A behavior analysis method and system based on security monitoring images

By identifying and intercepting the monitoring user images in security monitoring image analysis, and optimizing the behavior analysis network, the problem of poor behavior analysis reliability in the prior art is solved, and higher analysis accuracy and reliability are achieved.

CN117115744BActive Publication Date: 2024-11-08GUANGZHOU CHANGHUI INFORMATION TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311160959.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-09-08
Publication Date
2024-11-08
Estimated Expiration
2043-09-08

AI Technical Summary

Technical Problem

In the prior art, when directly performing behavioral analysis on security monitoring images, the reliability of behavioral analysis is poor, and there is a problem of interfering with information and affecting the analysis results.

Method used

By collecting images in the area to be monitored, identifying and intercepting the monitoring user images, forming the monitoring image sequence to be analyzed, and optimizing the behavior analysis network to form a target behavior analysis network, using this network to perform behavior analysis, and output abnormal user behavior results. .

Benefits of technology

It improves the reliability of behavioral analysis, reduces the interference of invalid information on the analysis, and enhances the accuracy of the analysis results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117115744B_ABST
    Figure CN117115744B_ABST
Patent Text Reader

Abstract

The present invention provides a behavior analysis method and system based on security monitoring images, which relate to the field of artificial intelligence technology. In the present invention, image acquisition is performed in the monitored area to form a set of monitoring images to be processed; monitoring user identification processing is performed on each frame of monitoring images to be processed respectively to determine each frame of monitoring images to be processed with monitoring users in multiple frames of monitoring images to be processed, and monitoring user interception processing is performed on the determined monitoring images to be processed to form monitoring images to be analyzed, and the monitoring images to be analyzed are sorted to form a monitoring image sequence to be analyzed; network optimization processing is performed on the behavior analysis network to be optimized to form a target behavior analysis network; the target behavior analysis network is used to perform behavior analysis on the monitoring image sequence to be analyzed, and the behavior analysis results are output. Based on the above method, the reliability of behavior analysis can be improved to a certain extent.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a behavior analysis method and system based on security monitoring images. Background Art

[0002] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or digital computer-controlled computing to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence.

[0003] Artificial intelligence is the study of the design principles and implementation methods of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making. Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level technology and software-level technology.

[0004] In the field of security monitoring, it is an important means to analyze the behavior of the monitored users in the collected monitoring images, such as abnormal behavior analysis. However, in the prior art, generally after the front end collects the monitoring images, the back-end equipment directly performs behavior analysis on the front-end monitoring images, such as analysis based on artificial intelligence technology. Although artificial intelligence technology is relatively mature, directly analyzing the monitoring images will also make the reliability of behavior analysis poor. Summary of the invention

[0005] In view of this, an object of the present invention is to provide a behavior analysis method and system based on security monitoring images, so as to improve the reliability of behavior analysis to a certain extent.

[0006] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0007] A behavior analysis method based on security monitoring images, comprising:

[0008] Performing image acquisition in the area to be monitored to form a set of monitored images to be processed corresponding to the area to be monitored, wherein the set of monitored images to be processed includes multiple frames of monitored images to be processed;

[0009] Performing monitoring user identification processing on each frame of the monitoring image to be processed respectively, so as to determine each frame of the monitoring image to be processed having a monitoring user among the multiple frames of the monitoring image to be processed, and performing monitoring user interception processing on the determined monitoring image to be processed to form a corresponding monitoring image to be analyzed, and sorting the monitoring images to be analyzed according to the chronological relationship between the corresponding image acquisition times to form a corresponding monitoring image sequence to be analyzed;

[0010] Performing network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network;

[0011] The target behavior analysis network is used to perform behavior analysis on the monitored image sequence to be analyzed, and a behavior analysis result corresponding to the monitored area is output, wherein the behavior analysis result is used to reflect whether there is abnormal user behavior in the monitored area.

[0012] In some preferred embodiments, in the above-mentioned behavior analysis method based on security monitoring images, the step of performing network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network includes:

[0013] Extracting an initial monitoring image data cluster, each initial monitoring image data in the initial monitoring image data cluster does not have abnormal behavior identification data, and the abnormal behavior identification data is used to reflect whether there is abnormal user behavior;

[0014] Performing a masking process on at least one initial exemplary monitoring image in each of the initial monitoring image data in the initial monitoring image data cluster to mask it by using preset masking information to form a corresponding intermediate monitoring image data cluster, wherein each of the initial monitoring image data includes a plurality of frames of initial exemplary monitoring images arranged in sequence;

[0015] Performing a loading process on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized, and restoring a corresponding restored monitoring image data cluster, wherein the intermediate monitoring image data includes multiple frames of intermediate exemplary monitoring images arranged in sequence, and each restored monitoring image data in the restored monitoring image data cluster includes multiple frames of restored exemplary monitoring images arranged in sequence;

[0016] According to the initial monitoring image data cluster and the restored monitoring image data cluster, the behavior analysis network to be optimized is subjected to network optimization processing to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

[0017] In some preferred embodiments, in the above-mentioned behavior analysis method based on security monitoring images, the intermediate monitoring image data cluster includes a first number of intermediate monitoring image data;

[0018] The step of loading and processing each intermediate monitoring image data in the intermediate monitoring image data cluster to load it into the established behavior analysis network to be optimized, and restoring the corresponding restored monitoring image data cluster includes:

[0019] Performing loading processing on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized;

[0020] Using the behavior analysis network to be optimized, restore an exemplary restored monitoring image corresponding to each preset masking information in the a-th intermediate monitoring image data;

[0021] The preset masking information in the ath intermediate monitoring image data is updated to be a restored exemplary monitoring image corresponding to the preset masking information in the ath intermediate monitoring image data, so as to obtain the ath restored monitoring image data in the restored monitoring image data cluster.

