Security detection method and system based on monitoring video network transmission

Abstract

The invention relates to the technical field of digital information transmission, and discloses a network transmission security detection method and system based on a monitoring video. The method comprises the following steps: firstly, encoding a monitoring video, dividing the monitoring video into monitoring video image frames, and screening abnormal points of the monitoring video frames by using a neural network to obtain a digital signal matrix of the monitoring video image frames; performing wavelet analysis on the digital signal matrix to obtain a reconstructed digital signal matrix, and evaluating the reconstructed digital signal matrix to obtain an abnormal degree value; detecting information divergence in the digital signal matrix by using an information divergence method, and detecting divergence distribution of an information divergence matrix to obtain an information divergence value; finally, the safety of the digital signal matrix is judged through the abnormal degree value and the information divergence value, the safety of monitoring video network transmission is reflected through the safety of the digital signal matrix, and monitoring video network transmission abnormity is detected through a genetic algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of digital information transmission, and in particular to a monitoring video network transmission security detection method and system thereof. Background Art

[0002] Chinese patent CN118041692A discloses a network security testing method and system based on intrusion detection technology, the method includes: performing dimensionality reduction processing on the access time array obtained from the time when the target database is accessed within a preset time period and the call duration array obtained from the call duration when the login interface is called when the target database is accessed multiple times within the preset time period, obtaining the reduced-dimensional access time array and the reduced-dimensional call duration array for network security analysis and calculation, obtaining the database security level, combining it with the number of times the target database is accessed within the preset time period, setting a threshold for the number of suspected intrusions, and combining multiple suspected intrusion behaviors with the database security level as the network security test result of the target database. The method solves the technical problem of the lack of detection of network intrusions in the prior art, resulting in low network security, realizes reasonable and accurate intrusion monitoring, and improves network security.

[0003] Traditional surveillance video equipment cannot issue timely warnings when massive amounts of data are being transmitted, cannot guarantee the security of surveillance video network transmission, and cannot monitor whether there are any abnormalities in video network transmission. It does not convert surveillance videos into digital signals, and the surveillance videos occupy too much memory during transmission. The security detection process is cumbersome and labor-intensive, posing a threat to the safe transmission of surveillance videos. Summary of the invention

[0004] In view of the problems in the related art, the present invention provides a monitoring video network transmission security detection method and system thereof to overcome the above-mentioned technical problems existing in the existing related art.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions: The present invention is a method for monitoring video network transmission security, comprising the following steps: S1. Acquire a surveillance video, encode the surveillance video to obtain a surveillance video image frame set to be processed, use a neural network to detect abnormalities in the surveillance video image frame set to be processed, obtain a surveillance video image frame set, and generate a digital signal matrix of the surveillance video image frame from the surveillance video image frame set; S2. The digital signal matrix of the monitoring video image frame is subjected to wavelet analysis to obtain a reconstructed digital signal matrix; the reconstructed digital signal matrix is ​​evaluated in the network space distribution according to the kernel density estimation method, and the abnormality degree value obtained after evaluating the reconstructed digital signal matrix is ​​used as the basis for judging whether the digital signal matrix is ​​safe, which is recorded as the first basis; S3, using the information divergence method to detect the information divergence in the digital signal matrix of the monitoring video image frame, establishing a regression model to obtain the information divergence matrix, detecting the divergence distribution of the information divergence matrix, and comparing the information divergence value of the digital signal in the information divergence matrix with the divergence threshold to determine whether the digital signal matrix is ​​safe, which is recorded as the second basis; S4. Detect the security of the digital signal matrix of the surveillance video image frame by using the first basis and the second basis, and detect the abnormality of the surveillance video network transmission based on the heuristic genetic algorithm.

[0006] The invention first encodes the surveillance video to divide it into surveillance video image frames, uses a neural network to screen abnormal points of the surveillance video frames, converts non-abnormal points into digital signals, and obtains a digital signal matrix of the surveillance video image frames. The method of use is clever, and it is relatively convenient to study the obtained digital signal matrix in the future; then the digital signal matrix of the surveillance video image frame is subjected to wavelet analysis to obtain a reconstructed digital signal matrix, and the reconstructed digital signal matrix is ​​evaluated according to the kernel density estimation method, and the abnormality degree value obtained after evaluating the reconstructed digital signal matrix is ​​used as a basis for judging whether the digital signal matrix is ​​safe; secondly, the information divergence method is used to detect the information divergence in the digital signal matrix of the surveillance video image frame, a regression model is established to obtain the information divergence matrix, the divergence distribution of the information divergence matrix is ​​detected, and the information divergence value of the digital signal in the information divergence matrix and the divergence threshold are used as a basis for comparing and judging whether the digital signal matrix is ​​safe; the security of the digital signal matrix of the video image frame is judged by two bases, and the method is rigorous and accurate; finally, the degree of digital signal loss during the transmission of the surveillance video through the network is judged by a genetic algorithm, and the abnormality of the surveillance video network transmission is detected.

