A big data management and control method based on digital twin three-dimensional scene

Abstract

The present invention relates to the technical field of three-dimensional scene data management and control. Specifically, it is a big data management and control method based on a digital twin three-dimensional scene, including analyzing the historical flow of people in different areas of a target object, dividing the target object area according to the historical flow of people, and determining the patrol form of security robots in different areas; judging the accuracy of emergency alarm instructions based on the monitoring video of the security robot; presetting the accuracy threshold of the emergency alarm instruction, and determining the emergency alarm instruction based on the accuracy of the emergency alarm instruction; when the emergency alarm instruction is accurate, the security robot in the adjacent area where the emergency occurs is deployed and simulated to obtain the security robot deployment score; based on the security robot deployment score, the security robot deployment decision is completed. The present invention reduces the patrol error of the security robot and increases the accuracy of the patrol information by judging the authenticity of the emergency.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional scene data management and control, and specifically to a big data management and control method based on a digital twin three-dimensional scene. Background Art

[0002] Digital twin 3D scenes refer to the use of digital technology and data from the physical world to construct a 3D virtual model corresponding to the real scene, so as to achieve real-time monitoring, simulation and optimization of the real scene. Digital twin 3D scenes can be applied to multiple fields, such as smart cities, industrial manufacturing, transportation, energy environment, etc. Digital twin 3D scenes provide visual, dynamic and intelligent auxiliary tools for decision-making. Based on digital twin 3D scenes, security robots can obtain surveillance videos of different areas in real time, monitor emergencies in real scenes in real time, and send alarms;

[0003] Under the existing technology, during the patrol process, security robots may have errors in monitoring real-time emergency events due to insufficient evidence collection range, camera failures, etc. When the security robot determines that an emergency has occurred in the real-time scene due to monitoring errors, the security robot sends false alarm commands, resulting in inaccurate information. In addition, the alarm command digital twin model makes deployment decisions for the security robot, which wastes security resources to a certain extent. Summary of the invention

[0004] The purpose of the present invention is to provide a big data management and control method based on digital twin three-dimensional scenes to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a big data management and control method based on a digital twin three-dimensional scene, the big data management and control method specifically comprising the following steps:

[0006] S100, analyzing the historical flow of people in different areas of the target object through the target recognition technology of the machine learning algorithm, and dividing the different areas of the target object according to the historical flow of people, and determining the patrol forms of the security robots in different areas of the target object; wherein the patrol forms of the security robots include: fixed-point patrol, round-trip patrol and circular patrol, and the patrol strategies corresponding to the security robots in different areas are different for different tasks and targets;

[0007] S200. Through the Internet of Things, cloud computing, big data and other technologies, a digital twin model corresponding to the target object is constructed in the virtual space to achieve real-time monitoring, simulation and optimization of the target object; when the digital twin model receives the emergency alarm instruction from the security robot, the accuracy of the emergency alarm instruction is judged based on the surveillance video of the security robot that sends the emergency alarm instruction and the security robots in the adjacent area where the security robot that sends the emergency alarm instruction is located;

[0008] S300, presetting an emergency alarm instruction accuracy threshold, and determining the emergency alarm instruction according to the accuracy of the emergency alarm instruction;

[0009] S400, when the emergency alarm instruction is accurate, since the security robots in the adjacent areas of the emergency monitor the entire process of the emergency in the video, in order to reduce data transmission, the security robots in the adjacent areas where the emergency occurs are deployed and simulated based on the historical flow of people in the area where the security robot sending the emergency alarm instruction is located and the adjacent areas to obtain the security robot deployment score;

[0010] S500: Screen the security robots for simulated deployment according to the deployment scores of the security robots to complete deployment decisions of the security robots to improve the response speed of emergency handling.