[0022] In some preferred embodiments, in the above-mentioned behavior analysis method based on security monitoring images, the step of updating the preset masking information in the a-th intermediate monitoring image data to update it to a restored exemplary monitoring image corresponding to the preset masking information in the a-th intermediate monitoring image data, and obtaining the a-th restored monitoring image data in the restored monitoring image data cluster includes:

[0023] Determine a first restored exemplary monitoring image from a plurality of restored exemplary monitoring images corresponding to the preset masking information in the ath intermediate monitoring image data; and update the preset masking information row in the ath intermediate monitoring image data to update it to the first restored exemplary monitoring image, thereby obtaining the ath restored monitoring image data;

[0024] The step of determining a first restored exemplary monitoring image from a plurality of restored exemplary monitoring images corresponding to the preset masking information in the ath intermediate monitoring image data comprises:

[0025] An initial exemplary monitoring image corresponding to the preset masking information in the a-th intermediate monitoring image data is found, and among multiple restored exemplary monitoring images corresponding to the preset masking information in the a-th intermediate monitoring image data, the initial exemplary monitoring image is screened out; and, possibility assessment parameters of each restored exemplary monitoring image that has not been screened out and corresponds to the preset masking information in the a-th intermediate monitoring image data are determined; and, the restored exemplary monitoring image corresponding to the possibility assessment parameter having the maximum value is marked as the corresponding first restored exemplary monitoring image.

[0026] In some preferred embodiments, in the above-mentioned behavior analysis method based on security monitoring images, the step of performing network optimization processing on the behavior analysis network to be optimized based on the initial monitoring image data cluster and the restored monitoring image data cluster to form a target behavior analysis network corresponding to the behavior analysis network to be optimized includes:

[0027] Performing loading processing on each initial monitoring image data in the initial monitoring image data cluster and each restored monitoring image data in the restored monitoring image data cluster, so as to load them into the behavior analysis network to be optimized;

[0028] Using the behavior analysis network to be optimized, an initial user behavior representation vector corresponding to each of the initial monitoring image data is mined, and a restored user behavior representation vector corresponding to each of the restored monitoring image data is mined;

[0029] The behavior analysis network to be optimized is subjected to network optimization processing according to the initial user behavior characterization vector and the restored user behavior characterization vector to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

[0030] In some preferred embodiments, in the above-mentioned behavior analysis method based on security monitoring images, the initial monitoring image data cluster includes a first number of initial monitoring image data, the restored monitoring image data cluster includes a first number of restored monitoring image data, and the bth restored monitoring image data corresponds to the bth initial monitoring image data;

[0031] The step of performing network optimization processing on the behavior analysis network to be optimized based on the initial user behavior characterization vector and the restored user behavior characterization vector to form a target behavior analysis network corresponding to the behavior analysis network to be optimized includes:

[0032] Marking the b-th restored monitoring image data to mark it as the relevant monitoring image data corresponding to the b-th initial monitoring image data, and marking other restored monitoring image data other than the b-th restored monitoring image data to mark them as the non-relevant monitoring image data corresponding to the b-th initial monitoring image data;

[0033] Based on the initial user behavior representation vector corresponding to the bth initial monitoring image data, the restored user behavior representation vector corresponding to the related monitoring image data corresponding to the bth initial monitoring image data, and the restored user behavior representation vector corresponding to the non-related monitoring image data corresponding to the bth initial monitoring image data, the behavior analysis network to be optimized is subjected to network optimization processing to form a corresponding target behavior analysis network.

[0034] In some preferred embodiments, in the above-mentioned behavior analysis method based on security monitoring images, the step of using the target behavior analysis network to perform behavior analysis on the monitoring image sequence to be analyzed and outputting the behavior analysis result corresponding to the monitored area includes:

[0035] Using the behavior feature mining subnetwork included in the target behavior analysis network, the behavior feature mining process is performed on the monitoring image sequence to be analyzed, so as to output a user behavior representation vector corresponding to the monitoring image sequence to be analyzed, wherein the user behavior representation vector is used to reflect the image semantics of the monitoring images to be analyzed included in the monitoring image sequence to be analyzed in the behavior dimension;

[0036] The behavior analysis output subnetwork included in the target behavior analysis network is used to perform behavior analysis on the user behavior representation vector, and a behavior analysis result corresponding to the area to be monitored is output.

