A smart campus safety monitoring and prediction management system

Abstract

The present invention discloses a smart campus security monitoring prediction management system, including a data acquisition unit, a preprocessing unit, a core processing unit and an optimization design unit. By dividing the grid area, video monitoring points and video monitoring files are uniformly configured and integrated to improve the control efficiency and effect. The edge computing model is then used for preprocessing, thereby extracting invalid video clips, reducing the amount of transmitted data and resource occupation space. Then, through the core processing, abnormal risk prediction, video energy consumption and monitoring image accuracy are obtained, and a corresponding change function is established to generate a parameter configuration scheme, thereby realizing dynamic configuration and flexible optimization of image parameters of video monitoring points in multiple grid areas, and finally achieving the effect of ensuring the precision and accuracy of security monitoring while reducing energy loss on the basis of expanding the scale of the smart campus.

Description

Technical Field

[0001] The present invention relates to the technical field of smart campus, and in particular to a smart campus security monitoring prediction management system. Background Art

[0002] A smart campus usually consists of a campus infrastructure with sensor networks and intelligent hardware as the core, and an intelligent software system deployed on a cloud server in a data center;

[0003] Smart campus security monitoring includes video monitoring and intelligent analysis. Specifically, it uses technologies such as video monitoring systems, face recognition, behavior analysis, and anomaly detection to analyze fire warning and fire prevention, entrance and exit management and safety precautions, emergency response and management, and security information, thereby using big data analysis to improve campus safety.

[0004] When the scale of smart campuses expands, security monitoring systems may face scalability challenges. Traditional architectures may not be able to effectively handle large amounts of data streams and events, resulting in a decrease in the accuracy of abnormality determination and risk prediction, making it difficult to conduct timely analysis and decision-making. In addition, as the scale expands, security monitoring systems usually require a large amount of computing resources and storage space to process and store video images, which can lead to high energy consumption and resource occupation.

[0005] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0006] The purpose of the present invention is to solve the scalability challenges of the security monitoring system caused by the expansion of the scale of smart campuses, including the decreased accuracy of abnormal judgment and risk prediction, the difficulty in timely analysis and decision-making, and the problems of high energy consumption and resource occupation. By dividing the grid area for unified configuration and integrated processing, the efficiency and effect of regulation are improved, pre-processing is performed through the edge computing model to reduce the amount of transmitted data and resource occupation space, and parameter configuration schemes are generated through core processing, thereby realizing dynamic configuration and flexible optimization of image parameters of video surveillance points in multiple grid areas, and finally achieving the precision and accuracy of security monitoring while reducing energy loss on the basis of expanding the scale of smart campuses.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A smart campus safety monitoring prediction management system includes a data acquisition unit, a preprocessing unit, a core processing unit and an optimization design unit, wherein the data acquisition unit, the preprocessing unit, the core processing unit and the optimization design unit are connected by signals;

[0009] The data acquisition unit is used to set the video monitoring points, collect and obtain the corresponding monitoring video files and send them to the preprocessing unit, wherein the video monitoring points are set by first dividing the campus into N grid areas, marking any grid area as i, setting a video monitoring points in the grid area i, and uniformly setting all the video monitoring points in the grid area i through the video parameter control model;

[0010] The preprocessing unit is used to establish a preprocessing model, compare and analyze the collected surveillance video files, determine the validity of the surveillance video files, extract valid surveillance video files and transmit them to the core processing unit;

[0011] The core processing unit is used to receive valid surveillance video files for in-depth analysis:

[0012] First, use image recognition technology to perform abnormality determination and risk prediction on video data to obtain abnormal risks, and establish an energy consumption analysis model to obtain video energy consumption, and then establish an image parameter analysis model to obtain monitoring image accuracy;

[0013] Then, a variation function analysis model is established to obtain the variation function F1 of abnormal risk and monitoring image accuracy, and the variation function F2 of video energy consumption and monitoring image accuracy. By combining the variation function F1 with the variation function F2, a parameter configuration scheme is generated.

[0014] The optimization design unit is used to establish a video parameter control model, dynamically control the surveillance video parameters of N areas respectively, and receive a parameter configuration plan to optimize the video parameter control model.

[0015] Furthermore, the preprocessing unit uses the edge computing model to preprocess the data on the computer edge device. The specific process of establishing the preprocessing model is as follows:

[0016] A1: Set the video acquisition cycle to Ts. There are a video monitoring points in the grid area i. Each video monitoring point collects a monitoring video file. Then, the collected a monitoring video files are input into the preprocessing model.