[0007] Preferably, the S1 comprises the following steps: S11, dividing the continuous video in the surveillance video into surveillance video images composed of frames, performing PCM encoding on each pixel in the surveillance video images composed of frames, digitizing the video signal, and completing the surveillance video encoding; recording the surveillance video images composed of frames as a surveillance video image frame set to be processed ,in Indicates Frame of surveillance video image to be processed; S12, using a neural network to detect abnormalities in the set of surveillance video image frames to be processed, the specific steps are as follows: S121, segmenting the surveillance video images in the surveillance video image frame set to be processed, setting the transparency of the pixels in the surveillance video images to ,and , using an energy function to reassign pixel transparency, wherein a pixel with a transparency of 1 represents a foreground pixel of the surveillance video image, and a pixel with a transparency of 0 represents a background pixel of the surveillance video image; S122, using a Gaussian density model to determine the density of foreground pixels and the density of background pixels, and completing the segmentation of the surveillance video image due to the difference between the density of the foreground pixels and the density of the background pixels; S123, setting the gray value of the foreground pixel of the monitoring video image to , Indicates the number of foreground pixels and establishes a foreground pixel location function , the calculation formula is as follows,

[0008] in, Indicates the maximum flow of surveillance video images; After locating the foreground pixel, select the Frame surveillance video image and the Frame surveillance video image, find the Frame surveillance video image and the The points with different foreground pixels in the surveillance video images of the frames are compared with the surveillance video images of adjacent frames in sequence, and the points with different foreground pixels are recorded as different point sets. ; S124. The different point sets may have anomalies, and a convolutional neural network is used to determine whether the different point sets have anomalies; the convolution kernel size in the convolutional neural network is set to 3×3, the number of convolution kernels is 128, the number of pooling layers is 1, and the learning rate is , the loss function is ; Collect different points After normalizing the different point data in the image, the normalized different point data is obtained, and the normalized different point data is input into the input layer of the convolutional neural network. After training, until convergence, the possible abnormal point is output; the probability that the possible abnormal point is an abnormal point is calculated, and the calculation formula is as follows:

[0009] in, represents the probability that a possible outlier is an outlier, Indicates normalized different point data; Setting anomaly thresholds , if the probability that the outlier is an outlier is less than , then the abnormal point may not be an abnormal point; otherwise, the abnormal point may be an abnormal point, and the surveillance video image in the surveillance video image frame set to be processed corresponding to the abnormal point is discarded to obtain the surveillance video image frame set; S13, set the monitoring video image frame set to ,in Indicates The monitoring video image of the frame; the digital signal set corresponding to the first monitoring video image in the monitoring video image frame set is , generate the digital signal matrix of the surveillance video image frame , , in, Indicates The surveillance video image of the frame corresponds to the A digital signal.

[0010] Preferably, S2 comprises the following steps: S21, the digital signal matrix of the monitoring video image frame is obtained by wavelet analysis to reconstruct the digital signal matrix, and the specific steps are as follows: S211, setting the transient signal set of the digital signal matrix ,in Indicates A transient signal; set represents the time sampling basis, represents the orthogonal wavelet basis constituting the transient signal set, and the calculation formula is as follows:

[0011] Calculate the other orthogonal wavelet bases of the transient signal set in sequence, denoted as the orthogonal wavelet base set ,in The transient signal set orthogonal wavelet bases; S212, using a multi-resolution analysis method to perform abnormal signal detection on the orthogonal wavelet basis set, setting any orthogonal wavelet basis in the orthogonal wavelet basis set, denoted as , For about The orthogonal complementary space of the transient signal set is set as , the transient signal in the transient signal matrix is ​​calculated based on the orthogonal wavelet basis. The calculation formula is as follows:

[0012] in, Indicates that the transient signal in the transient signal matrix is The projection on Indicates that the transient signal in the transient signal matrix is projection on; All transient signals in the transient signal set are calculated in sequence, and the transient signal matrix as follows, , in, Indicates The surveillance video image of the frame corresponds to the The transient signal of an orthogonal wavelet basis; S213, setting the orthogonal decomposition time parameter in the transient signal matrix to ,and , and represents the scaling function parameters of the transient signal matrix, Represents the scaling function of the transient signal matrix, and calculates the orthogonal wavelet packets of two adjacent transient signals of the transient signal matrix and , the calculation formula is as follows,

[0013]

[0014] Decompose and calculate the orthogonal wavelet basis of all adjacent signals in the transient signal matrix to obtain an orthogonal wavelet basis matrix, which is the reconstructed digital signal matrix. , , in, Indicates The surveillance video image of the frame corresponds to the Reconstructed digital signal based on orthogonal wavelet basis; S22, taking a signal sample set from the reconstructed digital signal matrix ,in represents the number of signal samples and , set the reconstructed digital signal matrix as a two-dimensional network space, calculate the kernel density of the signal sample set in the two-dimensional network space, the calculation formula is as follows,

[0015] in, represents the kernel density, represents the kernel function and , ; The kernel density of the signal sample set is calculated, and the kernel density of the signal sample set is set as the kernel density of a signal sample of a network communication link in the two-dimensional network space. The signal sample set is calculated in sequence The kernel density of the group signal sample set constitutes the network communication link in the two-dimensional network space Kernel density of signal samples; Calculate network communication links in two-dimensional network space Abnormality value of the kernel density of the signal sample , the calculation formula is as follows,

[0016] S23, set the abnormality threshold to , when the abnormality value is less than or equal to When , the network communication link in the two-dimensional network space is secure, that is, the digital signal matrix is ​​secure; otherwise, the digital signal matrix is ​​not secure.

[0017] The invention monitors the transient signal set of the digital signal matrix of the video image frame, obtains the transient signal matrix after processing with an orthogonal wavelet basis, and decomposes the transient signal matrix to obtain a reconstructed digital signal matrix; a network space is established based on the reconstructed digital signal matrix, and the sample kernel density of each network communication link is calculated in the network space to obtain the abnormality degree value of the sample kernel density. The abnormality degree value is used as a basis for judging whether the digital signal matrix is ​​safe. This method has high detection efficiency and small error.

