A security monitoring management system based on artificial intelligence

Abstract

The present invention discloses a security monitoring management system based on artificial intelligence, and relates to the technical field of security monitoring. The present invention comprises a security module, a personnel terminal, a monitoring end camera, a main control module, n security end cameras, a flow controller, a monitoring screen and a video recorder, the output end of the security module and the monitoring end camera is connected to the input end of the main control module, the output end of the main control module is connected to the input end of the personnel terminal, the port of the main control module establishes communication with the port of the flow controller, the output end of the security end camera is connected to the input end of the flow controller, and the output end of the flow controller is respectively connected to the monitoring screen and the input end of the video recorder, the present invention dynamically adjusts the transmission bit rate of each picture by analyzing the observation direction of the security guard, can reduce the transmission pressure of the first image information, save transmission bandwidth resources, and can analyze the risk index of each person by constructing a correlation network, and give early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of security monitoring, and in particular to a security monitoring management system based on artificial intelligence. Background Art

[0002] Smart community refers to the integration and application of new generation information technologies such as the Internet of Things, cloud computing, and mobile Internet to provide community residents with a safe, comfortable, convenient, modern, and intelligent living environment, thereby forming a new form of community management based on information-based and intelligent social management and services. Among them, the security monitoring system can achieve 24-hour all-round monitoring and has artificial intelligence analysis functions, such as abnormal behavior detection and crowd gathering warning, which effectively improves the community's security level and is an important part of the smart community.

[0003] The existing smart community security monitoring system only has the functions of image acquisition and storage, and cannot perform real-time image analysis based on artificial intelligence. When the number of security monitoring systems within the community is too large, on the one hand, it will generate greater data transmission pressure, and on the other hand, it cannot perform correlation analysis on every person entering and leaving the community, and cannot provide early warning. Therefore, optimizing security monitoring data transmission based on artificial intelligence and analyzing monitoring to build a correlation network are technical problems that technical personnel in this field need to solve. Summary of the invention

[0004] In view of the deficiencies of the prior art, the present invention provides a security monitoring and management system based on artificial intelligence, which solves the problems raised in the above background technology.

[0005] To achieve the above purpose, the present invention is implemented through the following technical solutions: an artificial intelligence-based security monitoring and management system, including a security module, a personnel terminal, a monitoring end camera, a main control module, n security end cameras, a flow controller, a monitoring screen and a video recorder, the output end of the security module and the monitoring end camera is connected to the input end of the main control module, the output end of the main control module is connected to the input end of the personnel terminal, the port of the main control module establishes communication with the port of the flow controller, the output end of the security end camera is connected to the input end of the flow controller, the output end of the flow controller is respectively connected to the input end of the monitoring screen and the video recorder, and the output end of the security end camera is connected to the input end of the security module;

[0006] The security camera is divided into a door lock recognition monitoring unit and a road monitoring unit. The road monitoring unit is installed on the side of the road in the community. The road monitoring unit is used to obtain the first image information in the area and transmit it to the flow controller. The flow controller forwards the first image information to the monitoring screen. The video recorder is used to store the first image information. The monitoring screen allows the security guard to view the first image information.

[0007] The monitoring end camera obtains the second image information in the monitoring room and transmits it to the main control module. The main control module numbers the display screens on the monitoring screen. Each numbered display screen corresponds to the security end camera one by one. When the security guard's head faces a display screen in the monitoring screen, the main control module executes the head posture analysis program to obtain the security guard's head posture. The main control module finds the corresponding display screen number according to the security guard's head posture and transmits it to the flow controller. The flow controller sets the transmission bit rate of the display screen to high quality according to the number of the display screen, and the flow controller sets the transmission bit rate of the remaining display screens to general. The main control module executes the head posture analysis program and fails to recognize the face. The main control module outputs the result of not recognizing the face to the flow controller. The flow controller sets the transmission bit rate of all display screens to balanced. The main control module determines the display screen watched by the security guard according to the head posture of the security guard, and then transmits the number corresponding to the display screen watched by the security guard to the flow controller. The flow controller increases the transmission bit rate of the display screen corresponding to the number, improves the clarity of the image, and at the same time, the flow controller reduces the transmission bit rate of the display screen that has not been viewed. When no one is watching the monitoring screen, the flow controller reduces the transmission bit rate of the overall display screen;