[0017] Mark any surveillance video file as Wa, and extract the video data of the surveillance video file Wa, wherein the video data includes video duration, energy consumption parameters, and image parameters;

[0018] A2: Since the video acquisition cycle is Ts, the video duration is Ts, and the image acquisition frequency Fs is preset. The surveillance video file is imaged to obtain x image files, and then a comparative analysis is performed to obtain the deviation of the x image files. The specific process is as follows:

[0019] A2-1: Compare and analyze two image files that are adjacent to each other in the image acquisition cycle. For x image files, a total of (x-1) comparisons are performed. Any image file is marked as P1, and the image file obtained in the next image acquisition cycle closest to image file P1 is marked as P2. Image file P1 and image file P2 are compared and analyzed using image recognition technology. After comparing and analyzing (x-1) groups of image files, a formula is established to obtain the image file deviation Dwa of the surveillance video file Wa.

[0020] A2-2: Setting the effective threshold value Y0 of the image file deviation degree Dwa: When the image file deviation degree Dwa is lower than the effective threshold value Y0, the monitoring video file Wa is determined to be invalid; otherwise, the monitoring video file Wa is determined to be valid;

[0021] A3: Preprocess the a surveillance video files in the grid area i in turn, extract the valid surveillance video files, and integrate them into the valid surveillance file group Ui, and then transmit the preprocessed surveillance video files: Taking the grid area as the unit, the valid surveillance file group generated by integrating the N grid areas after preprocessing is synchronously transmitted to the core processing unit for in-depth analysis.

[0022] Furthermore, the specific process of the core processing unit performing in-depth analysis is as follows:

[0023] B1: Analyze the abnormal changes of the effective monitoring file group Ui of the grid area i, determine the abnormal situation of campus safety, generate an alarm prompt signal, accumulate the number of alarm prompt signals of the grid area i, and obtain the abnormal cumulative number of the grid area i;

[0024] B2: Preset the abnormal accumulation times of grid area i in any video acquisition cycle as Cs, take the video acquisition cycle Ts as the X-axis and the abnormal accumulation times Cs as the Y-axis, and establish a dynamic curve change graph Si of the video acquisition cycle Ts-abnormal accumulation times Cs of grid area i;

[0025] B2-1: Establish a curve analysis model and conduct in-depth analysis of the dynamic curve. The specific process is as follows:

[0026] B2-11: First, obtain the coordinates and slopes of all points on the dynamic curve Si, mark any point as q (Xq, Yq), obtain the point closest to point q and mark it as p (Xp, Yp), and then obtain the slope Kq of point q;

[0027] B2-12: Then establish a fluctuation analysis model to obtain the fluctuation coefficient Xσi of the dynamic curve Si. The specific process is: the dynamic curve Si is assumed to have Ni points, and the average value of the vertical coordinates of all points of the dynamic curve Si is obtained. , and then find the standard deviation σy of the ordinate of the curve, and find the average value of the slope of all points of the dynamic curve Si , and then calculate the standard deviation σk of the curve slope, and then combine the standard deviation σy of the curve ordinate and the standard deviation σk of the curve slope, assign corresponding weight factor coefficients respectively, and establish a formula to obtain the fluctuation coefficient Xσi of the dynamic curve Si;

[0028] B2-13: Set the threshold value Ys of abnormal cumulative times Cs. When the ordinate of the dynamic curve Si exceeds the threshold value Ys, extract the corresponding curve segment and mark it as a highly abnormal curve segment;

[0029] B2-13-1: Ns height anomaly curve segments are preset. First, the time length values ​​corresponding to the height anomaly curve segments are obtained and accumulated, and then the total value Tzs of the height anomaly time period is obtained;

[0030] B2-13-2: Analyze the height anomaly curve segments, mark any height anomaly curve segment as Sg, and assume that there are Ng points on the height anomaly curve segment Sg. Mark any point as m (Xm, Ym), and calculate the average value of the ordinates of all points on the height anomaly curve segment Sg. , and then find the standard deviation σg of the ordinate of the height anomaly curve segment Sg;

[0031] B2-13-3: By combining the number of highly abnormal curve segments Ns, the total value of the highly abnormal time period Tzs and the standard deviation σg of the ordinate of the highly abnormal curve segment Sg, the corresponding weight factor coefficient is assigned, and a formula is established to obtain the highly abnormal risk coefficient Xyc of the dynamic curve Si;

[0032] B2-14: Then, by combining the volatility coefficient Xσi of the dynamic curve Si and the highly abnormal risk coefficient Xyc, the corresponding weight factor coefficient is assigned, and a formula is established to obtain the abnormal risk prediction index FXi.

