Security monitoring method and system based on image recognition

Abstract

The application provides a kind of security monitoring method and system based on image recognition, according to the regional configuration information of supervision end, determine the multiple monitoring sub-regions corresponding to monitoring data, risk source identification is carried out to each monitoring sub-region, determine the risk object in monitoring sub-region;The object attribute of risk object is obtained, and the safety layer corresponding to the risk object is generated according to the object attribute, and the display data is obtained by hiding the risk of the monitoring data based on the safety layer;According to the calibration adjustment strategy corresponding to the display data, the display data is analyzed according to the security calibration model, and the layer attribute parameter corresponding to the display data is adjusted based on the calibration adjustment strategy, and the safety adjustment data of display data is obtained.

Description

Technical Field

[0001] This invention relates to data processing technology, and more particularly to a security monitoring method and system based on image recognition. Background Technology

[0002] With the rapid development of technology, information may be extracted through various channels, such as by viewing surveillance videos and analyzing the video content, which could lead to the leakage of information technology. Therefore, ensuring that confidential information is not leaked and guaranteeing information security has received increasing attention.

[0003] In existing technologies, viewing permissions are typically set for surveillance videos. For example, personnel with high permissions can view the surveillance videos, while those without permissions cannot view them, thus ensuring information security. However, once the permission information is decrypted, the information in the corresponding surveillance videos still poses a risk of leakage because there is no information masking.

[0004] Therefore, how to hide information in surveillance video images, reduce the risk of information leakage, and improve the security of surveillance video information has become an urgent problem to be solved. Summary of the Invention

[0005] This invention provides a security monitoring method and system based on image recognition, which can hide information in images in surveillance videos, reduce the risk of information leakage, and improve the security of surveillance video information.

[0006] A first aspect of the present invention provides a security monitoring method based on image recognition, comprising:

[0007] Based on the regional configuration information of the regulatory end, multiple monitoring sub-regions corresponding to the monitoring data are determined, and risk sources are identified in each monitoring sub-region to determine the risk objects in the monitoring sub-region;

[0008] Obtain the object attributes of the risk object, generate a security layer corresponding to the risk object based on the object attributes, and hide the risk in the monitoring data based on the security layer to obtain the display data;

[0009] Based on the safety calibration model, a missing data analysis is performed on the displayed data to determine the corresponding calibration adjustment strategy.

[0010] Based on the calibration adjustment strategy, the layer attribute parameters corresponding to the displayed data are adjusted to obtain the safety adjustment data of the displayed data.

[0011] Optionally, in one possible implementation of the first aspect, the process includes obtaining the object attributes of the risk object, generating a security layer corresponding to the risk object based on the object attributes, and performing risk hiding on the monitoring data based on the security layer to obtain display data, including:

[0012] When the object attribute of the risk object is a stable attribute, the operation status of the risk object is identified, and the sub-location of the region corresponding to the risk object whose operation status is on is obtained.

[0013] Based on the sub-location of the area, a type of security layer corresponding to the risk object is retrieved, and the monitoring data is overlaid with the type of security layer to obtain the displayed data; or...

[0014] When the object attribute of the risk object is a variable attribute, obtain the displacement information corresponding to the risk object, and generate a second-class safety layer corresponding to the risk object based on the displacement information.

[0015] The monitoring data is overlaid with the second type of security layer to obtain the displayed data. The security layer includes a first type of security layer and a second type of security layer.

[0016] Optionally, in one possible implementation of the first aspect, when the object property of the risk object is a stable property, the operational state of the risk object is identified, and the sub-location of the region corresponding to the risk object whose operational state is "on" is obtained, including:

[0017] Extract the first contour of the risk object, and determine the first pixel point whose pixel value is located within the black screen pixel range in the first contour;

[0018] The black screen ratio corresponding to the risk object is determined based on the ratio of the number of the first pixel to the total number of pixels in the first contour.

[0019] The operating state corresponding to the risk object whose black screen ratio is less than the black screen threshold is determined to be the on state, and the sub-location of the risk object in the monitoring sub-region is obtained.