[0011] Furthermore, the specific method of analyzing the historical flow of people in different areas of the target object in S100 is as follows:

[0012] S101. Obtain historical surveillance videos of the target object through historical data information, and extract frames from the historical surveillance videos to obtain surveillance images sij, where sij represents the jth surveillance image obtained by extracting frames from the ith surveillance video of the target object, a∈{1, 2, 3...I}, i=1, 2, 3...I, I represents the number of historical surveillance videos of the target object, j=1, 2, 3...J, J represents the number of surveillance images obtained by extracting frames from any historical surveillance video, where the historical surveillance videos of the target object are historical surveillance videos of different areas in the target object;

[0013] S102, using HOG feature descriptors and support vector machine classifiers to detect historical surveillance videos to extract frames and obtain pedestrians in surveillance images sij, matching pedestrians in previous and next frames through tracking algorithms, thereby obtaining pedestrian movement trajectories, traversing historical surveillance videos, and obtaining the number of people passing through the i-th area in the historical surveillance videos as mi; the specific method of traversing the historical surveillance videos is to set a cross-line area in any area, and when the movement trajectory of pedestrians intersects with the cross-line area, the number of people passing through any area is accumulated. Since the number of people passing through any area is calculated, there is no need to screen pedestrians even if there is a situation where pedestrians repeatedly enter and exit the surveillance video range of any area of ​​the target object;

[0014] S103. Extract the time points of surveillance images si1 and siJ as ti1 and tiJ respectively, and calculate the historical flow of people in the i-th area of ​​the target object according to the formula: ui=mi / (tiJ-ti1), where mi represents the number of people passing through the i-th area of ​​the target object obtained from the historical surveillance video.

[0015] Furthermore, the specific method for determining the patrol mode of the security robot in different areas of the target object in S100 is as follows:

[0016] S104, forming a regional historical flow of people set according to the historical flow of people in different areas of the target object, randomly selecting k initial clustering centers ck from the regional historical flow of people set according to big data, dividing the regional historical flow of people set into k different historical flow of people areas, wherein the patrol form of the security robot is divided according to the historical flow of people in the area, when the regional flow of people is large, in order to avoid road congestion, etc., the security robot in the area with large flow of people should use the form of fixed-point patrol; when the regional flow of people is moderate, in order to ensure that the road is unobstructed, the security robot in the area with moderate flow of people should use the form of round-trip patrol, when the regional flow of people is small, circular patrol is used to ensure that the security robot in the area can be deployed at any time; calculate the distance between the historical flow of people set data and the initial clustering center ck in space, obtain the initial clustering center ck closest to any historical flow of people set data, and allocate any historical flow of people set data to the cluster corresponding to the initial clustering center ck; wherein, the historical flow of people in the area is inversely proportional to the degree of flexible deployment of security robots in the area;

[0017] The formula for calculating the distance between the historical crowd flow concentration data and the initial cluster center ck in space is:

[0018]

[0019] Where x represents any historical traffic flow data, i.e., the target data object, ck represents the kth initial cluster center, Z represents the dimension of the target data object, and Xz, ckz represent the zth attribute values ​​of x and ck respectively;

[0020] The formula for calculating the sum of squared errors (SSE) in the entire data set is:

[0021]

[0022] The size of SSE indicates the quality of the clustering result. Since SSE is a non-convex function (kmeans), SSE cannot guarantee the global optimal solution, but can only ensure the local optimal solution. However, it can be repeated several times, and the one with the smallest SSE is selected as the final clustering result.

[0023] S105. The target object region division set is obtained by K-Means algorithm as Uk, Uk={Uk1, Uk2, Uk3...UkG}, wherein UkG represents the region of the Gth target object of the kth type of security patrol, g=1, 2, 3...G,

[0024] Furthermore, the specific method for judging the accuracy of the emergency alarm instruction received by the digital twin model from the security robot in S200 is as follows:

[0025] S201. When the security robot detects that an emergency has occurred, it sends a surveillance video clip of the emergency to the data terminal and sends an emergency alarm instruction to the digital twin model; when the digital twin model receives the emergency alarm instruction sent by the security robot, it obtains the surveillance video of the security robot in the adjacent area where the security robot that sends the emergency alarm instruction is located through the digital twin model, and the surveillance video clip extracted from the surveillance video of the security robot in the adjacent area is the same as the time clip of the emergency surveillance video;