[0018] Preferably, S3 comprises the following steps: S31, using the information divergence method to detect the information divergence in the digital signal matrix of the monitoring video image frame, the specific steps are as follows: S311, setting the digital signal matrix to represent the discrete space, represents the total probability distribution of the discrete space, express Traffic properties at the time, express Traffic properties at the time, express The distribution probability of express The distribution probability of the information divergence method is used to calculate the traffic similarity in the discrete space , Represents the order, and the calculation formula is as follows:

[0019] The length of the digital signal of the traffic similarity in the discrete space is set to , the number of digital signals of traffic similarity in the discrete space is , The variation range of the length of the digital signal representing the traffic similarity in the discrete space is then the traffic distribution probability in the discrete space , the calculation formula is as follows,

[0020] S312: A regression model is established based on the traffic similarity in the discrete space and the traffic distribution probability in the discrete space to detect changes in information divergence in the discrete space. The regression model as follows,

[0021] in, represents the range of information divergence in discrete space, represents a time series vector; Calculate the information divergence of the digital signal in the digital signal matrix, and obtain the information divergence matrix through the regression model , , in, Indicates The surveillance video image of the frame corresponds to the The information dispersion of a digital signal; S32, Setting Represents a time interval, a time series collection ,in Indicates time series, the information divergence of the digital signal in the information divergence matrix varies within a range of , Indicates the range of IP address changes. represents the distribution vector of IP addresses, represents the distribution vector of the port, Indicates the port range. Represents the distribution probability of the information divergence of the digital signal in the information divergence matrix, and calculates the divergence distribution vector of the information divergence matrix , the calculation formula is as follows,

[0022] The number of occurrences of digital signals in the information divergence matrix is ​​counted, and the information divergence value of the digital signal in the information divergence matrix is ​​calculated according to the divergence distribution vector of the information divergence matrix. , the calculation formula is as follows,

[0023] in, represents the adjacent distribution vector of digital signals in the information divergence matrix, represents the order of the regression model; S33, setting the divergence threshold to , when the information divergence value of the digital signal in the information divergence matrix is ​​less than or equal to When , the information divergence value of the digital signal in the information divergence matrix meets the digital signal matrix security condition, that is, the digital signal matrix is ​​secure; otherwise, the digital signal matrix is ​​not secure.

[0024] The invention represents discrete space with a digital signal matrix, calculates digital signals with the same traffic similarity in the discrete space, establishes a regression model using traffic similarity, describes spatial divergence in discrete space, and obtains an information divergence matrix. The information divergence value of the digital signal generated by the information divergence matrix is ​​used as a basis for judging whether the digital signal matrix is ​​secure, which can effectively meet the requirements of traffic transmission security detection for its detection accuracy under network abnormalities.

[0025] Preferably, S4 comprises the following steps: S41, during the monitoring video transmission process, The digital signal corresponding to the surveillance video image of the frame generates a digital signal matrix of the surveillance video image frame, and the abnormality degree value is used as the first basis for detecting the security of the digital signal matrix of the surveillance video image frame; the information divergence value of the digital signal in the information divergence matrix is ​​used as the second basis for detecting the security of the digital signal matrix of the surveillance video image frame; when the abnormality degree value is less than or equal to the abnormality degree threshold and the information divergence value of the digital signal in the information divergence matrix is ​​less than or equal to the divergence threshold, the digital signal matrix is ​​safe, and the surveillance video network transmission is safe at this time; otherwise, the surveillance video network transmission is unsafe; S42, using a genetic algorithm to extract abnormal signals in the digital signal matrix, the specific steps are as follows: S421, take the anomaly detection data set in the digital signal matrix as ,in Indicates Anomaly detection data, set The distribution probability of anomaly detection data at each moment is , The resilience of the anomaly detection data at each moment is ,but The distribution probability calculation formula of moment anomaly detection data is as follows:

[0026] The first abnormal probability data set is obtained from the distribution probability of the abnormal detection data ,in Indicates the first abnormal probability data set The anomaly probability of the anomaly data to be detected; Assume that the anomaly data set to be detected is ,in Indicates anomaly data to be detected, and calculate the anomaly probability of all anomaly data to be detected in the anomaly data set to be detected, which is recorded as the second anomaly probability data set ,in Indicates the second abnormal probability data set The abnormal probability of the abnormal data to be detected; S422, set the abnormal probability threshold , calculate the absolute deviation of the first abnormal probability data set and the second abnormal probability data set to obtain an absolute deviation set. When the absolute deviation in the absolute deviation set is less than or equal to When , the corresponding abnormal probability data in the absolute deviation set is taken out to form an abnormal data set, which is recorded as the abnormal signal set ,in Indicates An abnormal signal and ; S43, the abnormal signal set , set the first signal clustering center of the abnormal signal set to , calculate the cosine distance from the first signal cluster center of the abnormal signal set to the abnormal signal in the abnormal signal set to obtain a first cosine distance set; then select the abnormal signal corresponding to the largest cosine distance from the first cosine distance set as the second signal cluster center of the abnormal signal set ; Repeat iteration and set the cosine distance threshold , until the difference between the signal cluster centers of adjacent abnormal signal sets is less than , stop the iteration and get the final signal clustering center of the abnormal signal set ; S44, obtaining an abnormal signal matrix of a digital signal matrix from the abnormal signal set , , in, Indicates The surveillance video image of the frame corresponds to the An abnormal digital signal; The final signal clustering center set of the digital signal matrix is ​​obtained from the final signal clustering center of the abnormal signal set, and the distance between the final signal clustering center set and the abnormal signal in the abnormal signal matrix is ​​calculated to obtain the abnormal signal distance matrix; in the fitness adjustment process of the genetic algorithm, different weights are given to the abnormal signal distances in the abnormal signal distance matrix, and the variance matrix of the signal distance matrix is ​​calculated. , , in, Indicates The surveillance video image of the frame corresponds to the distance variance; Set the variance matrix Line The column variance weight is , Indicates the variance matrix Line The variance of the column, Represents the input value, fitness function The calculation formula is as follows,