[0020] Optionally, in one possible implementation of the first aspect, when the object attribute of the risk object is a variable attribute, the displacement information corresponding to the risk object is obtained, and a second-class safety layer corresponding to the risk object is generated based on the displacement information, including:

[0021] Multiple image frames corresponding to the monitoring data are arranged in chronological order to obtain a monitoring sequence, and the image frames in the monitoring sequence are selected sequentially.

[0022] Obtain the second contour of the risk object in the current selected frame, determine the selected area in the next selected frame based on the positioning point corresponding to the second contour, and determine the displacement distance of the risk object in the adjacent selected frame based on the selected area;

[0023] The movement characteristics of the risk object are determined based on the displacement distance, and two types of safety layers are generated based on the movement characteristics and the second contour of the risk object. The movement characteristics include dynamic features and static features.

[0024] Optionally, in one possible implementation of the first aspect, obtaining the second contour of the risk object in the current selected frame, determining a selected area in the next selected frame based on the positioning point corresponding to the second contour, and determining the displacement distance of the risk object in adjacent selected frames based on the selected area, includes:

[0025] Extract the second contour corresponding to the risk object in the current selected frame, and determine the center point of the second contour as the positioning point;

[0026] Obtain the shortest distance between the positioning point and the boundary of the selected area. When the shortest distance is greater than the critical distance threshold, obtain the regional position of the selected area in the current selected frame, and determine the monitoring sub-region corresponding to the regional position in the next selected frame as the selected area.

[0027] When the shortest distance is less than or equal to the critical distance threshold, the movement trajectory of the risk object is determined based on the location points corresponding to the risk object in multiple selected frames corresponding to the historical time period.

[0028] Obtain the location of the risk object corresponding to the unchanged movement trajectory, and determine the monitoring sub-region corresponding to the location in the next selected frame as the selected area;

[0029] The tangential direction corresponding to the endpoint of the changed action trajectory is determined as the prediction direction. The monitoring sub-regions adjacent to the selected area and located in the prediction direction, as well as the area position of the selected area, are obtained. The monitoring sub-region corresponding to the area position in the next selected frame is determined as the selected area.

[0030] Risk sources are identified in the selected area, and the second contour corresponding to the risk object in the selected area is obtained. The displacement distance is obtained based on the point distance between the positioning points corresponding to the same risk object in adjacent selected frames.

[0031] Optionally, in one possible implementation of the first aspect, the directional characteristics of the hazardous object are determined based on the displacement distance, and a second type of safety layer is generated based on the directional characteristics and the second contour of the hazardous object. The directional characteristics include dynamic features and static features, including:

[0032] The initial layer corresponding to the monitoring data is retrieved, and an occlusion contour corresponding to the second contour is generated in the initial layer. The occlusion pixel value is retrieved to update the occlusion contour to obtain the second type of security layer.

[0033] The movement characteristics of a risk object whose displacement distance is less than the movement threshold are determined to be static characteristics, and the movement characteristics of a risk object whose displacement distance is greater than or equal to the movement threshold are determined to be dynamic characteristics.

[0034] When the risk object is a static feature, the two types of security layers corresponding to the risk object in adjacent selected frames are the same;

[0035] When the risk object is a dynamic feature, two types of security layers are generated based on the second contour corresponding to each risk object in the adjacent selected frames.

[0036] Optionally, in one possible implementation of the first aspect, performing a missing data analysis on the displayed data based on a security calibration model to determine the calibration adjustment strategy corresponding to the displayed data includes:

[0037] Based on the security calibration model, a preset number of image frames are randomly selected from the displayed data as calibration frames. The abnormal contours selected by the monitoring terminal in the calibration frames are obtained, and the image frames in the monitoring data corresponding to the calibration frames are determined as comparison frames.

[0038] Determine a reference contour in the comparison frame that has the same location point as the abnormal contour, and compare the similarity between the reference contour and the abnormal contour.