[0026] S202, extract frames from the surveillance video clips of the emergency and the surveillance video clips of the security robot in the adjacent area to obtain a surveillance video image, remove noise from the surveillance video image through a fuzzy filter and grayscale the surveillance video image, detect the contour of the surveillance video image using an edge detection filter, and generate surveillance video image information sets Q and Qr, respectively, Q = {Q1, Q2, Q3...QV}, Qr = {Qr1, Qr2, Qr3...QrV}, Qv represents the Vth surveillance video image information extracted from the surveillance video clip of the emergency and subjected to image processing, Qr represents the rth surveillance video clip extracted from the surveillance video clip of the security robot in the adjacent area and subjected to image processing The surveillance video picture information set for image processing, QrV represents the surveillance video clip of the security robot in the rth adjacent area frame-by-frame extraction and image processing of the Vth surveillance video picture information; r = 1, 2, 3 ... R, R represents the number of adjacent areas where the security robot that sends the emergency alarm instruction is located, v = 1, 2, 3 ... V, V represents the number of surveillance video pictures obtained by frame extraction of the surveillance video, because the surveillance video clip extracted from the surveillance video of the security robot in the adjacent area is the same as the time clip of the emergency surveillance video, so the number of surveillance video pictures obtained by frame extraction of the surveillance video clip of the emergency surveillance video clip and the surveillance video clip of the security robot in the adjacent area is consistent;

[0027] S203: construct an information accuracy analysis model based on the surveillance video image information set to analyze and obtain the accurate value of the emergency alarm instruction sent by the security robot.

[0028] Furthermore, the specific method of constructing the information accuracy analysis model in S203 is as follows:

[0029] S203-1. Map the surveillance video image information to a spatial coordinate system according to the surveillance video image information set, using the Euclidean distance formula:

[0030]

[0031] Calculate the similarity between the emergency surveillance video clip and the surveillance video clip of the security robot in the rth adjacent area, where qv is represented by the vector of the vth surveillance video image information in the emergency surveillance video clip in the spatial coordinate system, and qrv is represented by the vector of the vth surveillance video image information in the surveillance video clip of the security robot in the rth adjacent area in the spatial coordinate system;

[0032] S203-2, through the formula: Calculate the exact value of the emergency alarm instruction sent by the security robot.

[0033] Further, the specific method for determining the emergency warning instruction in S300 is as follows: Due to the problem of the similarity of the surveillance video clips caused by the surveillance video angles of the security robots in different regions, the accurate threshold of the preset emergency warning instruction is D'. When D < D', it indicates that the emergency warning instruction sent by the security robot is inaccurate, and the digital twin model sends an instruction to the data terminal to delete the surveillance video clip of the emergency event. When D ≥ D', it indicates that the emergency warning instruction sent by the security robot is accurate, and the digital twin model receives the emergency warning instruction sent by the security robot for warning reminder.

[0034] Further, the specific method for analyzing the deployment score in the deployment simulation of the security robot in S400 is as follows:

[0035] S401. Obtain the coordinates of the security robot that sends the emergency warning instruction as (x0, y0) and the coordinates of the security robots in the adjacent areas of the area where the security robot that sends the emergency warning instruction is located as (xr, yr) through the digital twin model, and obtain the historical pedestrian flow ur in the adjacent area of the security robot that sends the emergency warning instruction.

[0036] S402. Calculate the distance lr from the security robot in the adjacent area to the area where the security robot that sends the emergency warning instruction is located according to the formula lr = [(x0 - xr) 2 +(y0 - yr) 2 1 / 2 , and construct a deployment score model:

[0037]

[0038] Calculate the deployment score of the security robot in the adjacent area, where h is the initial value of the deployment score, β is a coefficient.

[0039] Further, it is characterized in that: the specific method for the security robot to make a deployment decision in S500 is as follows: Sort the security robots in the adjacent areas according to the deployment scores from large to small, and select the security robot with the highest deployment score to deploy to the area where the emergency event occurs.