[0033] Furthermore, the specific process of establishing the energy consumption analysis model is as follows:

[0034] C1: Get the energy consumption parameters of a video monitoring points in grid area i:

[0035] The energy consumption parameters include the voltage and current of the video monitoring point. Any video monitoring point in the grid area i is marked as Ja. The power Pa of the video monitoring point Ja is obtained by combining the voltage Va and current Ia of the video monitoring point Ja.

[0036] C2: Combine the power Pa of the video monitoring point Ja and the video duration Ts to obtain the energy consumption Was of the video monitoring point Ja during the video acquisition cycle;

[0037] C3: Then, the comprehensive energy consumption coefficient NHi of the grid area i is obtained by accumulating and summing the energy consumption of a video surveillance points in the grid area i.

[0038] Furthermore, the specific process of establishing the image parameter analysis model is as follows:

[0039] The monitoring image accuracy of grid area i is obtained by collecting image parameters. The image parameters include image frame rate zl, brightness ld and pixel xs. The corresponding weight factor coefficients are assigned to the image frame rate zl, brightness ld and pixel xs respectively, and a formula is established to obtain the monitoring image accuracy coefficient JDi.

[0040] Furthermore, the specific process of establishing the change function analysis model is as follows:

[0041] Then construct a dynamic curve of the energy consumption coefficient NHi of the grid area i and the monitoring image accuracy coefficient JDi, and obtain a change function F1 of the energy consumption coefficient NHi of the grid area i and the monitoring image accuracy coefficient JDi;

[0042] Then, a dynamic curve of the abnormal risk prediction index FXi and the monitoring image accuracy coefficient JDi of the grid area i is constructed, and a variation function F2 of the abnormal risk prediction index FXi and the monitoring image accuracy coefficient JDi of the grid area i is obtained.

[0043] Furthermore, the specific process of generating a parameter configuration scheme is as follows:

[0044] Since the image accuracy coefficient of the same grid area i is the same, JDi=NHi / F1=F2*FXi, the energy consumption coefficient NHi of the grid area i is combined with the change function F1 of the monitoring image accuracy coefficient JDi, and the abnormal risk prediction index FXi is combined with the change function F2 of the monitoring image accuracy coefficient JDi to establish a formula to obtain the parameter configuration influence index PZi;

[0045] The standard interval of the parameter configuration influence index PZi is set, the threshold is compared and the corresponding regulation design is performed. The video acquisition cycle is Ts, the time node t is set to divide the video acquisition cycle into two stages, and the video parameter regulation model is applied at the time node t to regulate the image parameters.

[0046] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0047] The present invention divides the video surveillance points and video surveillance files into grid areas for unified configuration and integrated processing, thereby improving the control efficiency and effect, and then performs preprocessing through the edge computing model to extract invalid video clips, reduce the amount of transmitted data and resource occupation space, and then processes the valid video files through the core to obtain abnormal risk prediction, video energy consumption and monitoring image accuracy, and establishes a corresponding change function to generate parameter configuration scheme, thereby realizing dynamic configuration and flexible optimization of image parameters of video surveillance points in multiple grid areas, and finally achieving the effect of ensuring the precision and accuracy of security monitoring while reducing energy loss on the basis of expanding the scale of smart campus. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Shows a schematic diagram of the modules of the present invention;

[0049] Figure 2 A schematic diagram of the process of the present invention is shown. DETAILED DESCRIPTION

[0050] 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.

[0051] Embodiment 1:

[0052] like Figure 1-2 As shown, a smart campus security monitoring prediction management system includes a data acquisition unit, a preprocessing unit, a core processing unit and an optimization design unit, wherein the data acquisition unit, the preprocessing unit, the core processing unit and the optimization design unit are signal connected;

[0053] The working steps are as follows:

[0054] S1: The optimization design unit establishes a video parameter control model, initializes and controls the monitoring video image parameters of N areas respectively, and the data acquisition unit sets the video monitoring points, collects and obtains the corresponding monitoring video files and sends them to the preprocessing unit;

[0055] The video surveillance point setting method is to first divide the campus into N grid areas, mark any grid area as i, set a video surveillance points in the grid area i, and uniformly set all the video surveillance points in the grid area i through the video parameter control model;

[0056] S2: The preprocessing unit establishes a preprocessing model, compares and analyzes the collected surveillance video files, determines the validity of the surveillance video files, extracts valid surveillance video files and transmits them to the core processing unit. The specific process is as follows:

[0057] The preprocessing unit uses the edge computing model to preprocess data on the computer edge device;

[0058] Set the video acquisition cycle to Ts, there are a video monitoring points in the grid area i, each video monitoring point collects a monitoring video file, and then inputs the collected a monitoring video files into the preprocessing model;

[0059] Mark any surveillance video file as Wa, and extract the video data of the surveillance video file Wa, wherein the video data includes video duration, energy consumption parameters, and image parameters;