[0039] Abnormal contours with a similarity greater than or equal to the similarity threshold are identified as unoccluded contours, while abnormal contours with a similarity less than the similarity threshold are identified as occluded or incomplete contours.

[0040] The display data that only contains the unobscured outline corresponds to the completion adjustment strategy; the display data that only contains the obscured and incomplete outline corresponds to the coverage adjustment strategy; and the display data that contains both the unobscured outline and the obscured and incomplete outline corresponds to the comprehensive adjustment strategy.

[0041] The calibration adjustment strategies include a completion adjustment strategy, a coverage adjustment strategy, and a comprehensive adjustment strategy.

[0042] Optionally, in one possible implementation of the first aspect, adjusting the layer attribute parameters corresponding to the displayed data based on the calibration adjustment strategy to obtain safe adjustment data for the displayed data includes:

[0043] The first number of unoccluded contours and the second number of occluded or incomplete contours in each monitoring sub-region of the displayed data are counted.

[0044] The sensitivity adjustment coefficient is obtained based on the ratio of the first quantity to the baseline sensitivity value, and the augmentation adjustment coefficient is obtained based on the ratio of the second quantity to the baseline augmentation value.

[0045] Based on the sensitivity adjustment coefficient, the sensitivity adjustment table is traversed to determine the preset sensitivity parameter corresponding to the preset coefficient range where the sensitivity adjustment coefficient is located as the first adjustment parameter;

[0046] The expansion correspondence table is traversed according to the expansion adjustment coefficient to determine the preset expansion parameter corresponding to the preset coefficient interval where the expansion adjustment coefficient is located as the second adjustment parameter.

[0047] Layer adjustment parameters are obtained based on the first adjustment parameter and / or the second adjustment parameter, and the layer adjustment parameters and the corresponding monitoring sub-regions are bound to obtain safety adjustment data.

[0048] Optionally, in one possible implementation of the first aspect, the sensitivity adjustment table and the expanded correspondence table are obtained through the following steps:

[0049] The system acquires multiple identification levels configured by the monitoring terminal. Each identification level is configured with a corresponding preset sensitivity parameter. The higher the identification level, the more preset sensitivity parameters are corresponding to that identification level.

[0050] Arrange the preset sensitivity parameters from largest to smallest according to the identification level to obtain a parameter sequence, and obtain the preset coefficient range configured by the monitoring terminal for each preset sensitivity parameter to obtain a sensitivity adjustment table. The higher the identification level, the larger the preset coefficient range corresponding to the preset sensitivity parameter.

[0051] The monitoring terminal is configured with multiple augmentation levels, each of which has a corresponding augmentation ratio. The higher the augmentation level, the greater the augmentation ratio.

[0052] The expansion sequence is obtained by arranging the expansion ratios from largest to smallest according to the expansion level, and the expansion correspondence table is obtained by obtaining the preset coefficient range configured by the regulatory end for each expansion ratio.

[0053] The preset expansion parameters include the expansion ratio. The larger the expansion level, the larger the preset coefficient range corresponding to the expansion ratio.

[0054] A second aspect of the present invention provides a security monitoring system based on image recognition, comprising:

[0055] The identification module is used to determine multiple monitoring sub-regions corresponding to the monitoring data based on the regional configuration information of the regulatory end, identify risk sources in each monitoring sub-region, and determine the risk objects in the monitoring sub-region.

[0056] The hiding module is used to obtain the object attributes of the risk object, generate a security layer corresponding to the risk object based on the object attributes, and hide the monitoring data based on the security layer to obtain the display data.

[0057] The analysis module is used to perform missing data analysis on the displayed data based on the security calibration model, and to determine the calibration adjustment strategy corresponding to the displayed data.

[0058] The adjustment module is used to adjust the layer attribute parameters corresponding to the displayed data based on the calibration adjustment strategy to obtain the safety adjustment data of the displayed data.

[0059] The adjustment module is used to adjust the layer attribute parameters corresponding to the displayed data based on the calibration adjustment strategy to obtain the safety adjustment data of the displayed data.

