Intelligent video monitoring method and system for intelligent community and medium

Abstract

The invention discloses an intelligent video monitoring method and system for an intelligent community and a medium, relates to the field of intelligent communities, and solves the problem that when a current community climbs over a community fence, most of the communities can only be distinguished by naked eyes of a user or checked through a monitoring picture. The video intelligent monitoring method specifically comprises the following steps: analyzing an infrared image in a smart community to obtain a person to be monitored; analyzing the real-time video data of the to-be-monitored personnel to obtain a root node feature matrix, a centripetal node feature matrix and a centrifugal node feature matrix corresponding to the to-be-monitored personnel; according to the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the to-be-monitored person, calculating regional feature vectors of four limbs corresponding to the to-be-monitored person; and analyzing the regional feature vectors of the corresponding four limbs of the to-be-monitored person, and judging the corresponding posture of the to-be-monitored person. According to the invention, intelligent monitoring and accurate identification of the fence crossing behavior are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of smart communities, and in particular to a method, system and medium for intelligent video monitoring for smart communities. Background Art

[0002] A smart community refers to a model that integrates various existing service resources in the community by utilizing various intelligent technologies and methods, such as the Internet of Things, cloud computing, mobile Internet, 5G, big data, artificial intelligence, blockchain and other new-generation information technologies, to provide community residents with a variety of convenient services such as government affairs, business, entertainment, education, medical care, and mutual assistance in life. It is a "cell" of a "smart city" and a new concept and new model of community management. Video surveillance in a smart community is an important part of the construction of a smart community. It combines modern video surveillance technology and intelligent management systems to provide the community with efficient and intelligent security monitoring solutions.

[0003] However, at present, when a community has a situation where someone climbs over the community fence, in most cases, it can only be identified by the naked eye of the user or by the staff through the monitoring screen. This method not only cannot accurately and effectively identify the situation of climbing over the community fence, but also greatly consumes human resources and causes waste; To this end, the present invention proposes a video intelligent monitoring method, system and medium for a smart community. Summary of the invention

[0004] The purpose of the present invention is to propose a video intelligent monitoring method, system and medium for a smart community to solve the problems raised in the above background technology.

[0005] In the first aspect, the purpose of the present invention can be achieved by the following technical solutions: A video intelligent monitoring method for a smart community, the video intelligent monitoring method is as follows: Step S1, collecting infrared images and real-time video data corresponding to the smart community; Step S2, analyzing the infrared images in the smart community to obtain the persons to be monitored; Step S3, analyzing the real-time video data of the monitored person to obtain a root node feature matrix, a centripetal node feature matrix, and a centrifugal node feature matrix corresponding to the monitored person; Step S4, calculating the regional feature vectors of the limbs of the person to be monitored according to the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the person to be monitored; Step S5, analyzing the regional feature vectors corresponding to the limbs of the monitored person, determining the corresponding posture of the monitored person, generating a warning signal or not performing any operation.

[0006] Furthermore, the step S2 includes the following sub-steps: Step S21, obtaining an infrared image corresponding to the smart community, and obtaining a pixel value XSu corresponding to each pixel in the infrared image, where u is the number of the pixel, u=1, 2, ..., m, and m is a positive integer; Step S22, dividing the pixel value corresponding to the pixel point into an R value component XRu, a G value component XGu and a B value component XBu; Step S23, calculating the gray value HDu corresponding to the pixel point by the formula HDu=A1×XRu+A2×XGu+A3×XBu; wherein A1>A2>A3; Step S24, comparing the grayscale value HDu corresponding to the pixel with the grayscale threshold; If the gray value of a pixel is greater than the gray threshold, the gray value of the corresponding pixel is recorded as 1; if the gray threshold of a pixel is less than or equal to the gray threshold, the gray value of the corresponding pixel is recorded as 0; Step S25, considering the pixel points with gray value of 1 as black, and considering the pixel points with gray value of 0 as white, so as to obtain a binary image; Match the black part in the binary image with the person outline preset in the database. If the match fails, the subsequent steps will not be performed. Step S26, if the match is successful, the black part in the binary image is recorded as the location of the person, and the corresponding person is recorded as the person to be monitored; According to the location of the personnel, the personnel skeleton binding points are set for the personnel to be monitored in the real-time video data, and the number of the personnel skeleton binding points is recorded as i, i=1, 2, ..., z, where z is a positive integer.