[0008] The flow controller executes the picture analysis program to filter the first image information to obtain the third image information, and the flow controller saves the third image information to the video recorder;

[0009] The door lock identification monitoring unit is installed at each entrance and exit of the community. The door lock identification monitoring unit is used to obtain facial images of people entering and leaving the community and transmit them to the security module. The security module executes a facial analysis program to assign an identity number to each person. The security module transmits the identity number to the main control module. The main control module executes an association analysis program based on artificial intelligence to obtain a correlation index Ptl, and constructs a correlation network based on the correlation index Ptl. The main control module calculates the risk index PFX of each person entering and leaving the community. The main control module presets a security threshold. When the risk index exceeds the security threshold, the main control module pushes a warning message to the personnel terminal, and the personnel terminal is equipped by the security guard.

[0010] Furthermore, when the head posture analysis program is executed, the distance from the camera position of the monitoring end to the seat of the security guard is a fixed value and remains unchanged, so the pixel ratio of the head area of ​​the security guard in the second image information is also a fixed value, the main control module presets a sliding window, the size of the sliding window is consistent with the size of the head of the security guard in the second image information, the main control module sets a three-dimensional face model, the main control module uses the front view of the three-dimensional face model as a reference, the size of the reference is consistent with the size of the sliding window, the sliding window starts from the upper left corner of the second image information and slides pixel by pixel in sequence to the lower right corner of the second image information, and each sliding of the sliding window calculates the similarity FM with the reference;

[0011] The main control module presets the judgment threshold, and the main control module uses the formula Calculate the similarity FM between the sliding window and the reference, where a is the length of the sliding window and the reference, b is the width of the sliding window and the reference, I is the pixel extracted from the reference, K is the pixel extracted from the sliding window, i is the horizontal coordinate of the pixel in the reference and the sliding window, and j is the vertical coordinate of the pixel in the reference and the sliding window. If the similarity FM is less than or equal to the judgment threshold, the sliding window continues to slide and calculate the similarity FM again until the similarity FM is greater than the judgment threshold, and the calculation of the similarity FM is stopped. The main control module marks the pixels in the sliding window whose similarity FM is greater than the judgment threshold as planar faces;

[0012] If the similarity FM is never greater than the judgment threshold when the sliding window slides to the lower right corner of the second image information, the main control module stops executing the head posture analysis program and outputs a result that no face is recognized;

[0013] The main control module sets five first nodes in the plane face, and the first nodes are respectively located at the two eyes of the plane face, the two corners of the mouth of the plane face, and the tip of the nose of the plane face. The main control module marks the distance between the two eyes of the first node as d1, the main control module marks the distance between the left eye and the tip of the nose of the first node as d2, the main control module marks the distance between the right eye and the tip of the nose of the first node as d3, the main control module marks the distance between the left corner of the mouth and the tip of the nose of the first node as d4, the main control module marks the distance between the right corner of the mouth and the tip of the nose of the first node as d5, and the main control module marks the distance between the two corners of the mouth of the first node as d6;

[0014] The main control module sets five second nodes in the three-dimensional face model, and the second nodes are respectively located at two eyes of the three-dimensional face model, two corners of the mouth of the three-dimensional face model, and the tip of the nose of the three-dimensional face model. The main control module marks the distance between the eyes of the second nodes as D1, the main control module marks the distance between the left eye and the tip of the nose of the second node as D2, the main control module marks the distance between the right eye and the tip of the nose of the second node as D3, the main control module marks the distance between the left corner of the mouth and the tip of the nose of the second node as D4, the main control module marks the distance between the right corner of the mouth and the tip of the nose of the second node as D5, and the main control module marks the distance between the two corners of the mouth of the second node as D6;

[0015] The main control module defines the vertical pitch angle of the 3D face model as Pitch, Pitch≤±30°. The head of the security guard can completely cover the display range of the monitoring screen by rotating it left and right at ±30°. The main control module defines the horizontal left and right angles of the 3D face model as Yaw, Yaw≤±30°. The main control module establishes an angle mapping collection. The main control module rotates the 3D face model in the order of Pitch first and then Yaw. The 3D face model gradually rotates from -30° to 30°. For every 1° rotation of the 3D face model, the main control module calculates D1, D2, D3, D4, D5 and D6 respectively. The main control module saves D1, D2, D3, D4, D5, D6, the corresponding Pitch and the corresponding Yaw to the angle mapping collection;