[0040] ​Compared with the prior art, the beneficial effects achieved by the present invention are as follows: the historical flow of people in different areas of the target object is analyzed through the target recognition technology of the machine learning algorithm, the target object area is divided according to the historical flow of people, and the patrol form of the security robot in different areas of the target object is determined. By dividing the target object area and determining the patrol form of the security robot in different areas, the patrol efficiency and quality are improved to a certain extent, and the patrol cost and risk are reduced; the present invention confirms the accuracy of the emergency alarm instruction sent by the security robot, reduces the errors in the monitoring of real-time scene emergencies caused by insufficient evidence collection range of the security robot, camera failure, etc., and ensures the accuracy of the information; the present invention makes deployment decisions on the security robot to improve the response speed of emergency handling. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0042] Figure 1 It is a flow chart of a big data management and control method based on a digital twin three-dimensional scene of the present invention. DETAILED DESCRIPTION

[0043] 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.

[0044] See also Figure 1 The present invention provides a technical solution: a big data management and control method based on a digital twin three-dimensional scene, the big data management and control method specifically comprising the following steps:

[0045] S100, analyzing the historical flow of people in different areas of the target object through the target recognition technology of the machine learning algorithm, and dividing the different areas of the target object according to the historical flow of people, and determining the patrol forms of the security robots in different areas of the target object; wherein the patrol forms of the security robots include: fixed-point patrol, round-trip patrol and circular patrol, and the patrol strategies corresponding to the security robots in different areas are different for different tasks and targets;

[0046] S200. Through the Internet of Things, cloud computing, big data and other technologies, a digital twin model corresponding to the target object is constructed in the virtual space to achieve real-time monitoring, simulation and optimization of the target object; when the digital twin model receives the emergency alarm instruction from the security robot, the accuracy of the emergency alarm instruction is judged based on the surveillance video of the security robot that sends the emergency alarm instruction and the security robots in the adjacent area where the security robot that sends the emergency alarm instruction is located;

[0047] S300, presetting an emergency alarm instruction accuracy threshold, and determining the emergency alarm instruction according to the accuracy of the emergency alarm instruction;

[0048] S400, when the emergency alarm instruction is accurate, since the security robots in the adjacent areas of the emergency monitor the entire process of the emergency in the video, in order to reduce data transmission, the security robots in the adjacent areas where the emergency occurs are deployed and simulated based on the historical flow of people in the area where the security robot sending the emergency alarm instruction is located and the adjacent areas to obtain the security robot deployment score;

[0049] S500: Screen the security robots for simulated deployment according to the deployment scores of the security robots to complete deployment decisions of the security robots to improve the response speed of emergency handling.

[0050] Furthermore, the specific method of analyzing the historical flow of people in different areas of the target object in S100 is as follows:

[0051] S101. Obtain historical surveillance videos of the target object through historical data information, and extract frames from the historical surveillance videos to obtain surveillance images sij, where sij represents the jth surveillance image obtained by extracting frames from the ith surveillance video of the target object, a∈{1, 2, 3...I}, i=1, 2, 3...I, I represents the number of historical surveillance videos of the target object, j=1, 2, 3...J, J represents the number of surveillance images obtained by extracting frames from any historical surveillance video, where the historical surveillance videos of the target object are historical surveillance videos of different areas in the target object;

[0052] S102, using HOG feature descriptors and support vector machine classifiers to detect historical surveillance videos to extract frames and obtain pedestrians in surveillance images sij, matching pedestrians in previous and next frames through tracking algorithms, thereby obtaining pedestrian movement trajectories, traversing historical surveillance videos, and obtaining the number of people passing through the i-th area in the historical surveillance videos as mi; the specific method of traversing the historical surveillance videos is to set a cross-line area in any area, and when the movement trajectory of pedestrians intersects with the cross-line area, the number of people passing through any area is accumulated. Since the number of people passing through any area is calculated, there is no need to screen pedestrians even if there is a situation where pedestrians repeatedly enter and exit the surveillance video range of any area of ​​the target object;

[0053] S103. Extract the time points of surveillance images si1 and siJ as ti1 and tiJ respectively, and calculate the historical flow of people in the i-th area of ​​the target object according to the formula: ui=mi / (tiJ-ti1), where mi represents the number of people passing through the i-th area of ​​the target object obtained from the historical surveillance video.