[0060] The beneficial effects of this invention are as follows:

[0061] 1. This invention can accurately hide information with security risks in surveillance videos, reducing the risk of information leakage and improving the security of confidential information. First, this invention can identify potential sources of information leakage in surveillance videos, allowing for the subsequent masking and hiding of the monitored data before display. This ensures that potentially dangerous information is not leaked without affecting the viewing of subsequent monitoring data, thus improving information security. Specifically, the complete area displayed in the video frame of the surveillance data can be divided into multiple monitoring sub-regions based on region configuration information. Since different monitoring sub-regions correspond to different levels of information leakage risk, the adjustment parameters for different monitoring sub-regions may also differ. Dividing into multiple monitoring sub-regions improves the accuracy of parameter adjustment, thereby increasing identification efficiency and the accuracy of information hiding, thus accelerating the data hiding process. Second, this invention can obtain the object attributes of the risky object and generate a corresponding security layer based on these attributes to mask and hide the risky object. Finally, this invention can adjust the layer attribute parameters corresponding to the displayed data according to a calibration adjustment strategy to improve the accuracy of information masking and reduce the risk of information leakage.

[0062] 2. This invention can generate security layers according to different object attributes corresponding to risk objects, thereby hiding monitoring data, reducing the risk of information leakage, and improving information security. Specifically, when the object attribute of a risk object is stable, the corresponding sub-position of the risk object is determined based on its on / off state, and a corresponding security layer is generated based on this sub-position to obscure and hide the risk object, reducing the risk of information leakage. When the object attribute of a risk object is dynamic, it indicates that the corresponding risk object may move. Therefore, this invention can obtain the displacement information of the corresponding risk object and generate a corresponding second-type security layer to improve the accuracy of information hiding and reduce the risk of information leakage.

[0063] 3. This invention can adjust the layer attribute parameters corresponding to the displayed data according to the calibration and adjustment strategy to obtain security adjustment data, so as to adjust the security layer and thus cover up abnormal outlines in the corresponding displayed data, thereby improving the accuracy of information covering and reducing the risk of information leakage. First, based on the abnormal contours under different conditions, corresponding calibration and adjustment strategies are determined to facilitate targeted adjustments later, making the adjusted parameters more accurate. This improves the accuracy of information occlusion and reduces the risk of information leakage. Specifically, the determined sensitivity adjustment coefficients are iterated through a sensitivity adjustment table to select a preset coefficient range. The preset sensitivity parameter corresponding to this range is used as the first adjustment parameter. This allows for subsequent adjustments to the sensitivity parameters of the corresponding monitoring sub-regions, resulting in more accurate identification of risk objects and preventing omissions. This improves the accuracy of information occlusion and reduces the risk of information leakage. Furthermore, the determined augmentation adjustment coefficients are iterated through an augmentation correspondence table to select a preset coefficient range. The preset augmentation parameter corresponding to this range is used as the second adjustment parameter. This allows for subsequent adjustments to the augmentation parameters of the corresponding monitoring sub-regions, ensuring that the occlusion contour completely covers the risk object, thereby improving the integrity of information occlusion and reducing the risk of information leakage. Attached Figure Description

[0064] Figure 1 A flowchart of a security monitoring method based on image recognition provided by the present invention;

[0065] Figure 2 This is a schematic diagram illustrating the display of data provided by the present invention;

[0066] Figure 3 A schematic diagram of determining a selected area provided by the present invention;

[0067] Figure 4This is a schematic diagram of the structure of a security monitoring system based on image recognition provided by the present invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0070] See Figure 1 This is a flowchart of a security monitoring method based on image recognition provided in an embodiment of the present invention. The security monitoring method based on image recognition includes steps S1 to S4, as detailed below:

[0071] S1. Based on the regional configuration information of the regulatory end, determine multiple monitoring sub-regions corresponding to the monitoring data, identify risk sources in each monitoring sub-region, and determine the risk objects in the monitoring sub-regions.