[0016] The main control module repeatedly matches the lengths of d1, d2, d3, d4, d5 and d6 with each D1, D2, D3, D4, D5 and D6 in the angle mapping collection. The matching accuracy range is ≤±0.3°. The matching accuracy setting range is to increase the probability of successful matching and reduce the impact of calculation errors. After the matching is successful, the main control module marks the vertical pitch angle Pitch corresponding to the successful matching as the head pitch angle;

[0017] The main control module marks the midpoints of the successfully matched d1, d2, d3, d4, d5 and d6 as the third nodes. The main control module counts the number of third nodes on the left and right sides of the vertical center line of the plane face. When the number of the third nodes on the left is greater than the number of the third nodes on the right, the main control module outputs a negative value for the horizontal left and right angle Yaw corresponding to the successful match and marks it as the horizontal angle of the head. When the number of the third nodes on the left is less than the number of the third nodes on the right, the main control module outputs a positive value for the horizontal left and right angle Yaw corresponding to the successful match and marks it as the horizontal angle of the head. When the number of the third nodes on the left is equal to the number of the third nodes on the right, the main control module marks the horizontal left and right angle Yaw corresponding to the successful match as the horizontal angle of the head.

[0018] The main control module combines the horizontal angle of the head and the pitch angle of the head to obtain the security guard's head posture.

[0019] Furthermore, when the image analysis program is executed, the flow controller subtracts the pixels of the previous and next frames of the first image information. If the subtraction result is zero, it means that the pixels of the previous and next frames have not changed. The flow controller counts the number of times g when the subtraction result is zero. The flow controller marks the frame with the highest number of times g as the first layer. If the subtraction result of the previous and next frames of the first image information is not zero, it means that the pixels of the previous and next frames have changed. The flow controller extracts the pixels with non-zero subtraction results and marks them as the second layer. The flow controller uses the first layer as the background and covers the second layer on the first layer to form the third image information. The flow controller forwards the third image information to the video recorder for storage.

[0020] When the road monitoring unit is powered off, the traffic controller clears the statistical times g corresponding to the road monitoring unit and re-executes the image analysis program. When the road monitoring unit needs to move the angle or position, the road monitoring unit will be powered off for safety reasons before the operation is performed. The statistical times g will affect the marking weight of the first layer when the angle and position change subsequently, so the statistical times g will be cleared after power failure.

[0021] Further, when the face analysis program is executed, the security module splits the face image into eye blocks, nose blocks and mouth blocks, the security module extracts the histogram feature h1 and the color feature h2 from the eye blocks, the security module extracts the histogram feature h3 and the color feature h4 from the nose blocks, and the security module extracts the histogram feature h5 and the color feature h6 from the mouth blocks, the histogram feature is a grayscale feature, the vertical axis of the histogram is the grayscale level, the horizontal axis of the histogram is the grayscale occurrence frequency, and the color feature is obtained by adding the hue mean, the saturation mean and the brightness mean;

[0022] The security module combines the histogram features h1, h3 and h5 to form a mixed histogram hx. The security module converts the mixed histogram hx into an MD5 sequence based on the hash algorithm. The security module adds the color features h2, h4 and h6 to obtain the color feature hy. The security module combines the MD5 sequence with the color feature hy to obtain the identity number corresponding to the face image.

[0023] Furthermore, when the correlation analysis program is executed, the main control module marks the time probability of the door lock identification monitoring unit identifying people entering and leaving the community as the time factor Ti, and the main control module marks the location probability of the road monitoring unit identifying people entering and leaving the community as the space factor Lo;

[0024] The main control module is based on the formula

[0025]

[0026] Calculate the correlation index Ptl between each person entering and leaving the community and each security camera, where is the average value of the spatial factor Lo, is the average value of the time factor Ti, q is the number of samples of people entering and leaving the community, s is the correlation coefficient, w1, w2, w3 and w4 are weight coefficients, and the weight coefficients w1, w2, w3, w4 and the correlation coefficient s are determined according to the specific number of security cameras and the density of installation;

[0027] The main control module aggregates the point-to-point correlation index Ptl between all people entering and leaving the community and the security cameras to form a correlation network.