[0007] Furthermore, the personnel skeleton binding points include the head, the center of the shoulder, the center of the abdomen, the left shoulder, the left elbow, the left hand, the right shoulder, the right elbow, the right hand, the left hip, the left knee, the left foot, the right hip, the right knee and the right foot.

[0008] Furthermore, the step S3 includes the following sub-steps: Step S31, decoding the real-time video data according to the time, obtaining multiple frames of video images, obtaining the corresponding personnel skeleton binding points in the video images at n adjacent times, and connecting all adjacent personnel skeleton binding points to obtain the personnel space-time diagram corresponding to the monitored personnel; Step S32: construct a skeleton node set based on the corresponding skeleton binding points of the personnel in the video images at n adjacent moments, denoted as GD={V ti |t=1, 2, ..., n}, where t represents the time number, V represents the binding point of the skeleton, and V ti Indicates the skeleton binding point of the person numbered i at time t; Step S33, recording the edges used to connect the adjacent personnel skeleton binding points as connecting edges, selecting a root node at each limb position corresponding to the personnel skeleton binding point, and recording the personnel skeleton binding points adjacent to the root node and the corresponding root node as the same binding point partition; Step S34, calculating the node spacing from the personnel skeleton binding points adjacent to the root node in the same binding point partition to the center of gravity of the personnel skeleton; wherein the node spacing is specifically the number of connection edges from the center of gravity of the skeleton to the corresponding personnel skeleton binding point; If the node distance between the personnel skeleton binding point and the center of gravity of the personnel skeleton is less than or equal to the reference value, the corresponding personnel skeleton binding point is recorded as a centripetal skeleton binding point; If the node distance between the personnel skeleton binding point and the center of gravity of the personnel skeleton is greater than the reference value, the corresponding personnel skeleton binding point is recorded as a centrifugal skeleton binding point.

[0009] Furthermore, the step S3 also includes the following sub-steps: Step S35, establishing a three-dimensional matrix JZ=(G, t, V) corresponding to the person to be monitored, where G represents the coordinates corresponding to the person's skeleton binding, t represents the time number, and V represents the person's skeleton binding point; Step S36, convolve the personnel space-time graph with the convolution kernel, and the graph convolution formula can be obtained: ; In the formula, A is the convolution kernel matrix, D is the degree matrix corresponding to the personnel space-time graph, and the elements on the diagonal of the degree matrix are the degrees of each vertex in the personnel space-time graph (that is, the number of connected edges adjacent to it); Step S37, since the personnel skeleton binding points include three types: root node, centripetal skeleton binding point and centrifugal skeleton binding point, the convolution kernels are also three types: Simplifying the graph convolution formula, we get: ; In the formula, k is the number of the convolution kernel, and different types of personnel skeleton binding points perform convolution operations with different convolution kernels; j is the number of the personnel skeleton binding point adjacent to Vti; Step S38, through the graph convolution formula, respectively obtain the root node feature matrix GJZ corresponding to multiple root nodes, the centripetal node feature matrix XJZ corresponding to the centripetal skeleton binding point, and the centrifugal node feature matrix LJZ of the centrifugal skeleton binding point.

[0010] Further, the binding point partitions include: a first hand binding point partition, a second hand binding point partition, a first leg binding point partition and a second leg binding point partition.