[0037] The embodiment of the present invention further provides a behavior analysis system based on security monitoring images, comprising:

[0038] A monitoring image acquisition module, used for acquiring images in a to-be-monitored area to form a to-be-processed monitoring image set corresponding to the to-be-monitored area, wherein the to-be-processed monitoring image set includes a plurality of frames of to-be-processed monitoring images;

[0039] A monitoring image screening module, used to perform monitoring user identification processing on each frame of the monitoring image to be processed, so as to determine each frame of the monitoring image to be processed having a monitoring user among the multiple frames of the monitoring image to be processed, and to perform monitoring user interception processing on the determined monitoring image to be processed, so as to form a corresponding monitoring image to be analyzed, and to sort the monitoring images to be analyzed according to the chronological relationship between the corresponding image acquisition times, so as to form a corresponding monitoring image sequence to be analyzed;

[0040] A network optimization processing module is used to perform network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network;

[0041] The abnormal behavior analysis module is used to use the target behavior analysis network to perform behavior analysis on the monitoring image sequence to be analyzed, and output the behavior analysis results corresponding to the monitored area, wherein the behavior analysis results are used to reflect whether there is abnormal user behavior in the monitored area.

[0042] In some preferred embodiments, in the above-mentioned behavior analysis system based on security monitoring images, the network optimization processing module is specifically used to:

[0043] Extracting an initial monitoring image data cluster, each initial monitoring image data in the initial monitoring image data cluster does not have abnormal behavior identification data, and the abnormal behavior identification data is used to reflect whether there is abnormal user behavior;

[0044] Performing a masking process on at least one initial exemplary monitoring image in each of the initial monitoring image data in the initial monitoring image data cluster to mask it by using preset masking information to form a corresponding intermediate monitoring image data cluster, wherein each of the initial monitoring image data includes a plurality of frames of initial exemplary monitoring images arranged in sequence;

[0045] Performing a loading process on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized, and restoring a corresponding restored monitoring image data cluster, wherein the intermediate monitoring image data includes multiple frames of intermediate exemplary monitoring images arranged in sequence, and each restored monitoring image data in the restored monitoring image data cluster includes multiple frames of restored exemplary monitoring images arranged in sequence;

[0046] According to the initial monitoring image data cluster and the restored monitoring image data cluster, the behavior analysis network to be optimized is subjected to network optimization processing to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

[0047] In some preferred embodiments, in the above-mentioned behavior analysis system based on security monitoring images, the behavior anomaly analysis module is specifically used to:

[0048] Using the behavior feature mining subnetwork included in the target behavior analysis network, the behavior feature mining process is performed on the monitoring image sequence to be analyzed, so as to output a user behavior representation vector corresponding to the monitoring image sequence to be analyzed, wherein the user behavior representation vector is used to reflect the image semantics of the monitoring images to be analyzed included in the monitoring image sequence to be analyzed in the behavior dimension;

[0049] The behavior analysis output subnetwork included in the target behavior analysis network is used to perform behavior analysis on the user behavior representation vector, and a behavior analysis result corresponding to the area to be monitored is output.

[0050] The embodiment of the present invention provides a behavior analysis method and system based on security monitoring images, which can first collect images in the monitored area to form a set of monitoring images to be processed; perform monitoring user identification processing on each frame of the monitoring image to be processed to determine each frame of the monitoring image to be processed with a monitoring user in multiple frames of monitoring images to be processed, and perform monitoring user interception processing on the determined monitoring images to be processed to form a monitoring image to be analyzed, and sort the monitoring images to be analyzed to form a monitoring image sequence to be analyzed; perform network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network; use the target behavior analysis network to perform behavior analysis on the monitoring image sequence to be analyzed, and output the behavior analysis results. Based on the above content, since the monitoring user identification and image interception processing will be performed on the monitoring image to be processed before the behavior analysis is performed, the monitoring image to be analyzed has a higher accuracy, that is, it avoids the interference of invalid information on the behavior analysis, thereby improving the reliability of the behavior analysis to a certain extent, thereby improving the problem of poor reliability of the behavior analysis in the prior art.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a structural block diagram of a behavior analysis platform based on security monitoring images provided in an embodiment of the present invention.

[0053] Figure 2 A flowchart of the steps of the behavior analysis method based on security monitoring images provided in an embodiment of the present invention.

[0054] Figure 3 A schematic diagram of various modules included in a behavior analysis system based on security monitoring images provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0056] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0057] like Figure 1 As shown, an embodiment of the present invention provides a behavior analysis platform based on security monitoring images. The behavior analysis platform based on security monitoring images may include a memory and a processor, and may also include other devices or components.

[0058] In detail, the memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, they can be electrically connected to each other through one or more communication buses or signal lines. The memory can store at least one software function module (computer program) that can exist in the form of software or firmware. The processor can be used to execute the executable computer program stored in the memory, thereby realizing the behavior analysis method based on security monitoring images provided in an embodiment of the present invention.

[0059] Optionally, in some embodiments, the memory may be, but is not limited to, random access memory (RAM), read only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc.

[0060] Optionally, in some embodiments, the processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0061] Optionally, in some implementations, the behavior analysis platform based on security monitoring images may be a server with data processing capabilities.