[0028] Furthermore, the risk index is calculated by executing the following procedure:

[0029] The main control module counts the number of times each person in and out of the community appears Cr, sets the weight w5 of the number of appearances, and the main control module uses the formula Calculate the average value Pta of all association indexes Ptl, where i is a random value from 1 to n of the number of security cameras, set the weight of the average value Pta to w6, and the main control module counts the number Px of association indexes Ptl of each person entering and leaving the community that is greater than the average value Pta, and sets the weight of the average value Px to w7. The specific values ​​of w5, w6 and w7 are determined according to the number of security cameras and the density of installation. w5, w6 and w7 simultaneously satisfy w5+w6+w7=1. The main control module calculates the average value Pta according to the formula

[0030]

[0031] Calculate the risk index PFX.

[0032] The present invention has the following beneficial effects:

[0033] 1. By analyzing the security guard's viewing direction and dynamically adjusting the transmission bit rate of each picture, the transmission pressure of the first image information can be reduced and transmission bandwidth resources can be saved.

[0034] 2. By analyzing the correlation index between each person entering and leaving the community and the security camera, and summarizing the correlation index to build a correlation network, the risk index of each person can be analyzed, and people with too high a risk index can be pushed to the personnel terminal, thereby achieving early warning.

[0035] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0037] Figure 1 This is a block diagram of a security monitoring and management system based on artificial intelligence in the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] See also Figure 1 The present invention provides a technical solution: a security monitoring and management system based on artificial intelligence, comprising a security module, a personnel terminal, a monitoring end camera, a main control module, n security end cameras, a flow controller, a monitoring screen and a video recorder, wherein the output end of the security module and the monitoring end camera is connected to the input end of the main control module, the output end of the main control module is connected to the input end of the personnel terminal, the port of the main control module establishes communication with the port of the flow controller, the output end of the security end camera is connected to the input end of the flow controller, the output end of the flow controller is respectively connected to the input end of the monitoring screen and the video recorder, and the output end of the security end camera is connected to the input end of the security module;

[0040] The security camera is divided into a door lock recognition monitoring unit and a road monitoring unit. The road monitoring unit is installed on the side of the road in the community. The road monitoring unit is used to obtain the first image information in the area and transmit it to the flow controller. The flow controller forwards the first image information to the monitoring screen. The video recorder is used to store the first image information. The monitoring screen allows the security guard to view the first image information.

[0041] The monitoring camera obtains the second image information in the monitoring room and transmits it to the main control module. The main control module numbers the display screens on the monitoring screen. Each numbered display screen corresponds to the security camera. When the security guard's head is facing a display screen in the monitoring screen, the main control module executes the head posture analysis program to obtain the security guard's head posture. The main control module finds the corresponding display screen number according to the security guard's head posture and transmits it to the flow controller. The flow controller sets the transmission bit rate of the display screen to high-quality 6000mbps according to the number of the display screen, and sets the transmission bit rate of the remaining display screens to general 2000mbps. If the main control module executes the head posture analysis program and fails to recognize a face, the main module will The result of identifying the face is output to the flow controller, and the flow controller sets the transmission bit rate of all display screens to a balanced 3000 mbps. The transmission bit rate is the amount of data transmitted per unit time for the first image information. The main control module determines the display screen watched by the security guard based on the head posture of the security guard, and then transmits the number corresponding to the display screen watched by the security guard to the flow controller. The flow controller increases the transmission bit rate of the display screen corresponding to the number to improve the clarity of the image. At the same time, the flow controller reduces the transmission bit rate of the display screen that is not being watched. When no one is watching the monitoring screen, the flow controller reduces the transmission bit rate of the overall display screen. By dynamically adjusting the transmission bit rate of each screen, the transmission pressure of the first image information can be reduced, saving transmission bandwidth resources.

[0042] The flow controller executes the picture analysis program to filter the first image information to obtain the third image information, and the flow controller saves the third image information to the video recorder;

[0043] Door lock recognition monitoring units are installed at each entrance and exit of the community. The door lock recognition monitoring units are used to obtain facial images of people entering and leaving the community and transmit them to the security module. The security module executes the facial analysis program to assign an identity number to each person. The security module transmits the identity number to the main control module. The main control module executes the association analysis program based on artificial intelligence to obtain the association index Ptl, and builds a correlation network based on the association index Ptl. The main control module calculates the risk index PFX of each person entering and leaving the community. The main control module presets a security threshold. When the risk index exceeds the security threshold of 95%, the main control module pushes an alert message to the personnel terminal, and the personnel terminal is equipped by the security guard.