[0060] Since the video acquisition cycle is Ts, the video duration is Ts, the image acquisition frequency Fs is preset, the monitoring video file is image intercepted, x image files are obtained, and then comparative analysis is performed to obtain the deviation of x image files. The specific process is as follows:

[0061] Compare and analyze two image files that are adjacent to each other in the image acquisition cycle. Then, a total of (x-1) comparisons are performed on x image files. Any image file is marked as P1, and the image file obtained in the next image acquisition cycle closest to image file P1 is marked as P2. Image file P1 and image file P2 are compared and analyzed by image recognition technology. The image recognition technology can be used to compare and analyze image parameters such as chromaticity, grayscale, and brightness:

[0062] If the image file P1 is inconsistent with the image file P2, it means that the content displayed by the video has changed dynamically, and the image deviation value Zp=1 is generated and marked; otherwise, it means that the content displayed by the video has not changed dynamically, and the image deviation value Zp=0 is generated and marked; after comparing and analyzing (x-1) groups of image files, a formula is established to obtain the image file deviation Dwa of the surveillance video file Wa:

[0063]

[0064] The effective threshold value Y0 of the image file deviation degree Dwa is set. When the image file deviation degree Dwa is lower than the effective threshold value Y0, the monitoring video file Wa is determined to be invalid; otherwise, the monitoring video file Wa is determined to be valid;

[0065] The effective threshold Y0 is set to compensate for the slight errors that may be caused by the image recognition technology during comparative analysis. If the (x-1) group of image files has a very low image deviation ratio, it means that there is a misjudgment in the image files with a very small number of frames. Only when the image deviation value reaches a certain ratio can the surveillance video file be characterized as effective.

[0066] Preprocess the a surveillance video files in the grid area i in turn, extract the valid surveillance video files, and integrate them into the valid surveillance file group Ui;

[0067] The specific process of transmitting the pre-processed surveillance video files is as follows: taking the grid area as the unit, the effective surveillance file group generated by integrating the pre-processed N grid areas is synchronously transmitted to the core processing unit for in-depth analysis;

[0068] Since the content displayed by the surveillance cameras on campus may be in a normal state, such as in corridors and pedestrian roads during classes, if the surveillance displays no dynamic changes, it directly indicates that there is no abnormal situation. Screening out such video files can avoid resource occupation and transmission energy consumption, thereby saving energy;

[0069] In order to ensure the correctness of the screening, a sampling verification step can be set to sample and verify the invalid video clips that have been screened out, and then conduct comparative analysis again through image recognition technology to avoid misscreening;

[0070] S3: The core processing unit receives valid surveillance video files for in-depth analysis:

[0071] S3-1: First, use image recognition technology to perform abnormality determination and risk prediction on video data to obtain abnormal risks, and establish an energy consumption analysis model to obtain video energy consumption, and then establish an image parameter analysis model to obtain monitoring image accuracy;

[0072] S3-1-1: Get abnormal risk first:

[0073] By combining shape analysis algorithms, sound processing algorithms and machine language technology, the abnormal changes of the effective monitoring file group Ui in the grid area i are deeply analyzed to determine the abnormal situation of campus safety, among which the abnormal situation of campus safety includes physical conflicts among students, traffic safety in the school and accidental disasters. The determination process uses existing technologies and generates an alarm prompt signal after abnormal determination.

[0074] The number of alarm prompt signals of the grid area i is accumulated to obtain the abnormal cumulative number of the grid area i. The abnormal cumulative number of the grid area i in any video acquisition cycle is preset as Cs. The video acquisition cycle Ts is taken as the X-axis and the abnormal cumulative number Cs is taken as the Y-axis. A dynamic curve change graph Si of the video acquisition cycle Ts-abnormal cumulative number Cs of the grid area i is established;

[0075] Establish a curve analysis model and conduct in-depth analysis of the dynamic curve. The specific process is as follows:

[0076] a1: First obtain the coordinates and slopes of all points of the dynamic curve Si, mark any point as q (Xq, Yq), obtain the point closest to point q and mark it as p (Xp, Yp), and then obtain the slope Kq of point q: ;

[0077] a2: Then establish the fluctuation analysis model to obtain the fluctuation coefficient Xσi of the dynamic curve Si. The specific process is:

[0078] a2-1: The preset dynamic curve Si has Ni points, and the average value of the vertical coordinates of all points of the dynamic curve Si is calculated. : , and then find the standard deviation σy of the ordinate of the curve: ;

[0079] a2-2: Calculate the average slope of all points on the dynamic curve Si : , and then find the standard deviation σk of the slope of the curve: ;