[0072] It should be noted that surveillance equipment in office areas often captures documents placed on desks or information displayed on computer screens that may pose security risks. Criminals could potentially steal this vulnerable information through the surveillance data, increasing the risk of information leakage. Therefore, to prevent data leakage and ensure information security, it is possible to identify potential sources of information leakage in the surveillance video. This allows for the subsequent masking and concealment of the monitored data. When the data is displayed after concealment, the risk of information leakage can be reduced without affecting the viewing of subsequent surveillance data, thereby improving information security.

[0073] Understandably, to quickly identify risky objects and their attributes, and thus implement appropriate data hiding methods to improve accuracy and efficiency, monitoring data can be divided into multiple monitoring sub-areas. For example, when the work area is an office, it can be divided into office areas, corridor areas, and green areas. Since different monitoring sub-areas correspond to different levels of information leakage risk—for instance, green areas typically do not contain document information—targeted risk identification can be performed based on the actual conditions of different areas during subsequent risky object tracking and identification. This reduces the need for image recognition in those areas, thereby reducing data processing volume. Furthermore, the adjustment parameters for different monitoring sub-areas may differ. Dividing the data into multiple monitoring sub-areas can improve the accuracy of parameter adjustment, thereby increasing identification efficiency and the accuracy of information hiding, thus accelerating the data hiding rate. These monitoring sub-areas can be numbered, such as office area number A, corridor area number B, and green area number C, to facilitate subsequent identification of risky object locations and information hiding.

[0074] Among them, the monitoring terminal is the information terminal of the personnel who monitor the monitoring data; the area configuration information is the configuration information for dividing the monitoring data into areas; the monitoring data is the video data for real-time monitoring of the work area; the monitoring sub-area is the sub-area after dividing the area in the monitoring data, such as office area, corridor area, green area, etc.; the risk source is the item that may cause information leakage risk, such as paper documents, computers, etc.; the risk object is the item that has information leakage risk, such as paper documents with text information, and computer screens that are displayed when the computer is turned on.

[0075] It's easy to understand that the type of risky object can be determined using OpenCV image processing technology. For example, images in video frames can be compared with existing computer equipment or paper documents to determine if the risky object is a computer or paper document.

[0076] S2, obtain the object attributes of the risk object, generate a security layer corresponding to the risk object based on the object attributes, and perform risk hiding on the monitoring data based on the security layer to obtain the display data.

[0077] Understandably, different risk objects have different properties. For example, computers are usually stationary and have high stability, while paper documents may shift in position in monitoring data as people handle them, making them less stable. Therefore, for risk objects with high stability—meaning their position won't change over a long period—a corresponding security layer can be pre-generated, and the same layer can be used for subsequent masking and hiding without continuously generating new layers to overlay and hide video frames based on position changes. However, for risk objects with low stability, their positions may differ across video frames. To hide these risk objects in real-time, new layers need to be generated for overlay and hiding. Therefore, to completely hide potentially dangerous information, improve hiding efficiency, and reduce the risk of information leakage, it's necessary to obtain the object's properties and generate corresponding security layers based on these properties for masking and hiding the risk object.

[0078] Among them, object attributes are the attributes of risk objects corresponding to different degrees of stability, including stability attributes and change attributes; security layer is a layer for hiding information about risk objects; and display data is video data that can be displayed and viewed after information is hidden.

[0079] Based on the above embodiments, the specific implementation of step S2 (obtaining the object attributes of the risk object, generating a security layer corresponding to the risk object based on the object attributes, and performing risk hiding on the monitoring data based on the security layer to obtain display data) can be as follows:

[0080] S21, when the object attribute of the risk object is a stable attribute, the operation status of the risk object is identified, and the sub-location of the region corresponding to the risk object whose operation status is open is obtained.

[0081] Understandably, when the risk object is a computer, monitoring devices can only capture the corresponding screen image when the computer is powered on and running, thus posing a risk of information leakage. Therefore, in order to accurately conceal the risk object, the corresponding sub-locations that need to be hidden can be determined based on the operating status of the risk object, so as to subsequently determine the corresponding security layer and reduce the risk of information leakage.