[0044] Among them, when the head posture analysis program is executed, the distance from the camera position of the monitoring end to the security guard's seat is a fixed value and remains unchanged, so the pixel ratio of the security guard's head area in the second image information is also a fixed value. The main control module presets a sliding window, and the size of the sliding window is consistent with the size of the security guard's head in the second image information. The main control module sets a three-dimensional face model. The main control module uses the front view of the three-dimensional face model as a reference. The size of the reference is consistent with the size of the sliding window. The sliding window starts from the upper left corner of the second image information and slides pixel by pixel in sequence to the lower right corner of the second image information. Each time the sliding window slides, the similarity FM is calculated with the reference.

[0045] The main control module presets the judgment threshold, and the main control module uses the formula Calculate the similarity FM between the sliding window and the reference, where a is the length of the sliding window and the reference, b is the width of the sliding window and the reference, I is the pixel extracted from the reference, K is the pixel extracted from the sliding window, i is the horizontal coordinate of the pixel in the reference and the sliding window, and j is the vertical coordinate of the pixel in the reference and the sliding window. If the similarity FM is less than or equal to the judgment threshold of 90%, the sliding window continues to slide and calculate the similarity FM again until the similarity FM is greater than the judgment threshold of 90%, then stop calculating the similarity FM, and the main control module marks the pixels in the sliding window whose similarity FM is greater than the judgment threshold of 90% as planar faces;

[0046] If the similarity FM is never greater than the judgment threshold 90% when the sliding window slides to the lower right corner of the second image information, the main control module stops executing the head posture analysis program and outputs a result that the face is not recognized;

[0047] The main control module sets five first nodes in the plane face, and the first nodes are respectively located at the two eyes of the plane face, the two corners of the mouth of the plane face, and the tip of the nose of the plane face. The main control module marks the distance between the two eyes of the first node as d1, the main control module marks the distance between the left eye and the tip of the nose of the first node as d2, the main control module marks the distance between the right eye and the tip of the nose of the first node as d3, the main control module marks the distance between the left corner of the mouth and the tip of the nose of the first node as d4, the main control module marks the distance between the right corner of the mouth and the tip of the nose of the first node as d5, and the main control module marks the distance between the two corners of the mouth of the first node as d6;

[0048] The main control module sets five second nodes in the three-dimensional face model, and the second nodes are respectively located at two eyes of the three-dimensional face model, two corners of the mouth of the three-dimensional face model, and the tip of the nose of the three-dimensional face model. The main control module marks the distance between the eyes of the second nodes as D1, the main control module marks the distance between the left eye and the tip of the nose of the second node as D2, the main control module marks the distance between the right eye and the tip of the nose of the second node as D3, the main control module marks the distance between the left corner of the mouth and the tip of the nose of the second node as D4, the main control module marks the distance between the right corner of the mouth and the tip of the nose of the second node as D5, and the main control module marks the distance between the two corners of the mouth of the second node as D6;

[0049] The main control module defines the vertical pitch angle of the 3D face model as Pitch, Pitch≤±30°. The head of the security guard can completely cover the display range of the monitoring screen by rotating it left and right at ±30°. The main control module defines the horizontal left and right angles of the 3D face model as Yaw, Yaw≤±30°. The main control module establishes an angle mapping collection. The main control module rotates the 3D face model in the order of Pitch first and then Yaw. The 3D face model gradually rotates from -30° to 30°. For every 1° rotation of the 3D face model, the main control module calculates D1, D2, D3, D4, D5 and D6 respectively. The main control module saves D1, D2, D3, D4, D5, D6, the corresponding Pitch and the corresponding Yaw to the angle mapping collection;

[0050] The main control module repeatedly matches the lengths of d1, d2, d3, d4, d5 and d6 with each D1, D2, D3, D4, D5 and D6 in the angle mapping collection. The matching accuracy range is ≤±0.3°. The matching accuracy setting range is to increase the probability of successful matching and reduce the impact of calculation errors. After the matching is successful, the main control module marks the vertical pitch angle Pitch corresponding to the successful matching as the head pitch angle;

[0051] The main control module marks the midpoints of the successfully matched d1, d2, d3, d4, d5 and d6 as the third nodes. The main control module counts the number of third nodes on the left and right sides of the vertical center line of the plane face. When the number of the third nodes on the left is greater than the number of the third nodes on the right, the main control module outputs a negative value for the horizontal left and right angle Yaw corresponding to the successful match and marks it as the horizontal angle of the head. When the number of the third nodes on the left is less than the number of the third nodes on the right, the main control module outputs a positive value for the horizontal left and right angle Yaw corresponding to the successful match and marks it as the horizontal angle of the head. When the number of the third nodes on the left is equal to the number of the third nodes on the right, the main control module marks the horizontal left and right angle Yaw corresponding to the successful match as the horizontal angle of the head.