[0080] a3: By combining the standard deviation of the ordinate of the curve σy and the standard deviation of the slope of the curve σk, the corresponding weight factor coefficients are assigned respectively, and the formula is established to obtain the fluctuation coefficient Xσi of the dynamic curve Si:

[0081]

[0082] Among them, ω1 and ω2 are the weight factor coefficients of the standard deviation σy of the vertical coordinate of the curve and the standard deviation σk of the slope of the curve, respectively, and ω1 and ω2 are both greater than 0. When the standard deviation σy of the vertical coordinate of the curve and the standard deviation σk of the slope of the curve are higher, it means that the fluctuation degree of the dynamic curve Si is higher, which means that the abnormal fluctuation risk of the grid area i is higher;

[0083] b1: Set the threshold value Ys of the abnormal cumulative times Cs. When the ordinate of the dynamic curve Si exceeds the threshold value Ys, the corresponding curve segment is extracted and marked as a highly abnormal curve segment. There are Ns highly abnormal curve segments preset. First, the time length value corresponding to the highly abnormal curve segment is obtained and accumulated, and then the total value Tzs of the highly abnormal time period is obtained;

[0084] b2: Analyze the height abnormal curve segments, mark any height abnormal curve segment as Sg, and presume that there are Ng points on the height abnormal curve segment Sg, and mark any point as m (Xm, Ym);

[0085] Calculate the average value of the ordinates of all points on the height anomaly curve segment Sg : , and then find the standard deviation σg of the ordinate of the height anomaly curve segment Sg: ;

[0086] b3: By combining the number of highly abnormal curve segments Ns, the total value of the highly abnormal time period Tzs and the standard deviation σg of the ordinate of the highly abnormal curve segment Sg, the corresponding weight factor coefficient is assigned, and a formula is established to obtain the highly abnormal risk coefficient Xyc of the dynamic curve Si:

[0087]

[0088] Among them, the height anomaly fluctuation coefficient of the dynamic curve Si is obtained by accumulating the vertical coordinate standard deviation of Ns height anomaly curve segments, μ1, μ2 and μ3 are the number of height anomaly curve segments Ns, the total value of the height anomaly time period Tzs and the weight factor coefficient of the height anomaly fluctuation coefficient, and μ1, μ2 and μ3 are greater than 0;

[0089] When the number of highly abnormal curve segments Ns, the total value of highly abnormal time period Tzs, and the highly abnormal fluctuation coefficient are higher, it means that the highly abnormal risk degree of the dynamic curve Si is higher, and the value of the highly abnormal risk coefficient Xyc is larger;

[0090] c1: Then, by combining the fluctuation coefficient Xσi of the dynamic curve Si and the highly abnormal risk coefficient Xyc, a formula is established to obtain the abnormal risk prediction index FXi:

[0091]

[0092] Among them, α1 and α2 are the weight factor coefficients of the volatility coefficient Xσi of the dynamic curve Si and the highly abnormal risk coefficient Xyc, respectively, and both α1 and α2 are greater than 0;

[0093] Since the number of abnormalities in grid area i in any video acquisition cycle is the frequency of abnormal situations in grid area i, abnormal risk prediction is performed by analyzing the frequency of abnormal situations, and targeted analysis is performed from two aspects: overall and high-risk local aspects to improve the accuracy of abnormal risk prediction;

[0094] S3-1-2: Get video energy consumption:

[0095] The specific process of establishing the energy consumption analysis model is as follows:

[0096] Get the energy consumption parameters at a video monitoring point in the grid area i. The energy consumption parameters include the voltage and current of the video monitoring point. Mark any video monitoring point in the grid area i as Ja. By combining the voltage Va and current Ia of the video monitoring point Ja, get the power Pa of the video monitoring point Ja: ;

[0097] Where, e is the filling factor, and e approaches 1;

[0098] The power Pa of the video monitoring point Ja and the video duration Ts are combined to obtain the energy consumption Was of the video monitoring point Ja during the video acquisition cycle: ;

[0099] Then the comprehensive energy consumption coefficient NHi of a video surveillance point in grid area i is obtained:

[0100]

[0101] S3-1-3: Obtain monitoring image accuracy:

[0102] The specific process of establishing the image parameter analysis model is as follows:

[0103] Image parameters include image frame rate zl, brightness ld and pixel xs. The monitoring image accuracy coefficient JDi of grid area i is obtained by collecting image parameters:

[0104]

[0105] Among them, β1, β2 and β3 are weight factor coefficients of image frame rate zl, brightness ld and pixel xs respectively. When the image frame rate zl, brightness ld and pixel xs are higher, the image accuracy coefficient is higher;