[0027] S45. During the transmission of the monitoring video through the network, the digital signal in the digital signal matrix will be lost at each network node. The probability that the digital signal in the digital signal matrix will be lost is , expected value ; The digital signal in the digital signal matrix is ​​lost at the expected value, and the threshold is set , when the digital signal in the digital signal matrix is ​​less than , the surveillance video network transmission is abnormal; otherwise, the surveillance video network transmission is normal.

[0028] The invention uses the abnormality degree value and the information divergence value as the basis for detecting the security of the digital signal matrix of the monitoring video image frame. The evaluation of the two bases makes the result accurate and can fully verify the security of the monitoring video network transmission. The genetic algorithm is then used to judge the degree of digital signal loss during the transmission of the monitoring video through the network, detect the abnormality of the monitoring video network transmission, and use the genetic algorithm to optimize the traditional anomaly detection algorithm. The anomaly detection function is realized through multiple rounds of iterations, thereby reducing errors and improving accuracy.

[0029] The present invention also discloses a system based on the monitoring video network transmission security detection method, which specifically includes: a monitoring video generation digital signal matrix module, an evaluation and reconstruction digital signal matrix abnormality value module, an information dispersion matrix information dispersion value determination module and a monitoring video network transmission abnormality detection module; The monitoring video digital signal matrix generation module is used to screen abnormal points of the monitoring video image frame of the monitoring video to obtain a digital signal matrix; The module for evaluating the abnormality degree value of the reconstructed digital signal matrix is ​​used to determine the abnormality degree value of the monitoring video network transmission after obtaining the reconstructed digital signal matrix through wavelet analysis; The information dispersion matrix information dispersion value determination module is used to detect the information dispersion in the digital signal matrix to determine the information dispersion of the monitoring video network transmission; The monitoring video network transmission anomaly detection module is used to monitor anomalies in the video network transmission process through a genetic algorithm.

[0030] The present invention has the following beneficial effects: 1. The invention first encodes the surveillance video to divide it into surveillance video image frames, uses a neural network to screen abnormal points of the surveillance video frames, converts non-abnormal points into digital signals, and obtains a digital signal matrix of the surveillance video image frames. The method of use is ingenious, and it is convenient to study the obtained digital signal matrix in the future; then the digital signal matrix of the surveillance video image frame is subjected to wavelet analysis to obtain a reconstructed digital signal matrix, and the reconstructed digital signal matrix is ​​evaluated according to the kernel density estimation method, and the abnormality degree value obtained after evaluating the reconstructed digital signal matrix is ​​used as a basis for judging whether the digital signal matrix is ​​safe; secondly, the information divergence method is used to detect the information divergence in the digital signal matrix of the surveillance video image frame, a regression model is established to obtain the information divergence matrix, the divergence distribution of the information divergence matrix is ​​detected, and the information divergence value of the digital signal in the information divergence matrix and the divergence threshold are used as a basis for comparing and judging whether the digital signal matrix is ​​safe; the security of the digital signal matrix of the video image frame is judged by two bases, and the method is rigorous and accurate; finally, the degree of digital signal loss during the transmission of the surveillance video through the network is judged by a genetic algorithm, and the abnormality of the surveillance video network transmission is detected.

[0031] 2. The invention obtains a transient signal matrix by monitoring the transient signal set of the digital signal matrix of the video image frame, processes it with an orthogonal wavelet basis, and decomposes the transient signal matrix to obtain a reconstructed digital signal matrix; a network space is established based on the reconstructed digital signal matrix, and the sample kernel density of each network communication link is calculated in the network space to obtain the abnormality degree value of the sample kernel density. The abnormality degree value is used as a basis for judging whether the digital signal matrix is ​​safe. This method has high detection efficiency and small error.

[0032] 3. The invention represents discrete space by using a digital signal matrix, calculates digital signals with the same traffic similarity in the discrete space, establishes a regression model using traffic similarity, describes spatial divergence in discrete space, and obtains an information divergence matrix. The information divergence value of the digital signal generated by the information divergence matrix is ​​used as a basis for judging whether the digital signal matrix is ​​safe, which can effectively meet the requirements of traffic transmission security detection under network abnormalities for its detection accuracy.

[0033] 4. The invention uses the abnormality degree value and the information divergence value as the basis for detecting the security of the digital signal matrix of the monitoring video image frame. The evaluation of the two bases makes the result accurate and can fully verify the security of the monitoring video network transmission. The genetic algorithm is then used to judge the degree of digital signal loss during the transmission of the monitoring video through the network, detect the abnormality of the monitoring video network transmission, and use the genetic algorithm to optimize the traditional anomaly detection algorithm. The anomaly detection function is realized through multiple rounds of iterations, thereby reducing errors and improving accuracy.

[0034] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying creative work.

[0036] Figure 1 The present invention provides a flow chart of monitoring video network transmission security detection based on the monitoring video network transmission security detection method. DETAILED DESCRIPTION

[0037] 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.

[0038] In the description of the present invention, it is necessary to understand that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.