[0052] The main control module combines the horizontal angle of the head and the pitch angle of the head to obtain the security guard's head posture.

[0053] When the picture analysis program is executed, the flow controller subtracts the pixels of the previous and next frames of the first image information. If the subtraction result is zero, it means that the pixels of the previous and next frames have not changed. The flow controller counts the number of times the subtraction result is zero g, and the flow controller marks the frame with the highest number of times g as the first layer. If the subtraction result of the previous and next frames of the first image information is not zero, it means that the pixels of the previous and next frames have changed. The flow controller extracts the pixels with non-zero subtraction results and marks them as the second layer. The flow controller uses the first layer as the background and covers the second layer on the first layer to form the third image information. The flow controller forwards the third image information to the video recorder for storage. By performing subtraction filtering on the first image information, unimportant pixels in the first image information are removed, and important information when the picture changes is retained, so as to save the storage space occupied by the video recorder.

[0054] When the road monitoring unit is powered off, the traffic controller clears the statistical times g corresponding to the road monitoring unit and re-executes the image analysis program. When the road monitoring unit needs to move the angle or position, the road monitoring unit will be powered off for safety reasons before the operation is performed. The statistical times g will affect the marking weight of the first layer when the angle and position change subsequently, so the statistical times g will be cleared after power failure.

[0055] When the face analysis program is executed, the security module splits the face image into eye blocks, nose blocks and mouth blocks, extracts histogram feature h1 and color feature h2 from the eye blocks, extracts histogram feature h3 and color feature h4 from the nose blocks, and extracts histogram feature h5 and color feature h6 from the mouth blocks. The histogram feature is a grayscale feature, the vertical axis of the histogram is the grayscale level, the grayscale level is 0 to 255, the horizontal axis of the histogram is the grayscale occurrence frequency, and the color feature is the sum of the hue mean, saturation mean and brightness mean;

[0056] The security module combines the histogram features h1, h3 and h5 to form a mixed histogram hx. The security module converts the mixed histogram hx into an MD5 sequence based on the hash algorithm. The security module adds the color features h2, h4 and h6 to obtain the color feature hy. The security module combines the MD5 sequence with the color feature hy to obtain the identity number corresponding to the face image.

[0057] When the correlation analysis program is executed, the main control module marks the time probability of the door lock recognition monitoring unit identifying people entering and leaving the community as the time factor Ti, and the main control module marks the location probability of the road monitoring unit identifying people entering and leaving the community as the space factor Lo;

[0058] The main control module is based on the formula

[0059]

[0060] Calculate the correlation index Ptl between each person entering and leaving the community and each security camera, where is the average value of the spatial factor Lo, is the average value of the time factor Ti, q is the number of samples of people entering and leaving the community, s is the correlation coefficient, w1, w2, w3 and w4 are weight coefficients, and the weight coefficients w1, w2, w3, w4 and the correlation coefficient s are determined according to the specific number of security cameras and the density of installation;

[0061] The main control module aggregates the point-to-point correlation index Ptl between all people entering and leaving the community and the security cameras to form a correlation network.

[0062] The risk index is calculated by executing the following procedure:

[0063] The main control module counts the number of times each person enters and leaves the community, Cr, and sets the weight w5 of the number of times. The main control module uses the formula Calculate the average value Pta of all correlation indexes Ptl, where i is a random value from 1 to n of the number of security cameras, set the weight of the average value Pta to w6, and the main control module counts the number Px of people entering and leaving the community whose correlation index Ptl is greater than the average value Pta, and sets the weight of the average value Px to w7. w5, w6 and w7 satisfy w5+w6+w7=1. The main control module calculates the average value Pta according to the formula

[0064]

[0065] Calculate the risk index PFX.