[0062] Combination Figure 2 The embodiment of the present invention further provides a behavior analysis method based on security monitoring images, which can be applied to the above-mentioned behavior analysis platform based on security monitoring images. The method steps defined in the process related to the behavior analysis method based on security monitoring images can be implemented by the behavior analysis platform based on security monitoring images.

[0063] The following will Figure 2 The specific process shown is explained in detail.

[0064] Step S110: collecting images in the area to be monitored to form a set of monitored images to be processed corresponding to the area to be monitored.

[0065] In an embodiment of the present invention, the behavior analysis platform based on security monitoring images can collect images in the monitored area to form a set of monitored images to be processed corresponding to the monitored area. The set of monitored images to be processed includes multiple frames of monitored images to be processed.

[0066] Step S120, performing monitoring user identification processing on each frame of the monitoring image to be processed respectively, so as to determine each frame of the monitoring image to be processed having a monitoring user among the multiple frames of the monitoring image to be processed, and performing monitoring user interception processing on the determined monitoring image to be processed to form a corresponding monitoring image to be analyzed, and sorting the monitoring images to be analyzed according to the chronological relationship between the corresponding image acquisition times to form a corresponding monitoring image sequence to be analyzed.

[0067] In an embodiment of the present invention, the behavior analysis platform based on security monitoring images can perform monitoring user identification processing on each frame of the monitoring image to be processed respectively, so as to determine each frame of the monitoring image to be processed with a monitoring user among the multiple frames of monitoring images to be processed, and perform monitoring user interception processing on the determined monitoring image to be processed to form a corresponding monitoring image to be analyzed (that is, part of the image that does not have the monitoring user can be screened out, such as the background image), and sort the monitoring images to be analyzed according to the chronological relationship between the corresponding image acquisition times to form a corresponding monitoring image sequence to be analyzed.

[0068] Step S130, performing network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network.

[0069] In the embodiment of the present invention, the behavior analysis platform based on security monitoring images can perform network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network.

[0070] Step S140, using the target behavior analysis network, performing behavior analysis on the monitoring image sequence to be analyzed, and outputting the behavior analysis result corresponding to the area to be monitored.

[0071] In an embodiment of the present invention, the behavior analysis platform based on security monitoring images can use the target behavior analysis network to perform behavior analysis on the monitoring image sequence to be analyzed, and output the behavior analysis results corresponding to the monitored area. The behavior analysis results are used to reflect whether there is abnormal user behavior in the monitored area.

[0072] Based on the above contents, such as step S110, step S120, step S130 and step S140, since the monitoring user identification and image capture processing will be performed on the monitoring image to be processed before the behavior analysis is performed, the formed monitoring image to be analyzed has higher accuracy, that is, it avoids the interference of invalid information on the behavior analysis, thereby improving the reliability of the behavior analysis to a certain extent, and thus improving the problem of poor reliability of the behavior analysis in the prior art.

[0073] Optionally, in some implementations, the step of performing network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network may further include the following contents:

[0074] Extracting an initial monitoring image data cluster, each initial monitoring image data in the initial monitoring image data cluster does not have abnormal behavior identification data, and the abnormal behavior identification data is used to reflect whether there is abnormal user behavior;

[0075] Performing a masking process on at least one initial exemplary monitoring image in each of the initial monitoring image data in the initial monitoring image data cluster, so as to mask it by using preset masking information, so as to form a corresponding intermediate monitoring image data cluster, wherein each of the initial monitoring image data includes a plurality of frames of initial exemplary monitoring images arranged in sequence (the specific content of the preset masking information is not limited, such as a frame of blank image, or a completely black image);

[0076] Performing a loading process on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized, and restoring a corresponding restored monitoring image data cluster, wherein the intermediate monitoring image data includes multiple frames of intermediate exemplary monitoring images arranged in sequence, and each restored monitoring image data in the restored monitoring image data cluster includes multiple frames of restored exemplary monitoring images arranged in sequence;

[0077] According to the initial monitoring image data cluster and the restored monitoring image data cluster, the behavior analysis network to be optimized is subjected to network optimization processing to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

[0078] Optionally, in some embodiments, the intermediate monitoring image data cluster includes a first number of intermediate monitoring image data. Based on this, the step of loading each intermediate monitoring image data in the intermediate monitoring image data cluster to load it into the established behavior analysis network to be optimized, and restoring the corresponding restored monitoring image data cluster may further include the following contents:

[0079] Performing loading processing on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized;

[0080] Using the behavior analysis network to be optimized, restore an exemplary restored monitoring image corresponding to each preset masking information in the a-th intermediate monitoring image data;

[0081] The preset masking information in the ath intermediate monitoring image data is updated to be a restored exemplary monitoring image corresponding to the preset masking information in the ath intermediate monitoring image data, so as to obtain the ath restored monitoring image data in the restored monitoring image data cluster.