[0011] Furthermore, the step S4 includes the following sub-steps: Step S41, obtaining the root node feature matrix GJZ corresponding to the monitored person, the centripetal node feature matrix XJZ corresponding to the centripetal skeleton binding point, and the centrifugal node feature matrix LJZ of the centrifugal skeleton binding point; Step S42, for the same binding point partition, the centripetal feature vector of the corresponding binding point partition is obtained by subtracting the root node feature matrix from the centripetal node feature matrix; the centrifugal feature vector is obtained by subtracting the root node feature matrix from the centrifugal node feature matrix; Step S43: Similarly, the centripetal feature vectors and the eccentric feature vectors of all the binding point partitions are calculated, and the centripetal feature vectors and the eccentric feature vectors are recorded as the regional feature vectors of the corresponding binding point partitions.

[0012] Furthermore, the step S5 includes the following sub-steps: Step S51, obtaining a first hand region feature vector, a second hand region feature vector, a first leg region feature vector, and a second leg region feature vector corresponding to the person to be monitored; Step S52, obtaining pre-trained comparison feature vectors corresponding to the four limbs of a person climbing over a fence; Step S53, matching the first leg region feature vector and the second leg region feature vector with the comparison feature vector respectively, the matching process is specifically as follows: If the centrifugal feature vector corresponding to the regional feature vector is collinear with the centrifugal feature vector of the comparison feature vector, and the sine value of the angle between the centripetal feature vector corresponding to the regional feature vector and the lower centripetal feature vector of the comparison feature vector is less than one-half, then the match is considered successful, otherwise the match is considered failed; Step S54: if the leg region feature vectors all fail to match, it is determined that the monitored person is performing a climbing behavior, and no operation is performed; If any leg region feature vector matches successfully, the first hand region feature vector and the second hand region feature vector are matched with the comparison feature vector; Step S55: if the hand region feature vectors all fail to match, it is determined that the monitored person is performing a climbing behavior, and no operation is performed; If any hand area feature vector matches successfully, it is determined that the monitored person has climbed over the fence, and a warning signal is generated and sent to the user terminal.

[0013] In a second aspect, a video intelligent monitoring system for a smart community includes: Data acquisition module, used to collect infrared images and real-time video data corresponding to the smart community; The infrared analysis module is used to analyze infrared images in the smart community and identify the persons to be monitored; The video data analysis module is used to analyze the real-time video data of the monitored personnel to obtain the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the monitored personnel; A feature vector calculation module is used to calculate the regional feature vectors of the limbs of the person to be monitored according to the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the person to be monitored; The feature vector analysis module is used to analyze the regional feature vectors corresponding to the limbs of the monitored person, determine the corresponding posture of the monitored person, generate a warning signal or do not perform any operation.

[0014] In a third aspect, a computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 8.

[0015] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: The present invention first collects infrared images and real-time video data corresponding to the smart community, and then analyzes the infrared images in the smart community to obtain the persons to be monitored, and then analyzes the real-time video data of the persons to be monitored to obtain the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the persons to be monitored, and calculates the regional feature vectors of the limbs corresponding to the persons to be monitored according to the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the persons to be monitored, and finally analyzes the regional feature vectors of the limbs corresponding to the persons to be monitored, judges the corresponding posture of the persons to be monitored, generates a warning signal or does not perform an operation, and the present invention realizes intelligent monitoring and accurate identification of fence climbing behavior, and effectively avoids a large consumption of human resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0017] Figure 1 is a flow chart of the method of the present invention; Figure 2 A schematic diagram of a skeleton binding of a person according to the present invention; Figure 3 It is a schematic diagram of the space-time diagram in the present invention; Figure 4 A schematic diagram of the binding point partitioning in the present invention; Figure 5 It is a structural schematic diagram of the computer device in the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than 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.