[0066] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A security monitoring and management system based on artificial intelligence, including a security module, a personnel terminal, a monitoring end camera, a main control module, n security end cameras, a flow controller, a monitoring screen and a video recorder, characterized in that: The output end of the security module and the monitoring end camera is connected to the input end of the main control module, the output end of the main control module is connected to the input end of the personnel terminal, the port of the main control module establishes communication with the port of the flow controller, the output end of the security end camera is connected to the input end of the flow controller, the output end of the flow controller is respectively connected to the input end of the monitoring screen and the video recorder, and the output end of the security end camera is connected to the input end of the security module; The security camera is divided into a door lock recognition monitoring unit and a road monitoring unit. The road monitoring unit is used to obtain the first image information in the area and transmit it to the flow controller. The flow controller forwards the first image information to the monitoring screen. The monitoring end camera acquires the second image information in the monitoring room and transmits it to the main control module, the main control module executes the head posture analysis program to obtain the head posture of the security guard, and the main control module finds the corresponding display screen number according to the head posture and transmits it to the flow controller; The flow controller sets the transmission bit rate of the display screen to high quality according to the number of the display screen, and sets the transmission bit rate of the remaining display screens to normal. The main control module executes the head posture analysis program and fails to recognize the face. The main control module outputs the result of not recognizing the face to the flow controller. The flow controller sets the transmission bit rate of all display screens to balance. The main control module determines the display screen watched by the security guard according to the head posture of the security guard, and then transmits the number corresponding to the display screen watched by the security guard to the flow controller. The flow controller increases the transmission bit rate of the display screen corresponding to the number, improves the clarity of the image, and at the same time reduces the transmission bit rate of the display screen that is not being viewed. When no one is viewing the monitoring screen, the flow controller reduces the transmission bit rate of the overall display screen. The door lock identification monitoring unit is used to obtain facial images of people entering and leaving the community and transmit them to the security module. The security module transmits the identity number to the main control module. The main control module executes the association analysis program to obtain the association index Ptl, and constructs a correlation network based on the association index Ptl. The main control module calculates the risk index PFX of each person entering and leaving the community. When the risk index exceeds the security threshold, the main control module pushes a warning message to the personnel terminal.

2. The security monitoring and management system based on artificial intelligence according to claim 1 is characterized in that: When the head posture analysis program is executed, the main control module presets a sliding window, and the size of the sliding window is consistent with the size of the security guard's head in the second image information, sets a three-dimensional face model, takes the front view of the three-dimensional face model as a reference, and the size of the reference is consistent with the size of the sliding window. The sliding window starts from the upper left corner of the second image information and slides pixel by pixel in sequence to the lower right corner of the second image information. Each time the sliding window slides, the similarity FM is calculated with the reference; Preset judgment threshold, according to the formula Calculate the similarity FM between the sliding window and the reference, where a is the length of the sliding window and the reference, b is the width of the sliding window and the reference, I is the pixel extracted from the reference, K is the pixel extracted from the sliding window, i is the horizontal coordinate of the pixel in the reference and the sliding window, and j is the vertical coordinate of the pixel in the reference and the sliding window. If the similarity FM is less than or equal to the judgment threshold, the sliding window continues to slide and calculate the similarity FM again until the similarity FM is greater than the judgment threshold, then stop calculating the similarity FM, and mark the pixels in the sliding window whose similarity FM is greater than the judgment threshold as planar faces; If the similarity FM is never greater than the judgment threshold when the sliding window slides to the lower right corner of the second image information, the head posture analysis program is stopped and the result that no face is recognized is output; A first node is set in a planar face, and the first nodes are respectively located at the eyes, the corners of the mouth, and the tip of the nose of the planar face, and the distance between the eyes of the first node is marked as d1, the distance between the left eye and the tip of the nose of the first node is marked as d2, the distance between the right eye and the tip of the nose of the first node is marked as d3, the distance between the left corner of the mouth and the tip of the nose of the first node is marked as d4, the distance between the right corner of the mouth and the tip of the nose of the first node is marked as d5, and the distance between the corners of the mouth of the first node is marked as d6; Setting second nodes in the 3D face model, the second nodes are respectively located at the eyes, mouth corners and nose tip of the 3D face model, marking the distance between the eyes of the second nodes as D1, marking the distance between the left eye and the nose tip of the second node as D2, marking the distance between the right eye and the nose tip of the second node as D3, marking the distance between the left mouth corner and the nose tip of the second node as D4, marking the distance between the right mouth corner and the nose tip of the second node as D5, and marking the distance between the mouth corners of the second nodes as D6; Define the vertical pitch angle of the 3D face model as Pitch, Pitch≤±30°, define the horizontal left and right angles of the 3D face model as Yaw, Yaw≤±30°, establish an angle mapping collection, rotate the 3D face model in the order of Pitch first and then Yaw, and gradually rotate the 3D face model from -30° to 30°. Calculate D1, D2, D3, D4, D5 and D6 for each 1° rotation of the 3D face model, and save D1, D2, D3, D4, D5, D6, Pitch and Yaw to the angle mapping collection; Repeatedly match the lengths of d1, d2, d3, d4, d5 and d6 with each D1, D2, D3, D4, D5 and D6 in the angle mapping collection, with a matching accuracy range of ≤±0.3°. After a successful match, mark the corresponding vertical pitch angle Pitch as the head pitch angle; Mark the midpoint of d1, d2, d3, d4, d5 and d6 as the third node, count the number of the third nodes on the left and right sides of the vertical center line of the plane face, when the number of the third nodes on the left side is greater than the number of the third nodes on the right side, output the corresponding horizontal left and right angle Yaw as a negative value and mark it as the horizontal angle of the head, when the number of the third nodes on the left side is less than the number of the third nodes on the right side, output the corresponding horizontal left and right angle Yaw as a positive value and mark it as the horizontal angle of the head, when the number of the third nodes on the left side is equal to the number of the third nodes on the right side, mark the corresponding horizontal left and right angle Yaw as the horizontal angle of the head; The head posture is obtained by combining the head horizontal angle and the head pitch angle.