[0082] Alternatively, in some embodiments, the step of updating the preset masking information in the ath intermediate monitoring image data to update the restored exemplary monitoring image corresponding to the preset masking information in the ath intermediate monitoring image data to obtain the ath restored monitoring image data in the restored monitoring image data cluster may further include the following contents:

[0083] Determining a first restored exemplary monitoring image from a plurality of restored exemplary monitoring images corresponding to preset masking information in the ath intermediate monitoring image data;

[0084] updating the preset mask information row in the ath intermediate monitoring image data to update it to the first restored exemplary monitoring image, thereby obtaining the ath restored monitoring image data;

[0085] Optionally, in some embodiments, the step of determining the first restored exemplary monitoring image from the multiple restored exemplary monitoring images corresponding to the preset masking information in the ath intermediate monitoring image data may further include the following contents:

[0086] Finding an initial exemplary monitoring image corresponding to the preset masking information in the a-th intermediate monitoring image data, and screening out the initial exemplary monitoring image among a plurality of restored exemplary monitoring images corresponding to the preset masking information in the a-th intermediate monitoring image data; and determining a possibility evaluation parameter (the possibility evaluation parameter is used to reflect the possibility of restoring the exemplary monitoring image, or the reliability) of each restored exemplary monitoring image that is not screened out and corresponds to the preset masking information in the a-th intermediate monitoring image data;

[0087] The restored exemplary monitoring image corresponding to the possibility assessment parameter with the maximum value is marked as the corresponding first restored exemplary monitoring image (that is, a first restored exemplary monitoring image that is different from the initial exemplary monitoring image is screened out from multiple restored exemplary monitoring images with preset masking information in the ath intermediate monitoring image data).

[0088] Optionally, in some embodiments, the step of performing network optimization processing on the behavior analysis network to be optimized based on the initial monitoring image data cluster and the restored monitoring image data cluster to form a target behavior analysis network corresponding to the behavior analysis network to be optimized may further include the following contents:

[0089] Performing loading processing on each initial monitoring image data in the initial monitoring image data cluster and each restored monitoring image data in the restored monitoring image data cluster, so as to load them into the behavior analysis network to be optimized;

[0090] Using the behavior analysis network to be optimized, an initial user behavior representation vector corresponding to each of the initial monitoring image data is mined, and a restored user behavior representation vector corresponding to each of the restored monitoring image data is mined;

[0091] The behavior analysis network to be optimized is subjected to network optimization processing according to the initial user behavior characterization vector and the restored user behavior characterization vector to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

[0092] Optionally, in some embodiments, the initial monitoring image data cluster may include a first number of initial monitoring image data, the restored monitoring image data cluster may include a first number of restored monitoring image data, the bth restored monitoring image data corresponds to the bth initial monitoring image data (i.e., the bth restored monitoring image data is formed based on the restoration processing of the intermediate monitoring image data corresponding to the bth initial monitoring image data), based on which, the step of performing network optimization processing on the behavior analysis network to be optimized based on the initial user behavior representation vector and the restored user behavior representation vector to form a target behavior analysis network corresponding to the behavior analysis network to be optimized may further include the following contents:

[0093] Marking the b-th restored monitoring image data to mark it as the relevant monitoring image data corresponding to the b-th initial monitoring image data, and marking other restored monitoring image data other than the b-th restored monitoring image data to mark them as the non-relevant monitoring image data corresponding to the b-th initial monitoring image data;

[0094] Based on the initial user behavior representation vector corresponding to the bth initial monitoring image data, the restored user behavior representation vector corresponding to the related monitoring image data corresponding to the bth initial monitoring image data, and the restored user behavior representation vector corresponding to the non-related monitoring image data corresponding to the bth initial monitoring image data, the behavior analysis network to be optimized is subjected to network optimization processing to form a corresponding target behavior analysis network.

[0095] Among them, it can be selected that, in some embodiments, the step of performing network optimization processing on the behavior analysis network to be optimized to form a corresponding target behavior analysis network based on the initial user behavior representation vector corresponding to the bth initial monitoring image data, the restored user behavior representation vector corresponding to the relevant monitoring image data corresponding to the bth initial monitoring image data, and the restored user behavior representation vector corresponding to the non-relevant monitoring image data corresponding to the bth initial monitoring image data, can further include the following contents:

[0096] Analyzing the relevant dimension matching coefficient between the initial user behavior representation vector corresponding to the b-th initial monitoring image data and the restored user behavior representation vector corresponding to the related monitoring image data corresponding to the b-th initial monitoring image data (i.e., calculating the vector similarity);

[0097] Analyzing the non-correlated dimension matching coefficient between the initial user behavior representation vector corresponding to the b-th initial monitoring image data and the restored user behavior representation vector corresponding to the non-correlated monitoring image data corresponding to the b-th initial monitoring image data (i.e., calculating the vector similarity);

[0098] According to the relevant dimension matching coefficient and the irrelevant dimension matching coefficient, the network learning cost value corresponding to the behavior analysis network to be optimized is analyzed (exemplarily, the relevant dimension matching coefficient and the network learning cost value may be negatively correlated, and the irrelevant dimension matching coefficient and the network learning cost value may be positively correlated);

[0099] According to the network learning cost value, the behavior analysis network to be optimized is subjected to network optimization processing to form a target behavior analysis network corresponding to the behavior analysis network to be optimized (the direction of the network optimization processing may be to increase the relevant dimension matching coefficient and to reduce the irrelevant dimension matching coefficient. Based on this, the behavior feature mining subnetwork included in the behavior analysis network to be optimized may be subjected to network optimization processing. In addition, the target behavior analysis network may further include a behavior analysis output subnetwork. The behavior analysis output subnetwork may include a softmax function for performing behavior analysis on a user behavior characterization vector. The behavior analysis output subnetwork may be optimized based on exemplary monitoring image data having behavior abnormality identification data).