[0019] Example 1: Please refer to Figure 1-Figure 4 As shown, the technical solution provided by the present invention is: a video intelligent monitoring method for a smart community, the method is used to monitor the scene of a person climbing over the fence of the smart community, and the method is specifically as follows: Step S1, collecting infrared images and real-time video data corresponding to the smart community; The infrared image is captured by the infrared camera, and the real-time video data is the video captured by the surveillance camera; both the infrared camera and the surveillance camera only monitor the location of the fence corresponding to the smart community; Step S2, analyzing the infrared images in the smart community to obtain the persons to be monitored; In this embodiment, step S2 includes the following sub-steps: Step S21, obtaining an infrared image corresponding to the smart community, and obtaining a pixel value XSu corresponding to each pixel in the infrared image, where u is the number of the pixel, u=1, 2, ..., m, and m is a positive integer; Step S22, dividing the pixel value corresponding to the pixel point into an R value component XRu, a G value component XGu and a B value component XBu; Step S23, calculate the gray value HDu corresponding to the pixel point by using the formula, the specific formula is: HDu=A1×XRu+A2×XGu+A3×XBu; where A1>A2>A3, A1, A2 and A3 are weight coefficients; Step S24, comparing the gray value HDu corresponding to the pixel with the gray threshold. If the gray value of the pixel is greater than the gray threshold, the gray value of the corresponding pixel is recorded as 1. If the gray threshold of the pixel is less than or equal to the gray threshold, the gray value of the corresponding pixel is recorded as 0. Step S25, treating pixels with a gray value of 1 as black and pixels with a gray value of 0 as white, thereby obtaining a binary image; matching the black portion of the binary image with the person contour preset in the database, and not proceeding to the subsequent steps if the match fails; The matching process specifically involves calculating the similarity between the black part of the binary image of the person and the person's outline. If the similarity is greater than 90%, the match is considered successful. Step S26, if the match is successful, the black part in the binary image is recorded as the location of the person, and the corresponding person is recorded as the person to be monitored; like Figure 2 As shown, according to the location of the personnel, the personnel skeleton binding points are set for the personnel to be monitored in the real-time video data, and the number of the personnel skeleton binding points is recorded as i, i=1, 2, ..., z, where z is a positive integer; in this implementation, the preferred value of z is 15; Among them, the specific skeleton binding points i are: 1-head, 2-shoulder center, 3-abdomen center, 4-left shoulder, 5-left elbow, 6-left hand, 7-right shoulder, 8-right elbow, 9-right hand, 10-left hip, 11-left knee, 12-left foot, 13-right hip, 14-right knee, 15-right foot.