[0106] S3-2: Establish a change function analysis model, obtain the change function F1 of the energy consumption coefficient NHi of the grid area i and the monitoring image accuracy coefficient JDi, and obtain the change function F2 of the abnormal risk prediction index FXi of the grid area i and the monitoring image accuracy coefficient JDi, and generate a parameter configuration plan by combining the change function F1 with the change function F2;

[0107] S3-2-1: The specific process of establishing the change function analysis model is as follows:

[0108] First, construct a dynamic curve of the energy consumption coefficient NHi of grid area i and the monitoring image accuracy coefficient JDi, and obtain the change function F1 of the energy consumption coefficient NHi of grid area i and the monitoring image accuracy coefficient JDi:

[0109] , when the image precision coefficient is higher, the energy consumption coefficient NHi will become higher. When the image parameters are lower, the power consumption will be reduced, and the energy consumption will be lower. For example, the screen brightness is reduced, which belongs to the low power consumption mode and can reduce energy consumption;

[0110] Then, a dynamic curve of the abnormal risk prediction index FXi and the monitoring image accuracy coefficient JDi of the grid area i is constructed, and the variation function F2 of the abnormal risk prediction index FXi and the monitoring image accuracy coefficient JDi of the grid area i is obtained:

[0111] ,When the abnormal risk degree is higher, the required monitoring image accuracy will become higher, and when the monitoring image accuracy is lower, the accuracy will be reduced, resulting in a lower prediction accuracy of the abnormal risk;

[0112] S3-2-2: The specific process of generating parameter configuration scheme is as follows:

[0113] Since the image accuracy coefficient of the same grid area i is the same, JDi=NHi / F1=F2*FXi, therefore, the higher the energy consumption coefficient NHi is, the higher the change function F1 will be, while the higher the abnormal risk prediction index FXi is, the lower the change function F2 will be;

[0114] The energy consumption coefficient NHi of grid area i and the change function F1 of the monitoring image accuracy coefficient JDi, as well as the change function F2 of the abnormal risk prediction index FXi and the monitoring image accuracy coefficient JDi, are combined to obtain the parameter configuration influence index PZi:

[0115]

[0116] Among them, υ1 and υ2 are weight factor coefficients of the variation function F1 and the variation function F2, respectively, and υ1 and υ2 are both greater than 0;

[0117] Set the standard interval of parameter configuration impact index PZi and compare the thresholds:

[0118] When the parameter configuration impact index PZi is lower than the standard range, it means that the abnormal risk is low and the comprehensive energy consumption is high, then the image parameters are reduced by the reduction range = PZi, thereby reducing energy consumption by reducing image accuracy;

[0119] When the parameter configuration impact index PZi is higher than the standard range, it means that the abnormal risk is high and the comprehensive energy consumption is low, then the image parameters are increased by an amount equal to PZi, thereby improving the accuracy of risk prediction by improving the image accuracy;

[0120] When the parameter configuration influence index PZi is in the standard range, it means that the abnormal risk and comprehensive energy consumption are relatively balanced, and the image parameters are not adjusted;

[0121] S4: The optimization design unit receives the parameter configuration scheme to optimize the video parameter control model, thereby dynamically controlling the surveillance video image parameters of the N areas respectively;

[0122] The video acquisition cycle is Ts, and the time node t is set to divide the video acquisition cycle into two stages. At the time node t, the video parameter control model is applied to control the image parameters. Then, the [0, t] time period is the measurement and evaluation stage, and the image parameters of the video monitoring point apply the initialized parameter configuration scheme. The (t, Ts) time period is the control application stage, and the image parameters of the video monitoring point apply the optimized parameter configuration scheme.

[0123] For example, image parameters include image frame rate zl, brightness ld and pixel xs. If the image frame rate zl0=n of the video surveillance point in the current grid area, it is determined that the image parameters need to be reduced, and the reduction range PZi=PZ1, then the adjusted image frame rate zl1=n*(1-PZ1), and the brightness ld and pixel xs are similar.

[0124] To summarize, the present invention divides the video surveillance points and video surveillance files into grid areas for unified configuration and integrated processing, thereby improving the control efficiency and effect, performing preprocessing through the edge computing model to extract invalid video clips, reducing the amount of transmitted data and resource occupancy space, and obtaining abnormal risk prediction, video energy consumption and monitoring image accuracy through core processing, and establishing a corresponding change function to generate a parameter configuration scheme, thereby realizing dynamic configuration and flexible optimization of image parameters of video surveillance points in multiple grid areas, so as to achieve the precision and accuracy of security monitoring while reducing energy loss on the basis of expanding the scale of smart campuses.