[0100] Optionally, in some embodiments, the step of using the target behavior analysis network to perform behavior analysis on the monitored image sequence to be analyzed and outputting the behavior analysis result corresponding to the monitored area may further include the following contents:

[0101] Using the behavior feature mining subnetwork included in the target behavior analysis network, the behavior feature mining process is performed on the monitoring image sequence to be analyzed, so as to output a user behavior representation vector corresponding to the monitoring image sequence to be analyzed, wherein the user behavior representation vector is used to reflect the image semantics of the monitoring images to be analyzed included in the monitoring image sequence to be analyzed in the behavior dimension;

[0102] The behavior analysis output subnetwork included in the target behavior analysis network is used to perform behavior analysis on the user behavior representation vector, and a behavior analysis result corresponding to the area to be monitored is output.

[0103] Combination Figure 3 The embodiment of the present invention further provides a behavior analysis system based on security monitoring images, which can be applied to the above-mentioned behavior analysis platform based on security monitoring images. The behavior analysis system based on security monitoring images can include the following contents:

[0104] A monitoring image acquisition module, used for acquiring images in a to-be-monitored area to form a to-be-processed monitoring image set corresponding to the to-be-monitored area, wherein the to-be-processed monitoring image set includes a plurality of frames of to-be-processed monitoring images;

[0105] A monitoring image screening module, used to perform monitoring user identification processing on each frame of the monitoring image to be processed, so as to determine each frame of the monitoring image to be processed having a monitoring user among the multiple frames of the monitoring image to be processed, and to perform monitoring user interception processing on the determined monitoring image to be processed, so as to form a corresponding monitoring image to be analyzed, and to sort the monitoring images to be analyzed according to the chronological relationship between the corresponding image acquisition times, so as to form a corresponding monitoring image sequence to be analyzed;

[0106] A network optimization processing module is used to perform network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network;

[0107] The abnormal behavior analysis module is used to use the target behavior analysis network to perform behavior analysis on the monitoring image sequence to be analyzed, and output the behavior analysis results corresponding to the monitored area, wherein the behavior analysis results are used to reflect whether there is abnormal user behavior in the monitored area.

[0108] Optionally, in some implementations, the network optimization processing module is specifically used to:

[0109] Extracting an initial monitoring image data cluster, each initial monitoring image data in the initial monitoring image data cluster does not have abnormal behavior identification data, and the abnormal behavior identification data is used to reflect whether there is abnormal user behavior;

[0110] Performing a masking process on at least one initial exemplary monitoring image in each of the initial monitoring image data in the initial monitoring image data cluster to mask it by using preset masking information to form a corresponding intermediate monitoring image data cluster, wherein each of the initial monitoring image data includes a plurality of frames of initial exemplary monitoring images arranged in sequence;

[0111] Performing a loading process on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized, and restoring a corresponding restored monitoring image data cluster, wherein the intermediate monitoring image data includes multiple frames of intermediate exemplary monitoring images arranged in sequence, and each restored monitoring image data in the restored monitoring image data cluster includes multiple frames of restored exemplary monitoring images arranged in sequence;

[0112] According to the initial monitoring image data cluster and the restored monitoring image data cluster, the behavior analysis network to be optimized is subjected to network optimization processing to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

[0113] Optionally, in some embodiments, the abnormal behavior analysis module is specifically used to:

[0114] Using the behavior feature mining subnetwork included in the target behavior analysis network, the behavior feature mining process is performed on the monitoring image sequence to be analyzed, so as to output a user behavior representation vector corresponding to the monitoring image sequence to be analyzed, wherein the user behavior representation vector is used to reflect the image semantics of the monitoring images to be analyzed included in the monitoring image sequence to be analyzed in the behavior dimension;

[0115] The behavior analysis output subnetwork included in the target behavior analysis network is used to perform behavior analysis on the user behavior representation vector, and a behavior analysis result corresponding to the area to be monitored is output.

[0116] In summary, the behavior analysis method and system based on security monitoring images provided by the present invention can first collect images in the monitored area to form a set of monitoring images to be processed; perform monitoring user identification processing on each frame of monitoring images to be processed respectively to determine each frame of monitoring images to be processed with monitoring users in multiple frames of monitoring images to be processed, and perform monitoring user interception processing on the determined monitoring images to be processed to form monitoring images to be analyzed, and perform sorting processing on the monitoring images to be analyzed to form a monitoring image sequence to be analyzed; perform network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network; use the target behavior analysis network to perform behavior analysis on the monitoring image sequence to be analyzed, and output the behavior analysis results. Based on the above content, since the monitoring user identification and image interception processing will be performed on the monitoring images to be processed before the behavior analysis is performed, the formed monitoring images to be analyzed have higher accuracy, that is, avoiding the interference of invalid information on the behavior analysis, thereby improving the reliability of the behavior analysis to a certain extent, and thus improving the problem of poor reliability of the behavior analysis existing in the prior art.

