difference between the action type of the target within the time domain sub-segment and the historical samples, improving the efficiency and accuracy of abnormality monitoring.

[0059] So far, the technical solution of the present invention has been described in combination with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0060] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A personnel safety intelligent monitoring system, characterized in that: include: A real-time monitoring module, including a plurality of millimeter wave monitoring units arranged in each area for monitoring target vital sign data and a plurality of image acquisition units for continuously acquiring regional images; A label setting module, which is connected to the real-time monitoring module, is used to extract the motion characteristics of each target in the regional image, calculate the abnormal characterization value in combination with the corresponding vital sign data, and set a potential abnormal label for each target; A feature extraction module connected to the label setting module is used to extract the spatiotemporal trajectory features of the target in response to the target being set with a potential abnormal label, and to estimate the image acquisition unit and the associated time domain segment associated with the target; A frame extraction analysis module connected to the feature extraction module for retrieving the regional image stored in the image acquisition unit based on the estimated result of the feature extraction module to determine the video segment to be extracted; A feature analysis module, which is connected to the frame extraction analysis module, and is used to identify the joint point graph corresponding to the target with the abnormal label in the video segment, so as to identify the dynamic characteristics of the target and evaluate the feature prominence of the target in each time domain sub-segment; A result analysis module connected to the feature analysis module is used to determine the time domain sub-segment to be analyzed based on the evaluation result of the feature analysis module on the feature prominence, determine the action type of the target in the time domain sub-segment, and determine whether the target is abnormal based on the historical samples of the corresponding action type; The motion feature is the sum of the target's lateral displacement and the target's longitudinal displacement, and the spatiotemporal trajectory feature includes the target's moving speed and moving direction.

2. The personnel safety intelligent monitoring system according to claim 1, characterized in that: The label setting module calculates the abnormal characterization value including: The ratio of the absolute value of the difference between the motion feature and the reference motion feature to the reference motion feature is used to determine the motion feature influencing factor; To determine vital signs, including respiratory rate and heart rate; The absolute value of the difference between the respiratory rate and the reference respiratory rate and the ratio of the reference respiratory rate are used to determine the respiratory rate influencing factor; The ratio of the absolute value of the difference between the heart rate and the reference heart rate to the reference heart rate is used to determine the heart rate influencing factor; The weighted sum of the motion characteristic influencing factor, the respiratory frequency influencing factor and the heart rate influencing factor is used to determine the abnormal characterization value.

3. The personnel safety intelligent monitoring system according to claim 1, characterized in that: The label setting module sets a potential abnormal label for each target, wherein: If the abnormal characterization value is outside the preset abnormal threshold range, a potential abnormal label is set for the target; If the abnormal characterization value is within the preset abnormal threshold range, no potential abnormal label is set for the target.

4. The personnel safety intelligent monitoring system according to claim 1, characterized in that: The feature extraction module estimates the image acquisition unit and the associated time domain segment associated with the target, including: Determine the path area of ​​the target based on the spatiotemporal trajectory characteristics; An image acquisition unit for determining the area within the pathway; To establish an association relationship between the image acquisition unit and the target; The method is used to predict an estimated arrival time at the route area based on the moving speed, and determine an associated time domain segment based on the estimated arrival time.

5. The personnel safety intelligent monitoring system according to claim 1, characterized in that: The frame extraction analysis module determines the video segment to be extracted, including: Used to extract the video segment captured by the image acquisition unit associated with the target within the associated time domain segment.

6. The personnel safety intelligent monitoring system according to claim 1, characterized in that: The joint points of the joint point graph of the feature analysis module include head joint points, hand joint points, foot joint points and chest joint points.

7. The intelligent personnel safety monitoring system according to claim 6, characterized in that: The feature analysis module identifies the dynamic features of the target, including: To determine the displacement of each joint point in the joint point diagram; Used to set the total displacement value as a dynamic feature.

8. The personnel safety intelligent monitoring system according to claim 1, characterized in that: The feature analysis module evaluates the feature prominence of the target in each time domain sub-segment, where: If the dynamic feature is greater than a predetermined dynamic feature, it is determined that the target has feature prominence in the time domain sub-segment; If the dynamic feature is less than or equal to the predetermined dynamic feature, it is determined that the target does not have feature prominence in the time domain sub-segment.

9. The personnel safety intelligent monitoring system according to claim 1, characterized in that: The result analysis module determines the time domain sub-segment to be analyzed based on the evaluation result of the feature salience by the feature analysis module, including: A time domain sub-segment used to determine the saliency of target features; The time domain sub-segment is used to determine the time domain sub-segment as the time domain sub-segment required for analysis.

10. The personnel safety intelligent monitoring system according to claim 1, characterized in that: The result analysis module determines whether the target is abnormal, including: Used to extract the regional image collected by the image acquisition unit in the time domain sub-segment; Comparing the regional image with the historical sample image to determine the similarity to the target; If the similarity is greater than or equal to the similarity threshold, it is determined that there is no abnormality in the target; If the similarity is less than the similarity threshold, it is determined that the target is abnormal.

Images