[0125] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technical personnel in this field for each group of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0126] The above formulas are all dimensionless and numerical calculations. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0127] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A smart campus safety monitoring prediction management system, characterized by: It includes a data acquisition unit, a preprocessing unit, a core processing unit and an optimization design unit, wherein the data acquisition unit, the preprocessing unit, the core processing unit and the optimization design unit are signal connected; The data acquisition unit is used to set the video monitoring points, collect and obtain the corresponding monitoring video files and send them to the preprocessing unit, wherein the video monitoring points are set by first dividing the campus into N grid areas, marking any grid area as i, setting a video monitoring points in the grid area i, and uniformly setting all the video monitoring points in the grid area i through the video parameter control model; The preprocessing unit is used to establish a preprocessing model, compare and analyze the collected surveillance video files, determine the validity of the surveillance video files, extract valid surveillance video files and transmit them to the core processing unit; The core processing unit is used to receive valid surveillance video files for in-depth analysis: First, use image recognition technology to perform abnormality determination and risk prediction on video data to obtain abnormal risks, and establish an energy consumption analysis model to obtain video energy consumption, and then establish an image parameter analysis model to obtain monitoring image accuracy; Then, a variation function analysis model is established to obtain the variation function F1 of the energy consumption coefficient NHi of the grid area i and the monitoring image accuracy coefficient JDi and the variation function F2 of the abnormal risk prediction index FXi of the grid area i and the monitoring image accuracy coefficient JDi. By combining the variation function F1 with the variation function F2, a parameter configuration scheme is generated. The specific process of generating the parameter configuration scheme is as follows: Since the image accuracy coefficients of the same grid area i are the same, JDi=NHi / F1=F2*FXi, the energy consumption coefficient NHi of the grid area i is combined with the change function F1 of the monitoring image accuracy coefficient JDi, and the abnormal risk prediction index FXi is combined with the change function F2 of the monitoring image accuracy coefficient JDi to establish a formula to obtain the parameter configuration influence index PZi: ; Among them, υ1 and υ2 are weight factor coefficients of the variation function F1 and the variation function F2, respectively, and υ1 and υ2 are both greater than 0; Set the standard interval of parameter configuration influence index PZi, compare the thresholds and make corresponding control designs, where the video acquisition cycle is Ts, set the time node t to divide the video acquisition cycle into two stages, and apply the video parameter control model to control the image parameters at the time node t; The optimization design unit is used to establish a video parameter control model, dynamically control the surveillance video parameters of N areas respectively, and receive parameter configuration solutions to optimize the video parameter control model; The preprocessing unit uses the edge computing model to preprocess the data on the computer edge device. The specific process of establishing the preprocessing model is as follows: A1: Set the video acquisition cycle to Ts. There are a video monitoring points in the grid area i. Each video monitoring point collects a monitoring video file. Then, the collected a monitoring video files are input into the preprocessing model. Mark any surveillance video file as Wa, and extract the video data of the surveillance video file Wa, wherein the video data includes video duration, energy consumption parameters, and image parameters; A2: Since the video acquisition cycle is Ts, the video duration is Ts, the image acquisition frequency Fs is preset, the monitoring video file is image intercepted, x image files are obtained, and then comparative analysis is performed to obtain the deviation of x image files. The specific process is: A2-1: Compare and analyze two image files that are adjacent to each other in the image acquisition cycle. Then, a total of (x-1) comparisons are performed for x image files. Any image file is marked as P1, and the image file obtained in the next image acquisition cycle closest to image file P1 is marked as P2. Image files P1 and P2 are compared and analyzed using image recognition technology. After comparing and analyzing (x-1) groups of image files, a formula is established to obtain the image file deviation Dwa of the surveillance video file Wa: , where Zp is the image deviation value. If the image file P1 is inconsistent with the image file P2, the image deviation value Zp=1 is generated and marked; otherwise, the image deviation value Zp=0 is generated and marked; A2-2: Setting the effective threshold value Y0 of the image file deviation degree Dwa: When the image file deviation degree Dwa is lower than the effective threshold value Y0, the monitoring video file Wa is determined to be invalid; otherwise, the monitoring video file Wa is determined to be valid; A3: Preprocess the a surveillance video files in the grid area i in turn, extract the valid surveillance video files, and integrate them into the valid surveillance file group Ui, and then transmit the preprocessed surveillance video files: Taking the grid area as the unit, the valid surveillance file group generated by integrating the N grid areas after preprocessing is synchronously transmitted to the core processing unit for in-depth analysis.