[0117] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A behavior analysis method based on security monitoring images, characterized in that: include: Performing image acquisition in the area to be monitored to form a set of monitored images to be processed corresponding to the area to be monitored, wherein the set of monitored images to be processed includes multiple frames of monitored images to be processed; Performing monitoring user identification processing on each frame of the monitoring image to be processed respectively, so as to determine each frame of the monitoring image to be processed having a monitoring user among the multiple frames of the monitoring image to be processed, and performing monitoring user interception processing on the determined monitoring image to be processed to form a corresponding monitoring image to be analyzed, and sorting the monitoring images to be analyzed according to the chronological relationship between the corresponding image acquisition times to form a corresponding monitoring image sequence to be analyzed; Performing network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network; Using the target behavior analysis network, the behavior analysis is performed on the monitored image sequence to be analyzed, and the behavior analysis result corresponding to the monitored area is output, where the behavior analysis result is used to reflect whether there is abnormal user behavior in the monitored area; The step of performing network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network includes: Extracting an initial monitoring image data cluster, each initial monitoring image data in the initial monitoring image data cluster does not have abnormal behavior identification data, and the abnormal behavior identification data is used to reflect whether there is abnormal user behavior; Performing a masking process on at least one initial exemplary monitoring image in each of the initial monitoring image data in the initial monitoring image data cluster to mask it by using preset masking information to form a corresponding intermediate monitoring image data cluster, wherein each of the initial monitoring image data includes a plurality of frames of initial exemplary monitoring images arranged in sequence; Performing a loading process on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized, and restoring a corresponding restored monitoring image data cluster, wherein the intermediate monitoring image data includes multiple frames of intermediate exemplary monitoring images arranged in sequence, and each restored monitoring image data in the restored monitoring image data cluster includes multiple frames of restored exemplary monitoring images arranged in sequence; According to the initial monitoring image data cluster and the restored monitoring image data cluster, the behavior analysis network to be optimized is subjected to network optimization processing to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

2. The behavior analysis method based on security monitoring images according to claim 1, characterized in that: The intermediate monitoring image data cluster includes a first number of intermediate monitoring image data; The step of loading and processing each intermediate monitoring image data in the intermediate monitoring image data cluster to load it into the established behavior analysis network to be optimized, and restoring the corresponding restored monitoring image data cluster includes: Performing loading processing on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized; Using the behavior analysis network to be optimized, restore an exemplary restored monitoring image corresponding to each preset masking information in the a-th intermediate monitoring image data; The preset masking information in the ath intermediate monitoring image data is updated to be a restored exemplary monitoring image corresponding to the preset masking information in the ath intermediate monitoring image data, so as to obtain the ath restored monitoring image data in the restored monitoring image data cluster.

3. The behavior analysis method based on security monitoring images as claimed in claim 2, characterized in that: The step of updating the preset masking information in the ath intermediate monitoring image data to update it to the restored exemplary monitoring image corresponding to the preset masking information in the ath intermediate monitoring image data to obtain the ath restored monitoring image data in the restored monitoring image data cluster includes: Determine a first restored exemplary monitoring image from a plurality of restored exemplary monitoring images corresponding to the preset masking information in the ath intermediate monitoring image data; and update the preset masking information row in the ath intermediate monitoring image data to update it to the first restored exemplary monitoring image, thereby obtaining the ath restored monitoring image data; The step of determining a first restored exemplary monitoring image from a plurality of restored exemplary monitoring images corresponding to the preset masking information in the ath intermediate monitoring image data comprises: An initial exemplary monitoring image corresponding to the preset masking information in the a-th intermediate monitoring image data is found, and among multiple restored exemplary monitoring images corresponding to the preset masking information in the a-th intermediate monitoring image data, the initial exemplary monitoring image is screened out; and, possibility assessment parameters of each restored exemplary monitoring image that has not been screened out and corresponds to the preset masking information in the a-th intermediate monitoring image data are determined; and, the restored exemplary monitoring image corresponding to the possibility assessment parameter having the maximum value is marked as the corresponding first restored exemplary monitoring image.