[0054] Furthermore, the specific method for determining the patrol mode of the security robot in different areas of the target object in S100 is as follows:

[0055] S104, forming a regional historical flow of people set according to the historical flow of people in different areas of the target object, randomly selecting k initial clustering centers ck from the regional historical flow of people set according to big data, dividing the regional historical flow of people set into k different historical flow of people areas, wherein the patrol form of the security robot is divided according to the historical flow of people in the area, when the regional flow of people is large, in order to avoid road congestion, etc., the security robot in the area with large flow of people should use the form of fixed-point patrol; when the regional flow of people is moderate, in order to ensure that the road is unobstructed, the security robot in the area with moderate flow of people should use the form of round-trip patrol, when the regional flow of people is small, circular patrol is used to ensure that the security robot in the area can be deployed at any time; since the patrol forms of the security robot include fixed-point patrol, round-trip patrol and circular patrol, K=3, calculating the distance between the historical flow of people set data and the initial clustering center ck in space, obtaining the initial clustering center ck closest to any historical flow of people set data, and allocating any historical flow of people set data to the cluster corresponding to the initial clustering center ck; wherein, the historical flow of people in the area is inversely proportional to the degree of flexible deployment of the security robot in the area;

[0056] The formula for calculating the distance between the historical crowd flow concentration data and the initial cluster center ck in space is:

[0057] L(x,ck)=[∑ Z z＝1(xz-ckz) 2 ] 1 / 2

[0058] Where x represents any historical traffic flow data, i.e., the target data object, ck represents the kth initial cluster center, Z represents the dimension of the target data object, and Xz, ckz represent the zth attribute values ​​of x and ck respectively;

[0059] The formula for calculating the sum of squared errors (SSE) in the entire data set is:

[0060]

[0061] The size of SSE indicates the quality of the clustering result. Since SSE is a non-convex function (kmeans), SSE cannot guarantee the global optimal solution, but can only ensure the local optimal solution. However, it can be repeated several times, and the one with the smallest SSE is selected as the final clustering result.

[0062] S105. The target object region division set is obtained by K-Means algorithm as Uk, Uk={Uk1, Uk2, Uk3...UkG}, wherein UkG represents the region of the Gth target object of the kth type of security patrol, g=1, 2, 3...G,

[0063] Furthermore, the specific method for judging the accuracy of the emergency alarm instruction received by the digital twin model from the security robot in S200 is as follows:

[0064] S201. When the security robot detects that an emergency has occurred, it sends a surveillance video clip of the emergency to the data terminal and sends an emergency alarm instruction to the digital twin model; when the digital twin model receives the emergency alarm instruction sent by the security robot, it obtains the surveillance video of the security robot in the adjacent area where the security robot that sends the emergency alarm instruction is located through the digital twin model, and the surveillance video clip extracted from the surveillance video of the security robot in the adjacent area is the same as the time clip of the emergency surveillance video;

[0065] S202, extract frames from the surveillance video clips of the emergency and the surveillance video clips of the security robot in the adjacent area to obtain a surveillance video image, remove noise from the surveillance video image through a fuzzy filter and grayscale the surveillance video image, detect the contour of the surveillance video image using an edge detection filter, and generate surveillance video image information sets Q and Qr, respectively, Q = {Q1, Q2, Q3...QV}, Qr = {Qr1, Qr2, Qr3...QrV}, Qv represents the Vth surveillance video image information extracted from the surveillance video clip of the emergency and subjected to image processing, Qr represents the rth surveillance video clip extracted from the surveillance video clip of the security robot in the adjacent area and subjected to image processing The surveillance video picture information set for image processing, QrV represents the surveillance video clip of the security robot in the rth adjacent area frame-by-frame extraction and image processing of the Vth surveillance video picture information; r = 1, 2, 3 ... R, R represents the number of adjacent areas where the security robot that sends the emergency alarm instruction is located, v = 1, 2, 3 ... V, V represents the number of surveillance video pictures obtained by frame extraction of the surveillance video, because the surveillance video clip extracted from the surveillance video of the security robot in the adjacent area is the same as the time clip of the emergency surveillance video, so the number of surveillance video pictures obtained by frame extraction of the surveillance video clip of the emergency surveillance video clip and the surveillance video clip of the security robot in the adjacent area is consistent;

[0066] S203: construct an information accuracy analysis model based on the surveillance video image information set to analyze and obtain the accurate value of the emergency alarm instruction sent by the security robot.