[0020] Step S3, analyzing the real-time video data of the monitored person to obtain a root node feature matrix, a centripetal node feature matrix, and a centrifugal node feature matrix corresponding to the monitored person; In this embodiment, step S3 includes the following sub-steps: Step S31, decode the real-time video data according to the time, obtain multiple frames of video images, obtain the corresponding personnel skeleton binding points in the video images at n adjacent times, connect all adjacent personnel skeleton binding points to obtain the personnel time-space diagram corresponding to the monitored personnel, such as Figure 3 As shown; Step S32: construct a skeleton node set based on the corresponding skeleton binding points of the personnel in the video images at n adjacent moments, denoted as GD={V ti |t=1, 2, ..., n}, where t represents the time number, V represents the binding point of the skeleton, and V ti Indicates the skeleton binding point of the person numbered i at time t; Step S33, as Figure 4 As shown, the edge used to connect the adjacent personnel skeleton binding points is recorded as the connecting edge, a root node is selected at each limb position corresponding to the personnel skeleton binding point, and the personnel skeleton binding points adjacent to the root node and the corresponding root node are recorded as the same binding point partition, and the binding point partition specifically includes: a first hand binding point partition, a second hand binding point partition, a first leg binding point partition, and a second leg binding point partition; Step S34, calculating the node spacing from the personnel skeleton binding points adjacent to the root node in the same binding point partition to the center of gravity of the personnel skeleton; wherein the node spacing is specifically the number of connection edges from the center of gravity of the skeleton to the corresponding personnel skeleton binding point; If the node distance between the personnel skeleton binding point and the center of gravity of the personnel skeleton is less than or equal to the reference value, the corresponding personnel skeleton binding point is recorded as a centripetal skeleton binding point; If the node distance between the personnel skeleton binding point and the center of gravity of the personnel skeleton is greater than the reference value, the corresponding personnel skeleton binding point is recorded as a centrifugal skeleton binding point; Among them, the centripetal bone binding point is closer to the center of gravity of the human skeleton, so when the monitored person is exercising, the bone binding point contains the characteristics of centripetal movement; the eccentric bone binding point is farther away from the center of gravity of the human skeleton, so when the monitored person is exercising, the bone binding point contains the characteristics of centripetal movement; Step S35, establishing a three-dimensional matrix JZ=(G, t, V) corresponding to the person to be monitored, where G represents the coordinates corresponding to the person's skeleton binding, t represents the time number, and V represents the person's skeleton binding point; Step S36, convolve the personnel space-time graph with the convolution kernel, and the graph convolution formula can be obtained: ; The above formula can be understood as taking the connection edge as the weight to calculate the weighted average of the skeleton binding points of the personnel; in the formula, A is the convolution kernel matrix, D is the degree matrix corresponding to the personnel space-time graph, and the elements on the diagonal of the degree matrix are the degrees of each vertex in the personnel space-time graph (that is, the number of connection edges adjacent to it); Step S37, since the personnel skeleton binding points include three types: root node, centripetal skeleton binding point and centrifugal skeleton binding point, the convolution kernels are also three types: Simplifying the graph convolution formula, we get: ; In the formula, k is the number of the convolution kernel, and different types of personnel skeleton binding points perform convolution operations with different convolution kernels; j is the number of the personnel skeleton binding point adjacent to Vti; Step S38, obtaining the root node feature matrix GJZ corresponding to the multiple root nodes, the centripetal node feature matrix XJZ corresponding to the centripetal skeleton binding point, and the centrifugal node feature matrix LJZ of the centrifugal skeleton binding point respectively through the graph convolution formula; Step S4, calculating the regional feature vectors of the limbs of the person to be monitored according to the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the person to be monitored; In this embodiment, step S4 includes the following sub-steps: Step S41, obtaining the root node feature matrix GJZ corresponding to the monitored person, the centripetal node feature matrix XJZ corresponding to the centripetal skeleton binding point, and the centrifugal node feature matrix LJZ of the centrifugal skeleton binding point; Step S42, for the same binding point partition, the centripetal feature vector of the corresponding binding point partition is obtained by subtracting the root node feature matrix from the centripetal node feature matrix; the centrifugal feature vector is obtained by subtracting the root node feature matrix from the centrifugal node feature matrix; Step S43, similarly, the centripetal feature vectors and the eccentric feature vectors of all the binding point partitions are calculated, and the centripetal feature vectors and the eccentric feature vectors are recorded as the regional feature vectors of the corresponding binding point partitions; Step S5, analyzing the regional feature vectors of the limbs of the monitored person, determining the corresponding posture of the monitored person, and then generating a warning signal or not performing any operation; In this embodiment, step S5 includes the following sub-steps: Step S51, obtaining a first hand region feature vector, a second hand region feature vector, a first leg region feature vector, and a second leg region feature vector corresponding to the person to be monitored; Step S52, obtaining pre-trained comparison feature vectors corresponding to the four limbs of a person climbing over a fence; Step S53, matching the first leg region feature vector and the second leg region feature vector with the comparison feature vector respectively, the matching process is specifically as follows: If the centrifugal feature vector corresponding to the regional feature vector is collinear with the centrifugal feature vector of the comparison feature vector, and the sine value of the angle between the centripetal feature vector corresponding to the regional feature vector and the lower centripetal feature vector of the comparison feature vector is less than one-half, then the match is considered successful, otherwise the match is considered failed; Step S54: if the leg region feature vectors all fail to match, it is determined that the monitored person is performing a climbing behavior, and no operation is performed; If any leg region feature vector matches successfully, the first hand region feature vector and the second hand region feature vector are matched with the comparison feature vector; Step S55: if the hand region feature vectors all fail to match, it is determined that the monitored person is performing a climbing behavior, and no operation is performed; If any hand region feature vector matches successfully, it is determined that the monitored person has climbed over the fence, and a warning signal is generated and sent to the user terminal; In this application, if corresponding calculation formulas appear, the above calculation formulas are all dimensionless and take their numerical calculations. The weight coefficients, proportional coefficients and other coefficients in the formulas are set to a result value obtained by quantifying each parameter. The size of the weight coefficient and the proportional coefficient can be determined as long as it does not affect the proportional relationship between the parameter and the result value.