4. The behavior analysis method based on security monitoring images according to claim 1, characterized in that: The step of performing network optimization processing on the behavior analysis network to be optimized based on the initial monitoring image data cluster and the restored monitoring image data cluster to form a target behavior analysis network corresponding to the behavior analysis network to be optimized includes: Performing loading processing on each initial monitoring image data in the initial monitoring image data cluster and each restored monitoring image data in the restored monitoring image data cluster, so as to load them into the behavior analysis network to be optimized; Using the behavior analysis network to be optimized, an initial user behavior representation vector corresponding to each of the initial monitoring image data is mined, and a restored user behavior representation vector corresponding to each of the restored monitoring image data is mined; The behavior analysis network to be optimized is subjected to network optimization processing according to the initial user behavior characterization vector and the restored user behavior characterization vector to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

5. The behavior analysis method based on security monitoring images as claimed in claim 4, characterized in that: The initial monitoring image data cluster includes a first number of initial monitoring image data, the restored monitoring image data cluster includes a first number of restored monitoring image data, and the bth restored monitoring image data corresponds to the bth initial monitoring image data; The step of performing network optimization processing on the behavior analysis network to be optimized based on the initial user behavior characterization vector and the restored user behavior characterization vector to form a target behavior analysis network corresponding to the behavior analysis network to be optimized includes: Marking the b-th restored monitoring image data to mark it as the relevant monitoring image data corresponding to the b-th initial monitoring image data, and marking other restored monitoring image data other than the b-th restored monitoring image data to mark them as the non-relevant monitoring image data corresponding to the b-th initial monitoring image data; Based on the initial user behavior representation vector corresponding to the bth initial monitoring image data, the restored user behavior representation vector corresponding to the related monitoring image data corresponding to the bth initial monitoring image data, and the restored user behavior representation vector corresponding to the non-related monitoring image data corresponding to the bth initial monitoring image data, the behavior analysis network to be optimized is subjected to network optimization processing to form a corresponding target behavior analysis network.

6. The behavior analysis method based on security monitoring images according to any one of claims 1 to 5, characterized in that: The step of using the target behavior analysis network to perform behavior analysis on the monitored image sequence to be analyzed and outputting the behavior analysis result corresponding to the monitored area includes: Using the behavior feature mining subnetwork included in the target behavior analysis network, the behavior feature mining process is performed on the monitoring image sequence to be analyzed, so as to output a user behavior representation vector corresponding to the monitoring image sequence to be analyzed, wherein the user behavior representation vector is used to reflect the image semantics of the monitoring images to be analyzed included in the monitoring image sequence to be analyzed in the behavior dimension; The behavior analysis output subnetwork included in the target behavior analysis network is used to perform behavior analysis on the user behavior representation vector, and a behavior analysis result corresponding to the area to be monitored is output.

7. A behavior analysis system based on security monitoring images, characterized in that: include: A monitoring image acquisition module, used for acquiring images in a to-be-monitored area to form a to-be-processed monitoring image set corresponding to the to-be-monitored area, wherein the to-be-processed monitoring image set includes a plurality of frames of to-be-processed monitoring images; A monitoring image screening module, used to perform monitoring user identification processing on each frame of the monitoring image to be processed, so as to determine each frame of the monitoring image to be processed having a monitoring user among the multiple frames of the monitoring image to be processed, and to perform monitoring user interception processing on the determined monitoring image to be processed, so as to form a corresponding monitoring image to be analyzed, and to sort the monitoring images to be analyzed according to the chronological relationship between the corresponding image acquisition times, so as to form a corresponding monitoring image sequence to be analyzed; A network optimization processing module is used to perform network optimization processing on the behavior analysis network to be optimized to form a target behavior analysis network; A behavior abnormality analysis module, used to use the target behavior analysis network to perform behavior analysis on the monitoring image sequence to be analyzed, and output a behavior analysis result corresponding to the monitored area, wherein the behavior analysis result is used to reflect whether there is abnormal user behavior in the monitored area; The network optimization processing module is specifically used for: Extracting an initial monitoring image data cluster, each initial monitoring image data in the initial monitoring image data cluster does not have abnormal behavior identification data, and the abnormal behavior identification data is used to reflect whether there is abnormal user behavior; Performing a masking process on at least one initial exemplary monitoring image in each of the initial monitoring image data in the initial monitoring image data cluster to mask it by using preset masking information to form a corresponding intermediate monitoring image data cluster, wherein each of the initial monitoring image data includes a plurality of frames of initial exemplary monitoring images arranged in sequence; Performing a loading process on each intermediate monitoring image data in the intermediate monitoring image data cluster so as to load it into the established behavior analysis network to be optimized, and restoring a corresponding restored monitoring image data cluster, wherein the intermediate monitoring image data includes multiple frames of intermediate exemplary monitoring images arranged in sequence, and each restored monitoring image data in the restored monitoring image data cluster includes multiple frames of restored exemplary monitoring images arranged in sequence; According to the initial monitoring image data cluster and the restored monitoring image data cluster, the behavior analysis network to be optimized is subjected to network optimization processing to form a target behavior analysis network corresponding to the behavior analysis network to be optimized.

8. The behavior analysis system based on security monitoring images as claimed in claim 7, characterized in that: The behavior anomaly analysis module is specifically used for: Using the behavior feature mining subnetwork included in the target behavior analysis network, the behavior feature mining process is performed on the monitoring image sequence to be analyzed, so as to output a user behavior representation vector corresponding to the monitoring image sequence to be analyzed, wherein the user behavior representation vector is used to reflect the image semantics of the monitoring images to be analyzed included in the monitoring image sequence to be analyzed in the behavior dimension; The behavior analysis output subnetwork included in the target behavior analysis network is used to perform behavior analysis on the user behavior representation vector, and a behavior analysis result corresponding to the area to be monitored is output.