[0039] This implementation discloses a method for detecting network transmission security based on surveillance video, which specifically includes the following contents: S1. Acquire a surveillance video, encode the surveillance video to obtain a surveillance video image frame set to be processed, use a neural network to detect abnormalities in the surveillance video image frame set to be processed, obtain a surveillance video image frame set, and generate a digital signal matrix of the surveillance video image frame from the surveillance video image frame set; The S1 comprises the following steps: S11, dividing the continuous video in the surveillance video into surveillance video images composed of frames, performing PCM encoding on each pixel in the surveillance video images composed of frames, digitizing the video signal, and completing the surveillance video encoding; recording the surveillance video images composed of frames as a surveillance video image frame set to be processed ,in Indicates Frame of surveillance video image to be processed; S12, using a neural network to detect abnormalities in the set of surveillance video image frames to be processed, the specific steps are as follows: S121, segmenting the surveillance video images in the surveillance video image frame set to be processed, setting the transparency of the pixels in the surveillance video images to ,and , using an energy function to reassign pixel transparency, wherein a pixel with a transparency of 1 represents a foreground pixel of the surveillance video image, and a pixel with a transparency of 0 represents a background pixel of the surveillance video image; S122, using a Gaussian density model to determine the density of foreground pixels and the density of background pixels, and completing the segmentation of the surveillance video image due to the difference between the density of the foreground pixels and the density of the background pixels; S123, setting the gray value of the foreground pixel of the monitoring video image to , Indicates the number of foreground pixels and establishes a foreground pixel location function , the calculation formula is as follows,

[0040] in, Indicates the maximum flow of surveillance video images; After locating the foreground pixel, select the Frame surveillance video image and the Frame surveillance video image, find the Frame surveillance video image and the The points with different foreground pixels in the surveillance video images of the frames are compared with the surveillance video images of adjacent frames in sequence, and the points with different foreground pixels are recorded as different point sets. ; S124. The different point sets may have anomalies, and a convolutional neural network is used to determine whether the different point sets have anomalies; the convolution kernel size in the convolutional neural network is set to 3×3, the number of convolution kernels is 128, the number of pooling layers is 1, and the learning rate is , the loss function is ; Collect different points After normalizing the different point data in the image, the normalized different point data is obtained, and the normalized different point data is input into the input layer of the convolutional neural network. After training, until convergence, the possible abnormal point is output; the probability that the possible abnormal point is an abnormal point is calculated, and the calculation formula is as follows:

[0041] in, represents the probability that a possible outlier is an outlier, Indicates normalized different point data; Setting anomaly thresholds , if the probability that the outlier is an outlier is less than , then the abnormal point may not be an abnormal point; otherwise, the abnormal point may be an abnormal point, and the surveillance video image in the surveillance video image frame set to be processed corresponding to the abnormal point is discarded to obtain the surveillance video image frame set; S13, set the monitoring video image frame set to ,in Indicates The monitoring video image of the frame; the digital signal set corresponding to the first monitoring video image in the monitoring video image frame set is , generate the digital signal matrix of the surveillance video image frame , , in, Indicates The surveillance video image of the frame corresponds to the A digital signal; S2. The digital signal matrix of the monitoring video image frame is subjected to wavelet analysis to obtain a reconstructed digital signal matrix; the reconstructed digital signal matrix is ​​evaluated in the network space distribution according to the kernel density estimation method, and the abnormality degree value obtained after evaluating the reconstructed digital signal matrix is ​​used as the basis for judging whether the digital signal matrix is ​​safe, which is recorded as the first basis; The S2 comprises the following steps: S21, the digital signal matrix of the monitoring video image frame is obtained by wavelet analysis to reconstruct the digital signal matrix, and the specific steps are as follows: S211, setting the transient signal set of the digital signal matrix ,in Indicates A transient signal; set represents the time sampling basis, represents the orthogonal wavelet basis constituting the transient signal set, and the calculation formula is as follows:

[0042] Calculate the other orthogonal wavelet bases of the transient signal set in sequence, denoted as the orthogonal wavelet base set ,in The transient signal set orthogonal wavelet bases; S212, using a multi-resolution analysis method to perform abnormal signal detection on the orthogonal wavelet basis set, setting any orthogonal wavelet basis in the orthogonal wavelet basis set, denoted as , For about The orthogonal complementary space of the transient signal set is set as , the transient signal in the transient signal matrix is ​​calculated based on the orthogonal wavelet basis. The calculation formula is as follows:

[0043] in, Indicates that the transient signal in the transient signal matrix is The projection on Indicates that the transient signal in the transient signal matrix is projection on; All transient signals in the transient signal set are calculated in sequence, and the transient signal matrix as follows, , in, Indicates The surveillance video image of the frame corresponds to the The transient signal of an orthogonal wavelet basis; S213, setting the orthogonal decomposition time parameter in the transient signal matrix to ,and , and represents the scaling function parameters of the transient signal matrix, Represents the scaling function of the transient signal matrix, and calculates the orthogonal wavelet packets of two adjacent transient signals of the transient signal matrix and , the calculation formula is as follows,

[0044]

[0045] Decompose and calculate the orthogonal wavelet basis of all adjacent signals in the transient signal matrix to obtain an orthogonal wavelet basis matrix, which is the reconstructed digital signal matrix. , , in, Indicates The surveillance video image of the frame corresponds to the Reconstructed digital signal based on orthogonal wavelet basis; S22, taking a signal sample set from the reconstructed digital signal matrix ,in represents the number of signal samples and , set the reconstructed digital signal matrix as a two-dimensional network space, calculate the kernel density of the signal sample set in the two-dimensional network space, the calculation formula is as follows,