[0021] Embodiment 2: Based on another concept of the same invention, the present invention also proposes a video intelligent monitoring system for a smart community, the system comprising: Data acquisition module, used to collect infrared images and real-time video data corresponding to the smart community; The infrared analysis module is used to analyze infrared images in the smart community and identify the persons to be monitored; The video data analysis module is used to analyze the real-time video data of the monitored personnel to obtain the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the monitored personnel; A feature vector calculation module is used to calculate the regional feature vectors of the limbs of the person to be monitored according to the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the person to be monitored; The feature vector analysis module is used to analyze the regional feature vectors corresponding to the limbs of the monitored person, determine the corresponding posture of the monitored person, generate a warning signal or do not perform any operation.

[0022] Embodiment three: Figure 5 The structural diagram of a computer device is illustrated, and the computer device may include: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. The processor may call the logic instructions in the memory to execute a video intelligent monitoring method for a smart community, and the method includes: collecting infrared images and real-time video data corresponding to the smart community; analyzing the infrared images in the smart community to obtain the persons to be monitored; analyzing the real-time video data of the persons to be monitored to obtain the root node feature matrix, the centripetal node feature matrix, and the centrifugal node feature matrix corresponding to the persons to be monitored; calculating the regional feature vectors of the limbs of the persons to be monitored based on the root node feature matrix, the centripetal node feature matrix, and the centrifugal node feature matrix corresponding to the persons to be monitored; analyzing the regional feature vectors of the limbs of the persons to be monitored, judging the posture of the persons to be monitored, and generating a warning signal or not performing any operation.

[0023] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0024] Embodiment 4: The present application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute a video intelligent monitoring method for a smart community provided by the above methods, which method includes: collecting infrared images and real-time video data corresponding to the smart community; analyzing the infrared images in the smart community to obtain the persons to be monitored; analyzing the real-time video data of the persons to be monitored to obtain the root node feature matrix, centripetal node feature matrix and centrifugal node feature matrix corresponding to the persons to be monitored; calculating the regional feature vectors of the limbs corresponding to the persons to be monitored based on the root node feature matrix, centripetal node feature matrix and centrifugal node feature matrix corresponding to the persons to be monitored; analyzing the regional feature vectors of the limbs corresponding to the persons to be monitored, judging the posture of the persons to be monitored, generating a warning signal or not performing any operation.

[0025] Embodiment 5: The present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute the above-mentioned video intelligent monitoring method, system and medium for a smart community. The method includes: collecting infrared images and real-time video data corresponding to the smart community; analyzing the infrared images in the smart community to obtain the persons to be monitored; analyzing the real-time video data of the persons to be monitored to obtain the root node feature matrix, centripetal node feature matrix and centrifugal node feature matrix corresponding to the persons to be monitored; calculating the regional feature vectors of the limbs corresponding to the persons to be monitored based on the root node feature matrix, centripetal node feature matrix and centrifugal node feature matrix corresponding to the persons to be monitored; analyzing the regional feature vectors of the limbs corresponding to the persons to be monitored, judging the posture of the persons to be monitored, generating a warning signal or not performing any operation.

[0026] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0027] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0028] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A video intelligent monitoring method for a smart community, characterized in that: The specific video intelligent monitoring method is as follows: Step S1, collecting infrared images and real-time video data corresponding to the smart community; Step S2, analyzing the infrared images in the smart community to obtain the persons to be monitored; Step S3, analyzing the real-time video data of the monitored person to obtain a root node feature matrix, a centripetal node feature matrix, and a centrifugal node feature matrix corresponding to the monitored person; Step S4, calculating the regional feature vectors of the limbs of the person to be monitored according to the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the person to be monitored; Step S5, analyzing the regional feature vectors corresponding to the limbs of the monitored person, determining the corresponding posture of the monitored person, and generating a warning signal or not performing any operation.