[0067] Furthermore, the specific method of constructing the information accuracy analysis model in S203 is as follows:

[0068] S203-1. Map the surveillance video image information to a spatial coordinate system according to the surveillance video image information set, using the Euclidean distance formula:

[0069]

[0070] Calculate the similarity between the emergency surveillance video clip and the surveillance video clip of the security robot in the rth adjacent area, where qv is represented by the vector of the vth surveillance video image information in the emergency surveillance video clip in the spatial coordinate system, and qrv is represented by the vector of the vth surveillance video image information in the surveillance video clip of the security robot in the rth adjacent area in the spatial coordinate system;

[0071] S203-2, through the formula: Calculate the exact value of the emergency alarm instruction sent by the security robot.

[0072] Further, the specific method for determining the emergency warning instruction in S300 is as follows: Due to the problem of the similarity of surveillance video segments caused by the surveillance video angles of security robots in different regions being inaccurate, a preset accurate threshold for the emergency warning instruction is D'. When D < D', it indicates that the emergency warning instruction sent by the security robot is inaccurate, and the digital twin model sends an instruction to the data terminal to delete the emergency surveillance video segment; when D ≥ D', it indicates that the emergency warning instruction sent by the security robot is accurate, and the digital twin model receives the emergency warning instruction sent by the security robot for warning reminder.

[0073] Further, the specific method for analyzing the deployment score when the security robot conducts deployment simulation in S400 is as follows:

[0074] S401. Obtain the coordinates of the security robot that sends the emergency warning instruction as (x0, y0) and the coordinates of the security robots in the adjacent areas of the area where the security robot that sends the emergency warning instruction is located as (xr, yr) through the digital twin model, and obtain the historical pedestrian flow in the adjacent area of the security robot that sends the emergency warning instruction as ur;

[0075] S402. Calculate the distance lr from the security robot in the adjacent area to the area where the security robot that sends the emergency warning instruction is located according to the formula lr = [(x0 - xr) 2 +(y0 - yr) 2 1 / 2 , and construct a deployment score model:

[0076]

[0077] Calculate the deployment score of the security robot in the adjacent area, where h is the initial value of the deployment score, β is a coefficient.

[0078] Further, it is characterized in that: the specific method for the security robot to make a deployment decision in S500 is as follows: Sort the security robots in the adjacent areas according to the deployment scores from large to small, and select the security robot with the highest deployment score to deploy to the area where the emergency occurs.

[0079] In this embodiment:

[0080] Based on the surveillance video picture information set, map the surveillance video picture information to the spatial coordinate system, and through the Euclidean distance formula:

[0081]

[0082] ​Calculate the similarity between the emergency surveillance video clip and the surveillance video clips of the security robots in the four adjacent areas in front, behind, left and right;

[0083] Set the emergency alarm instruction accuracy threshold to D'=2*R=8, where R is the number of adjacent areas to the area where the emergency occurs;

[0084] Example 1: The similarities between the emergency monitoring video clips and the monitoring video clips of the security robots in the four adjacent areas are set to d1=1, d2=0.8, d3=1.2, and d4=0.7, respectively, and the exact value of the emergency alarm instruction sent by the security robot is calculated to obtain:

[0085] Among them, D=3.4<D', and the digital twin model sends an instruction to the data terminal to delete the emergency monitoring video clip.

[0086] Example 2: The similarities between the emergency monitoring video clips and the monitoring video clips of the security robots in the four adjacent areas are set to d1=4, d2=4.5, d3=4.9, and d4=3.8 respectively, and the accurate value of the emergency alarm instruction sent by the security robot is calculated to obtain:

[0087] Among them, D=17.2>D', the digital twin model receives the emergency alarm command sent by the security robot to issue an alarm reminder;

[0088] Through the digital twin model, the coordinates of the security robot that sends the emergency alarm command are obtained as (2, 4) and the coordinates of the security robot in the area adjacent to the area where the security robot that sends the emergency alarm command is located are obtained as (xr, yr).