2. The smart campus safety monitoring prediction management system according to claim 1 is characterized by: The specific process of the core processing unit performing in-depth analysis is as follows: B1: Analyze the abnormal changes of the effective monitoring file group Ui of the grid area i, determine the abnormal situation of campus safety, generate an alarm prompt signal, accumulate the number of alarm prompt signals of the grid area i, and obtain the abnormal cumulative number of the grid area i; B2: Preset the abnormal accumulation times of grid area i in any video acquisition cycle as Cs, take the video acquisition cycle Ts as the X-axis and the abnormal accumulation times Cs as the Y-axis, and establish a dynamic curve change graph Si of the video acquisition cycle Ts-abnormal accumulation times Cs of grid area i; B2-1: Establish a curve analysis model and conduct in-depth analysis of the dynamic curve. The specific process is as follows: B2-11: First, obtain the coordinates and slopes of all points on the dynamic curve Si, mark any point as q (Xq, Yq), obtain the point closest to point q and mark it as p (Xp, Yp), and then obtain the slope Kq of point q; B2-12: Then establish a fluctuation analysis model to obtain the fluctuation coefficient Xσi of the dynamic curve Si. The specific process is: the dynamic curve Si is assumed to have Ni points, and the average value of the vertical coordinates of all points of the dynamic curve Si is obtained. , and then find the standard deviation σy of the ordinate of the curve, and find the average value of the slope of all points of the dynamic curve Si , and then calculate the standard deviation σk of the curve slope, and then combine the standard deviation σy of the curve ordinate and the standard deviation σk of the curve slope, assign corresponding weight factor coefficients respectively, and establish a formula to obtain the fluctuation coefficient Xσi of the dynamic curve Si; B2-13: Set the threshold value Ys of abnormal cumulative times Cs. When the ordinate of the dynamic curve Si exceeds the threshold value Ys, extract the corresponding curve segment and mark it as a highly abnormal curve segment; B2-13-1: Ns height anomaly curve segments are preset. First, the time length values ​​corresponding to the height anomaly curve segments are obtained and accumulated, and then the total value Tzs of the height anomaly time period is obtained; B2-13-2: Analyze the height anomaly curve segments, mark any height anomaly curve segment as Sg, and assume that there are Ng points on the height anomaly curve segment Sg. Mark any point as m (Xm, Ym), and calculate the average value of the ordinates of all points on the height anomaly curve segment Sg. , and then find the standard deviation σg of the ordinate of the height anomaly curve segment Sg; B2-13-3: By combining the number of highly abnormal curve segments Ns, the total value of the highly abnormal time period Tzs and the standard deviation σg of the ordinate of the highly abnormal curve segment Sg, the corresponding weight factor coefficient is assigned, and a formula is established to obtain the highly abnormal risk coefficient Xyc of the dynamic curve Si; B2-14: Then, by combining the volatility coefficient Xσi of the dynamic curve Si and the highly abnormal risk coefficient Xyc, the corresponding weight factor coefficient is assigned, and a formula is established to obtain the abnormal risk prediction index FXi.

3. The smart campus safety monitoring prediction management system according to claim 2 is characterized by: The specific process of establishing the energy consumption analysis model is as follows: C1: Get the energy consumption parameters of a video monitoring points in grid area i: The energy consumption parameters include the voltage and current of the video monitoring point. Any video monitoring point in the grid area i is marked as Ja. The power Pa of the video monitoring point Ja is obtained by combining the voltage Va and current Ia of the video monitoring point Ja. C2: Combine the power Pa of the video monitoring point Ja and the video duration Ts to obtain the energy consumption Was of the video monitoring point Ja during the video acquisition cycle; C3: Then, the comprehensive energy consumption coefficient NHi of the grid area i is obtained by accumulating and summing the energy consumption of a video surveillance points in the grid area i.

4. The smart campus safety monitoring prediction management system according to claim 3 is characterized by: The specific process of establishing the image parameter analysis model is as follows: The monitoring image accuracy of grid area i is obtained by collecting image parameters. The image parameters include image frame rate zl, brightness ld and pixel xs. The corresponding weight factor coefficients are assigned to the image frame rate zl, brightness ld and pixel xs respectively, and a formula is established to obtain the monitoring image accuracy coefficient JDi.

5. The smart campus safety monitoring prediction management system according to claim 4 is characterized by: The specific process of establishing the change function analysis model is as follows: Construct a dynamic curve of the energy consumption coefficient NHi of the grid area i and the monitoring image accuracy coefficient JDi, and obtain a change function F1 of the energy consumption coefficient NHi of the grid area i and the monitoring image accuracy coefficient JDi; Then, a dynamic curve of the abnormal risk prediction index FXi and the monitoring image accuracy coefficient JDi of the grid area i is constructed, and a variation function F2 of the abnormal risk prediction index FXi and the monitoring image accuracy coefficient JDi of the grid area i is obtained.

Images