2. According to claim 1, a method for intelligent video monitoring for a smart community is characterized in that: The step S2 includes the following sub-steps: Step S21, obtaining an infrared image corresponding to the smart community, and obtaining a pixel value XSu corresponding to each pixel in the infrared image, where u is the number of the pixel, u=1, 2, ..., m, and m is a positive integer; Step S22, dividing the pixel value corresponding to the pixel point into an R value component XRu, a G value component XGu and a B value component XBu; Step S23, calculating the gray value HDu corresponding to the pixel point by the formula HDu=A1×XRu+A2×XGu+A3×XBu; wherein A1>A2>A3; Step S24, comparing the grayscale value HDu corresponding to the pixel with the grayscale threshold; If the gray value of a pixel is greater than the gray threshold, the gray value of the corresponding pixel is recorded as 1; if the gray threshold of a pixel is less than or equal to the gray threshold, the gray value of the corresponding pixel is recorded as 0; Step S25, considering the pixel points with gray value of 1 as black, and considering the pixel points with gray value of 0 as white, so as to obtain a binary image; Match the black part in the binary image with the person outline preset in the database. If the match fails, the subsequent steps will not be performed. Step S26, if the match is successful, the black part in the binary image is recorded as the location of the person, and the corresponding person is recorded as the person to be monitored; According to the location of the personnel, the personnel skeleton binding points are set for the personnel to be monitored in the real-time video data, and the number of the personnel skeleton binding points is recorded as i, i=1, 2, ..., z, where z is a positive integer.

3. A method for intelligent video surveillance for a smart community according to claim 2, characterized in that: The personnel skeleton binding points include head, shoulder center, abdomen center, left shoulder, left elbow, left hand, right shoulder, right elbow, right hand, left hip, left knee, left foot, right hip, right knee and right foot.

4. The method for intelligent video monitoring for a smart community according to claim 2, characterized in that: The step S3 includes the following sub-steps: Step S31, decoding the real-time video data according to the time, obtaining multiple frames of video images, obtaining the corresponding personnel skeleton binding points in the video images at n adjacent times, and connecting all adjacent personnel skeleton binding points to obtain the personnel space-time diagram corresponding to the monitored personnel; Step S32: construct a skeleton node set based on the corresponding skeleton binding points of the personnel in the video images at n adjacent moments, denoted as GD={V ti |t=1, 2, ..., n}, where t represents the time number, V represents the binding point of the skeleton, and V ti Indicates the skeleton binding point of the person numbered i at time t; Step S33, recording the edges used to connect the adjacent personnel skeleton binding points as connecting edges, selecting a root node at each limb position corresponding to the personnel skeleton binding point, and recording the personnel skeleton binding points adjacent to the root node and the corresponding root node as the same binding point partition; Step S34, calculating the node spacing from the personnel skeleton binding points adjacent to the root node in the same binding point partition to the center of gravity of the personnel skeleton; wherein the node spacing is specifically the number of connection edges from the center of gravity of the skeleton to the corresponding personnel skeleton binding point; If the node distance between the personnel skeleton binding point and the center of gravity of the personnel skeleton is less than or equal to the reference value, the corresponding personnel skeleton binding point is recorded as a centripetal skeleton binding point; If the node distance between the personnel skeleton binding point and the center of gravity of the personnel skeleton is greater than the reference value, the corresponding personnel skeleton binding point is recorded as a centrifugal skeleton binding point.