[0089] ={(1, 5), (2, 3), (4, 9), (6, 4)}, and obtain the historical flow of people in the adjacent area of ​​the security robot that sends the emergency alarm command as ur ={10, 2, 1, 3};

[0090] According to the formula: lr = [(x0-xr) 2 +(y0-yr) 2 ] 1 / 2 The calculated distances from the security robots in the adjacent areas to the area where the security robots that send emergency alarm instructions are located are {1.414, 1, 5.385, 4}, and the deployment score model is constructed:

[0091]

[0092] The deployment scores of security robots in adjacent areas are calculated to be {67.172, 92, 86.23, 83}, and the deployment scores of security robots in adjacent areas are sorted from large to small to obtain {92, 86.23, 83, 67.172}. The security robot with a deployment score of 92 is selected to be deployed to the area where the emergency occurs.

[0093] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0094] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A big data management and control method based on digital twin three-dimensional scene, characterized in that: The big data management and control method specifically comprises the following steps: S100, analyzing the historical flow of people in different areas of the target object through the target recognition technology of the machine learning algorithm, and dividing the different areas of the target object according to the historical flow of people, and determining the patrol form of the security robot in different areas of the target object; S200, constructing a digital twin model corresponding to the target object in a virtual space, and when the digital twin model receives an emergency alarm instruction from the security robot, judging the accuracy of the emergency alarm instruction based on the surveillance video of the security robot sending the emergency alarm instruction and the security robots in the adjacent area where the security robot sending the emergency alarm instruction is located; S300, presetting an emergency alarm instruction accuracy threshold, and determining the emergency alarm instruction according to the accuracy of the emergency alarm instruction; S400, when the emergency alarm instruction is accurate, based on the historical flow of people in the area where the security robot sending the emergency alarm instruction is located and the adjacent areas, a deployment simulation is performed on the security robots in the adjacent areas where the emergency occurs to obtain a security robot deployment score; The specific method of analyzing the deployment score by the deployment simulation of the security robot in S400 is as follows: S401. The coordinates of the security robot that sends the emergency alarm instruction are obtained through the digital twin model as (x0, y0) and the coordinates of the security robots in the area adjacent to the area where the security robot that sends the emergency alarm instruction is located are obtained as (xr, yr). The historical flow of people in the area adjacent to the security robot that sends the emergency alarm instruction is obtained as ur; S402: Calculate the distance from the security robot in the adjacent area to the area where the security robot that sends the emergency alarm instruction is located according to the formula: lr = [(x0-xr) 2 +(y0-yr) 2 ] 1 / 2 , build a deployment score model: The deployment scores of security robots in adjacent areas are calculated, where h is the initial value of the deployment score. β is the coefficient; S500: Screen the security robots for simulated deployment according to the deployment scores of the security robots to complete deployment decisions of the security robots.

2. The big data management and control method based on digital twin three-dimensional scene according to claim 1 is characterized in that: The specific method of analyzing the historical flow of people in different areas of the target object in S100 is as follows: S101. Obtain historical surveillance videos of the target object through historical data information, and extract frames from the historical surveillance videos to obtain surveillance images sij, where sij represents the jth surveillance image obtained by extracting frames from the i-th surveillance video of the target object, a∈{1, 2, 3...I}, i=1, 2, 3...I, I represents the number of historical surveillance videos of the target object, j=1, 2, 3...J, J represents the number of surveillance images obtained by extracting frames from any historical surveillance video; S102, using HOG feature descriptors and support vector machine classifiers to detect historical surveillance videos to extract frames to obtain pedestrians in surveillance images sij, matching pedestrians in previous and next frames through a tracking algorithm, thereby obtaining pedestrian movement trajectories, traversing historical surveillance videos, and obtaining the number of people passing through the i-th area in the historical surveillance videos as mi; the specific method of traversing the historical surveillance videos is to set a cross-line area in any area, and when the movement trajectory of the pedestrians intersects with the cross-line area, the number of people passing through any area is accumulated; S103. Extract the time points of surveillance images si1 and siJ as ti1 and tiJ respectively, and calculate the historical flow of people in the i-th area of ​​the target object according to the formula: ui=mi / (tiJ-ti1), where mi represents the number of people passing through the i-th area of ​​the target object obtained from the historical surveillance video.