[0046] in, represents the kernel density, represents the kernel function and , ; The kernel density of the signal sample set is calculated, and the kernel density of the signal sample set is set as the kernel density of a signal sample of a network communication link in the two-dimensional network space. The signal sample set is calculated in sequence The kernel density of the group signal sample set constitutes the network communication link in the two-dimensional network space Kernel density of signal samples; Calculate network communication links in two-dimensional network space Abnormality value of the kernel density of the signal sample , the calculation formula is as follows, S23, set the abnormality threshold to , when the abnormality value is less than or equal to When , the network communication link in the two-dimensional network space is secure, that is, the digital signal matrix is ​​secure; otherwise, the digital signal matrix is ​​not secure; S3, using the information divergence method to detect the information divergence in the digital signal matrix of the monitoring video image frame, establishing a regression model to obtain the information divergence matrix, detecting the divergence distribution of the information divergence matrix, and comparing the information divergence value of the digital signal in the information divergence matrix with the divergence threshold to determine whether the digital signal matrix is ​​safe, which is recorded as the second basis; The S3 comprises the following steps: S31, using the information divergence method to detect the information divergence in the digital signal matrix of the monitoring video image frame, the specific steps are as follows: S311, setting the digital signal matrix to represent the discrete space, represents the total probability distribution of the discrete space, express Traffic properties at the time, express Traffic properties at the time, express The distribution probability of express The distribution probability of the information divergence method is used to calculate the traffic similarity in the discrete space , Represents the order, and the calculation formula is as follows:

[0047] The length of the digital signal of the traffic similarity in the discrete space is set to , the number of digital signals of traffic similarity in the discrete space is , The variation range of the length of the digital signal representing the traffic similarity in the discrete space is then the traffic distribution probability in the discrete space , the calculation formula is as follows,

[0048] S312: A regression model is established based on the traffic similarity in the discrete space and the traffic distribution probability in the discrete space to detect changes in information divergence in the discrete space. The regression model as follows,

[0049] in, represents the range of information divergence in discrete space, represents a time series vector; Calculate the information divergence of the digital signal in the digital signal matrix, and obtain the information divergence matrix through the regression model , , in, Indicates The surveillance video image of the frame corresponds to the The information dispersion of a digital signal; S32, Setting Represents a time interval, a time series collection ,in Indicates time series, the information divergence of the digital signal in the information divergence matrix varies within a range of , Indicates the range of IP address changes. represents the distribution vector of IP addresses, represents the distribution vector of the port, Indicates the port range. Represents the distribution probability of the information divergence of the digital signal in the information divergence matrix, and calculates the divergence distribution vector of the information divergence matrix , the calculation formula is as follows,

[0050] The number of occurrences of digital signals in the information divergence matrix is ​​counted, and the information divergence value of the digital signal in the information divergence matrix is ​​calculated according to the divergence distribution vector of the information divergence matrix. , the calculation formula is as follows,

[0051] in, represents the adjacent distribution vector of digital signals in the information divergence matrix, represents the order of the regression model; S33, setting the divergence threshold to , when the information divergence value of the digital signal in the information divergence matrix is ​​less than or equal to When , the information divergence value of the digital signal in the information divergence matrix meets the digital signal matrix security condition, that is, the digital signal matrix is ​​secure; otherwise, the digital signal matrix is ​​unsafe; S4. Detecting the security of the digital signal matrix of the surveillance video image frame by using the first basis and the second basis, and detecting the abnormality of surveillance video network transmission based on the heuristic genetic algorithm; The S4 comprises the following steps: S41, during the monitoring video transmission process, The digital signal corresponding to the surveillance video image of the frame generates a digital signal matrix of the surveillance video image frame, and the abnormality degree value is used as the first basis for detecting the security of the digital signal matrix of the surveillance video image frame; the information divergence value of the digital signal in the information divergence matrix is ​​used as the second basis for detecting the security of the digital signal matrix of the surveillance video image frame; when the abnormality degree value is less than or equal to the abnormality degree threshold and the information divergence value of the digital signal in the information divergence matrix is ​​less than or equal to the divergence threshold, the digital signal matrix is ​​safe, and the surveillance video network transmission is safe at this time; otherwise, the surveillance video network transmission is unsafe; S42, using a genetic algorithm to extract abnormal signals in the digital signal matrix, the specific steps are as follows: S421, take the anomaly detection data set in the digital signal matrix as ,in Indicates Anomaly detection data, set The distribution probability of anomaly detection data at each moment is , The resilience of the anomaly detection data at each moment is ,but The distribution probability calculation formula of moment anomaly detection data is as follows:

[0052] The first abnormal probability data set is obtained from the distribution probability of the abnormal detection data ,in Indicates the first abnormal probability data set The anomaly probability of the anomaly data to be detected; Assume that the anomaly data set to be detected is ,in Indicates anomaly data to be detected, and calculate the anomaly probability of all anomaly data to be detected in the anomaly data set to be detected, which is recorded as the second anomaly probability data set ,in Indicates the second abnormal probability data set The abnormal probability of the abnormal data to be detected; S422, set the abnormal probability threshold , calculate the absolute deviation of the first abnormal probability data set and the second abnormal probability data set to obtain an absolute deviation set. When the absolute deviation in the absolute deviation set is less than or equal to When , the corresponding abnormal probability data in the absolute deviation set is taken out to form an abnormal data set, which is recorded as the abnormal signal set ,in Indicates An abnormal signal and ; S43, the abnormal signal set , set the first signal clustering center of the abnormal signal set to , calculate the cosine distance from the first signal cluster center of the abnormal signal set to the abnormal signal in the abnormal signal set to obtain a first cosine distance set; then select the abnormal signal corresponding to the largest cosine distance from the first cosine distance set as the second signal cluster center of the abnormal signal set ; Repeat iteration and set the cosine distance threshold , until the difference between the signal cluster centers of adjacent abnormal signal sets is less than , stop the iteration and get the final signal clustering center of the abnormal signal set ; S44, obtaining an abnormal signal matrix of a digital signal matrix from the abnormal signal set , , in, Indicates The surveillance video image of the frame corresponds to the An abnormal digital signal; The final signal clustering center set of the digital signal matrix is ​​obtained from the final signal clustering center of the abnormal signal set, and the distance between the final signal clustering center set and the abnormal signal in the abnormal signal matrix is ​​calculated to obtain the abnormal signal distance matrix; in the fitness adjustment process of the genetic algorithm, different weights are given to the abnormal signal distances in the abnormal signal distance matrix, and the variance matrix of the signal distance matrix is ​​calculated. , , in, Indicates The surveillance video image of the frame corresponds to the The distance variance; Set the variance matrix Line The column variance weight is , Indicates the variance matrix Line The variance of the column, Represents the input value, fitness function The calculation formula is as follows,

[0053] S45. During the transmission of the monitoring video through the network, the digital signal in the digital signal matrix will be lost at each network node. The probability that the digital signal in the digital signal matrix will be lost is , expected value ; The digital signal in the digital signal matrix is ​​lost at the expected value, and the threshold is set , when the digital signal in the digital signal matrix is ​​less than , the surveillance video network transmission is abnormal; otherwise, the surveillance video network transmission is normal.

[0054] This embodiment also discloses a security detection system based on surveillance video network transmission, which specifically includes: a surveillance video generation digital signal matrix module, an evaluation and reconstruction digital signal matrix abnormality value module, an information dispersion matrix information dispersion value determination module and a surveillance video network transmission abnormality detection module; The monitoring video digital signal matrix generation module is used to screen abnormal points of the monitoring video image frame of the monitoring video to obtain a digital signal matrix; The module for evaluating the abnormality degree value of the reconstructed digital signal matrix is ​​used to determine the abnormality degree value of the monitoring video network transmission after obtaining the reconstructed digital signal matrix through wavelet analysis; The information dispersion matrix information dispersion value determination module is used to detect the information dispersion in the digital signal matrix to determine the information dispersion of the monitoring video network transmission; The monitoring video network transmission anomaly detection module is used to monitor anomalies in the video network transmission process through a genetic algorithm.

[0055] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0056] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can understand and use the invention well.

Claims

1. A method for detecting network transmission security based on surveillance video, characterized in that: The steps include: S1. Acquire a surveillance video to obtain a surveillance video image frame set to be processed, use a neural network to detect abnormal points of the surveillance video image frame set to be processed, discard the abnormal points to obtain a surveillance video image frame set, and generate a digital signal matrix of the surveillance video image frame from the surveillance video image frame set; S2. The digital signal matrix of the monitoring video image frame is obtained by wavelet analysis to obtain a reconstructed digital signal matrix, and the reconstructed digital signal matrix is ​​evaluated according to the kernel density estimation method. The abnormality degree value obtained after evaluating the reconstructed digital signal matrix is ​​used as the basis for judging whether the digital signal matrix of the monitoring video image frame is safe, which is recorded as the first basis; S3, using the information divergence method to detect the information divergence in the digital signal matrix of the monitoring video image frame, establishing a regression model to obtain the information divergence matrix, and using the information divergence value of the digital signal in the information divergence matrix as the basis for judging whether the digital signal matrix of the monitoring video image frame is safe, recorded as the second basis; S4. Detect the security of the digital signal matrix of the monitoring video image frame through the first basis and the second basis, and detect abnormalities in the monitoring video network transmission.

2. The method for detecting the security of surveillance video network transmission according to claim 1, characterized in that: The S1 comprises the following steps: S11, dividing the continuous video in the surveillance video into surveillance video images composed of frames, and recording the surveillance video images composed of the frames as a surveillance video image frame set to be processed; S12, using a neural network to detect abnormal points of the surveillance video image frame set to be processed, and discarding the abnormal points to obtain the surveillance video image frame set; S13, generating a digital signal matrix of monitoring video image frames from the monitoring video image frame set.

3. The method for detecting the security of surveillance video network transmission according to claim 2 is characterized in that: The S12 comprises the following steps: S121, segmenting the surveillance video images in the surveillance video image frame set to be processed; S122, comparing surveillance video images of adjacent frames in sequence, and recording points with different foreground pixels as different point sets; S123, the different point sets may have anomalies, and a convolutional neural network is used to determine whether the different point sets have anomalies; the convolution kernel size in the convolutional neural network is set to 3×3, the number of convolution kernels is 128, the number of pooling layers is 1, and the learning rate is , the loss function is ; Normalize the different point data in the different point set to obtain normalized different point data, input the normalized different point data into the convolutional neural network input layer, and after training until convergence, output possible abnormal points; calculate the probability that the possible abnormal point is an abnormal point; Setting anomaly thresholds , if the probability that the outlier is an outlier is less than , then the abnormal point may not be an abnormal point; otherwise, the abnormal point may be an abnormal point, and the surveillance video image in the surveillance video image frame set to be processed corresponding to the abnormal point is discarded to obtain the surveillance video image frame set.