5. A method for intelligent video surveillance for a smart community according to claim 4, characterized in that: The step S3 further comprises the following sub-steps: Step S35, establishing a three-dimensional matrix JZ=(G, t, V) corresponding to the person to be monitored, where G represents the coordinates corresponding to the person's skeleton binding, t represents the time number, and V represents the person's skeleton binding point; Step S36, convolve the personnel space-time graph with the convolution kernel, and the graph convolution formula can be obtained: ; In the formula, A is the convolution kernel matrix, D is the degree matrix corresponding to the personnel space-time graph, and the elements on the diagonal of the degree matrix are the degrees of each vertex in the personnel space-time graph; Step S37, since the personnel skeleton binding points include three types: root node, centripetal skeleton binding point and centrifugal skeleton binding point, the convolution kernels are also three types: Simplifying the graph convolution formula, we get: ; In the formula, k is the number of the convolution kernel, and different types of personnel skeleton binding points perform convolution operations with different convolution kernels; j is the number of the personnel skeleton binding point adjacent to Vti; Step S38, through the graph convolution formula, respectively obtain the root node feature matrix GJZ corresponding to multiple root nodes, the centripetal node feature matrix XJZ corresponding to the centripetal skeleton binding point, and the centrifugal node feature matrix LJZ of the centrifugal skeleton binding point.

6. A method for intelligent video surveillance for a smart community according to claim 4, characterized in that: The binding point partitions include: a first hand binding point partition, a second hand binding point partition, a first leg binding point partition, and a second leg binding point partition.

7. A method for intelligent video surveillance for a smart community according to claim 5, characterized in that: The step S4 includes the following sub-steps: Step S41, obtaining the root node feature matrix GJZ corresponding to the monitored person, the centripetal node feature matrix XJZ corresponding to the centripetal skeleton binding point, and the centrifugal node feature matrix LJZ of the centrifugal skeleton binding point; Step S42, for the same binding point partition, the centripetal feature vector of the corresponding binding point partition is obtained by subtracting the root node feature matrix from the centripetal node feature matrix; the centrifugal feature vector is obtained by subtracting the root node feature matrix from the centrifugal node feature matrix; Step S43: Similarly, the centripetal feature vectors and the eccentric feature vectors of all the binding point partitions are calculated, and the centripetal feature vectors and the eccentric feature vectors are recorded as the regional feature vectors of the corresponding binding point partitions.

8. A method for intelligent video surveillance for a smart community according to claim 7, characterized in that: The step S5 comprises the following sub-steps: Step S51, obtaining a first hand region feature vector, a second hand region feature vector, a first leg region feature vector, and a second leg region feature vector corresponding to the person to be monitored; Step S52, obtaining pre-trained comparison feature vectors corresponding to the four limbs of a person climbing over a fence; Step S53, matching the first leg region feature vector and the second leg region feature vector with the comparison feature vector respectively, the matching process is specifically as follows: If the centrifugal feature vector corresponding to the regional feature vector is collinear with the centrifugal feature vector of the comparison feature vector, and the sine value of the angle between the centripetal feature vector corresponding to the regional feature vector and the lower centripetal feature vector of the comparison feature vector is less than one-half, then the match is considered successful, otherwise the match is considered failed; Step S54: if the leg region feature vectors all fail to match, it is determined that the monitored person is performing a climbing behavior, and no operation is performed; If any leg region feature vector matches successfully, the first hand region feature vector and the second hand region feature vector are matched with the comparison feature vector; Step S55: if the hand region feature vectors all fail to match, it is determined that the monitored person is performing a climbing behavior, and no operation is performed; If any hand area feature vector matches successfully, it is determined that the monitored person has climbed over the fence, and a warning signal is generated and sent to the user terminal.

9. A video intelligent monitoring system for a smart community, characterized in that: In combination with a method for intelligent video surveillance for a smart community as described in any one of claims 1 to 8, the method comprises: Data acquisition module, used to collect infrared images and real-time video data corresponding to the smart community; The infrared analysis module is used to analyze infrared images in the smart community and identify the persons to be monitored; The video data analysis module is used to analyze the real-time video data of the monitored personnel to obtain the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the monitored personnel; A feature vector calculation module is used to calculate the regional feature vectors of the limbs of the person to be monitored according to the root node feature matrix, the centripetal node feature matrix and the centrifugal node feature matrix corresponding to the person to be monitored; The feature vector analysis module is used to analyze the regional feature vectors corresponding to the limbs of the monitored person, determine the corresponding posture of the monitored person, generate a warning signal or do not perform any operation.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images