3. The security monitoring and management system based on artificial intelligence according to claim 1 is characterized in that: When the image analysis program is executed, the flow controller performs pixel subtraction of the previous and next frames of the first image information. If the subtraction result is zero, the number of times the subtraction result is zero (g) is counted, and the frame with the highest number of times (g) is marked as the first layer. If the pixel subtraction result of the previous and next frames of the first image information is not zero, it means that the pixels of the previous and next frames have changed. The pixels with non-zero subtraction results are extracted and marked as the second layer. With the first layer as the background, the second layer is covered on the first layer to form the third image information, and the third image information is forwarded to the video recorder for storage. When the road monitoring unit is powered off, the flow controller clears the statistical times g corresponding to the road monitoring unit and re-executes the picture analysis program.

4. The security monitoring and management system based on artificial intelligence according to claim 1 is characterized in that: When the face analysis program is executed, the security module splits the face image into eye blocks, nose blocks and mouth blocks, extracts histogram feature h1 and color feature h2 from the eye blocks, extracts histogram feature h3 and color feature h4 from the nose blocks, and extracts histogram feature h5 and color feature h6 from the mouth blocks, where the histogram features are grayscale features; The histogram features h1, h3 and h5 are combined to form a mixed histogram hx, which is converted into an MD5 sequence based on the hash algorithm. The color features h2, h4 and h6 are added to obtain the color feature hy. The MD5 sequence and the color feature hy are combined to obtain the identity number corresponding to the face image.

5. The security monitoring and management system based on artificial intelligence according to claim 1 is characterized in that: When the correlation analysis program is executed, the main control module marks the time probability of the door lock recognition monitoring unit identifying people entering and leaving the community as the time factor Ti, and marks the location probability of the road monitoring unit identifying people entering and leaving the community as the space factor Lo; According to the formula Calculate the correlation index Ptl between each person entering and leaving the community and each security camera, where is the average value of the spatial factor Lo, is the average value of the time factor Ti, q is the number of people entering and leaving the community, s is the correlation coefficient, and w1, w2, w3 and w4 are weight coefficients; The point-to-point correlation index Ptl between all people entering and leaving the community and the security cameras is aggregated to form a correlation network.

6. The security monitoring and management system based on artificial intelligence according to claim 1 is characterized in that: The risk index is calculated by executing the following procedure: The main control module counts the number of times each person in and out of the community appears Cr, sets the weight w5 of the number of appearances, and the main control module uses the formula Calculate the average value Pta of all association indexes Ptl, where i is a random value from 1 to n of the number of security cameras, set the weight of the average value Pta to w6, count the number Px of people entering and leaving the community whose association index Ptl is greater than the average value Pta, set the weight of the average value Px to w7, w5, w6 and w7 satisfy w5+w6+w7=1, according to the formula Calculate the risk index PFX.

Images