4. The method for detecting the security of surveillance video network transmission according to claim 3 is characterized in that: The S2 comprises the following steps: S21, the digital signal matrix of the monitoring video image frame is obtained by wavelet analysis to reconstruct the digital signal matrix; S22. Take a signal sample set from the reconstructed digital signal matrix, calculate the signal sample kernel density of the signal sample set, obtain the abnormality degree value of the signal sample kernel density of the network communication link in the two-dimensional network space, and use the abnormality degree value as the basis for judging whether the digital signal matrix of the monitoring video image frame is safe, which is recorded as the first basis.

5. The method for detecting the security of surveillance video network transmission according to claim 4 is characterized in that: Set the abnormality threshold to , when the abnormality value is less than or equal to When , the network communication link in the two-dimensional network space is secure, indicating that the digital signal matrix of the monitoring video image frame is secure; Otherwise the digital signal matrix of the monitoring video image frame is not safe.

6. The method for monitoring video network transmission security detection according to claim 5 is characterized in that: The S3 comprises the following steps: S31, using the information divergence method to detect the information divergence in the digital signal matrix of the monitoring video image frame to obtain an information divergence matrix; S32. Calculate the information divergence value of the digital signal in the information divergence matrix, and use the information divergence value as a basis for judging whether the digital signal matrix of the monitoring video image frame is safe, which is recorded as the second basis.

7. The method for detecting the security of surveillance video network transmission according to claim 6 is characterized in that: Set the divergence threshold to , when the information divergence value of the digital signal in the information divergence matrix is ​​less than or equal to When , the information divergence value of the digital signal in the information divergence matrix meets the digital signal matrix security condition of the monitoring video image frame, indicating that the digital signal matrix of the monitoring video image frame is safe; Otherwise the digital signal matrix of the monitoring video image frame is not safe.

8. The method for detecting the security of surveillance video network transmission according to claim 7 is characterized in that: The S4 comprises the following steps: S41, detecting the security of the digital signal matrix of the monitoring video image frame by using the first basis and the second basis; S42, using a genetic algorithm to extract abnormal signals in the digital signal matrix of the monitoring video image frame, the specific steps are as follows: S421, taking the abnormal detection data set in the digital signal matrix of the monitoring video image frame as ,in Indicates Anomaly detection data, the first anomaly probability data set is obtained by the distribution probability of the anomaly detection data ,in Indicates the first abnormal probability data set The anomaly probability of the anomaly data to be detected; Assume that the anomaly data set to be detected is ,in Indicates anomaly data to be detected, and calculate the anomaly probability of all anomaly data to be detected in the anomaly data set to be detected, which is recorded as the second anomaly probability data set ,in Indicates the second abnormal probability data set The abnormal probability of the abnormal data to be detected; S422, set the abnormal probability threshold , calculate the absolute deviation of the first abnormal probability data set and the second abnormal probability data set to obtain an absolute deviation set, and set the absolute deviation in the absolute deviation set to be less than or equal to The corresponding abnormal probability data is taken out to form an abnormal signal set ,in Indicates An abnormal signal and ; S43, the abnormal signal set , set the first signal clustering center of the abnormal signal set to , calculate the cosine distance from the first signal cluster center of the abnormal signal set to each abnormal signal in the abnormal signal set to obtain a first cosine distance set; then select the abnormal signal corresponding to the largest first cosine distance from the first cosine distance set as the second signal cluster center of the abnormal signal set ; Repeat the iteration and set the cosine distance threshold , until the difference between the signal cluster centers of adjacent abnormal signal sets is less than , stop the iteration and get the final signal clustering center of the abnormal signal set ; S44, obtaining an abnormal signal matrix of a digital signal matrix of a monitoring video image frame from the abnormal signal set, obtaining a final signal clustering center set of a digital signal matrix from the final signal clustering center of the abnormal signal set, calculating the distances from the final signal clustering center set to each abnormal signal in the abnormal signal matrix, and obtaining an abnormal signal distance matrix; in the fitness adjustment process of the genetic algorithm, assigning different weights to the abnormal signal distances in the abnormal signal distance matrix, and calculating the variance matrix of the signal distance matrix; S45, calculating the fitness function by the variance matrix , during the transmission of the monitoring video through the network, every time it reaches a network node, the digital signal in the digital signal matrix of the monitoring video image frame will be lost. The probability that the digital signal in the digital signal matrix of the monitoring video image frame will be lost is , expected value ; The digital signal in the digital signal matrix of the monitoring video image frame is lost at the expected value, and the threshold is set , when the digital signal in the digital signal matrix of the monitoring video image frame is less than , monitoring video network transmission abnormality; Otherwise, there is no abnormality in the surveillance video network transmission.

9. A system for implementing the method for detecting security of surveillance video network transmission according to any one of claims 1 to 8, characterized in that: Specifically include: A monitoring video digital signal matrix generation module, a module for evaluating and reconstructing the abnormality degree value of the digital signal matrix, an information divergence matrix information divergence value determination module, and a monitoring video network transmission abnormality detection module; The monitoring video digital signal matrix generation module is used to screen abnormal points of the monitoring video image frame set of the monitoring video to obtain a digital signal matrix of the monitoring video image frame; The module for evaluating the abnormality degree value of the reconstructed digital signal matrix is ​​used to determine the abnormality degree value of the monitoring video network transmission after obtaining the reconstructed digital signal matrix through wavelet analysis; The information divergence matrix information divergence value determination module is used to detect the information divergence in the digital signal matrix of the monitoring video image frame to determine the information divergence of the monitoring video network transmission; The monitoring video network transmission anomaly detection module is used to monitor anomalies in the video network transmission process through a genetic algorithm.

Images