3. The big data management and control method based on digital twin three-dimensional scene according to claim 2 is characterized in that: The specific method for determining the patrol mode of the security robot in different areas of the target object in S100 is as follows: S104, forming a regional historical flow of people set according to the historical flow of people in different areas of the target object, randomly selecting k initial clustering centers ck from the regional historical flow of people set according to big data, calculating the distance between the historical flow of people set data and the initial clustering center ck in space, obtaining the initial clustering center ck closest to any historical flow of people set data, and assigning any historical flow of people set data to the cluster corresponding to the initial clustering center ck; wherein, the historical flow of people in a region is inversely proportional to the degree of flexible deployment of security robots in the region; S105. The target object region division set is obtained by K-Means algorithm as Uk, Uk={Uk1, Uk2, Uk3...UkG}, wherein UkG represents the region of the Gth target object of the kth type of security patrol, g=1, 2, 3...G, 4. The big data management and control method based on digital twin three-dimensional scene according to claim 3 is characterized in that: The specific method for judging the accuracy of the emergency alarm instruction received by the digital twin model from the security robot in S200 is as follows: S201. When the security robot detects an emergency, it sends the emergency monitoring video clip to the data terminal and sends an emergency alarm instruction to the digital twin model; When the digital twin model receives an emergency alarm instruction sent by the security robot, the digital twin model obtains the surveillance video of the security robot in the adjacent area where the security robot that sends the emergency alarm instruction is located, and the surveillance video segment extracted from the surveillance video of the security robot in the adjacent area is the same as the time segment of the emergency surveillance video; S202, extract frames from the surveillance video clips of the emergency and the surveillance video clips of the security robots in the adjacent areas to obtain surveillance video images, remove noise from the surveillance video images through a fuzzy filter and grayscale the surveillance video images, use an edge detection filter to detect the contour of the surveillance video images, and generate surveillance video image information sets Q and Qr, where Q = {Q1, Q2, Q3...QV}, Qr = {Qr1, Qr2, Qr3...QrV}, and Qv represents the frame extraction of the surveillance video clips of the emergency and image processing. The Vth surveillance video image information processed, Qr represents the surveillance video image information set of the surveillance video segment of the security robot in the rth adjacent area frame-by-frame extraction and image processing, QrV represents the Vth surveillance video image information of the surveillance video segment of the security robot in the rth adjacent area frame-by-frame extraction and image processing; r = 1, 2, 3 ... R, R represents the number of adjacent areas where the security robot that sends the emergency alarm instruction is located, v = 1, 2, 3 ... V, V represents the number of surveillance video images obtained by frame extraction of the surveillance video; S203: construct an information accuracy analysis model based on the surveillance video image information set to analyze and obtain the accurate value of the emergency alarm instruction sent by the security robot.

5. The big data management and control method based on digital twin three-dimensional scene according to claim 4 is characterized in that: The specific method for constructing the information accuracy analysis model in S203 is as follows: S203-1. Based on the monitoring video image information set, map the monitoring video image information into the spatial coordinate system through the formula: Calculate the similarity between the monitoring video segment of the emergency event and the monitoring video segment of the security robot in the r-th adjacent area, where qv represents the vector of the v-th monitoring video image information in the monitoring video segment of the emergency event in the spatial coordinate system, and qrv represents the vector of the v-th monitoring video image information in the monitoring video segment of the security robot in the r-th adjacent area; S203-2, through the formula: Calculate the exact value of the emergency alarm instruction sent by the security robot.

6. The big data management and control method based on digital twin three-dimensional scene according to claim 5 is characterized in that: The specific method for determining the emergency event alarm instruction in S300 is as follows: Preset the accurate threshold of the emergency event alarm instruction as D'. When D < D', the digital twin model sends an instruction to the data terminal to delete the monitoring video segment of the emergency event; when D ≥ D', the digital twin model receives the emergency event alarm instruction sent by the security robot for alarm reminder.

7. The big data management and control method based on digital twin three-dimensional scene according to claim 6 is characterized by: The specific method for the security robot to make a deployment decision in S500 is as follows: Sort the security robots in the adjacent areas according to the deployment scores from large to small, and select the security robot with the highest deployment score to deploy to the area where the emergency event occurs.

Images