[0082] Among them, the stability attribute is the object attribute that the location of the risk object will not change in the long term, the operating status is the operating status of the risk object with the stability attribute, such as the on state or the off state. The on state means that the screen of the risk object is on, which poses a risk of information leakage. The area sub-location is the location of the risk object in the monitoring sub-area, which is pre-configured.

[0083] It is worth mentioning that the specifications and dimensions of the security layer correspond to the dimensions of the video frames in the monitoring data. Furthermore, there may be multiple risk objects in the same monitoring sub-area. For example, there may be multiple computers in the same office area, and different computers have corresponding location numbers, such as A1, A2, A3, etc. Each risk object has a corresponding security layer. That is, in the layer with the same specifications, the location area corresponding to the risk object can be filled with black, and the size of the black-filled area is the same as the size of the risk object image in the corresponding video frame. This is so that when the corresponding security layer is retrieved later, the corresponding risk object can be hidden or covered, reducing the risk of information leakage.

[0084] It is easy to understand that within the same monitoring sub-area, there may be multiple risk objects with stable properties. Different risk objects have corresponding locations. For example, computer 1 corresponds to location A1, computer 2 corresponds to location A2, computer 3 corresponds to location A3, and so on. Therefore, the location of the risk object that needs to be hidden can be determined by the sub-location of the area, thereby determining the pre-generated security layer so as to hide the information, reduce the risk of information leakage with security risks, and improve information security.

[0085] Based on the above embodiments, the specific implementation of step S21 (when the object attribute of the risk object is a stable attribute, the operational status of the risk object is identified, and the sub-location of the region corresponding to the risk object whose operational status is open is obtained) can be as follows:

[0086] S211, extract the first contour of the risk object, and determine the pixel point whose pixel value is located in the black screen pixel interval in the first contour as the first pixel point.

[0087] Understandably, since the pixel values ​​displayed by a risky object are inconsistent in different operating states—for example, the computer screen is black when it is off, but can be other colors when it is on—it is necessary to determine the operating state of the risky object so as to make an accurate judgment based on the pixel values ​​of the corresponding pixels and generate the corresponding safety layer.

[0088] Wherein, the first contour is the image contour of the risk object, the black screen pixel range is the range of pixel values ​​corresponding to the black screen, and the first pixel point is the pixel point of the risk object corresponding to the black screen, that is, the pixel point of the risk object corresponding to the first contour whose pixel value is located within the black screen pixel range.

[0089] S212, determine the black screen ratio corresponding to the risk object based on the ratio of the number of the first pixel to the total number of pixels in the first contour.

[0090] It is understandable that, since there may be some non-black markings in the area corresponding to the computer outline, such as the white logo information on the computer, when the computer is off, there may be some pixels whose pixel values ​​are not in the black screen pixel range. When the computer is on, the number of pixels whose pixel values ​​are not in the black screen pixel range will be much greater than when the computer is off. Therefore, the accurate operating status can be determined based on the black screen ratio, thereby generating the corresponding security layer.

[0091] Among them, the black screen ratio is the proportion of black pixels corresponding to the risky object, that is, the ratio of the number of first pixels to the total number of pixels in the first outline.

[0092] S213, determine that the operating state of the risk object whose black screen ratio is less than the black screen threshold is the on state, and obtain the regional sub-location of the risk object in the monitoring sub-region.

[0093] Understandably, the black screen threshold is the threshold of black screen pixels in the black screen state of the risk object, such as 95%. When the black screen ratio is greater than the black screen threshold, it means that the current corresponding screen is in a black screen off state. When the black screen ratio is less than the black screen threshold, it means that the pixels in the current screen that are not in the black screen pixel range have other pixels besides the relevant logo information, so it can be determined that the running state is on.

[0094] It is easy to understand that when the running status is on, it means that the current risk object has the risk of information leakage. Thus, the coordinates of the risk object in the corresponding video can be obtained. When the corresponding coordinates correspond to the coordinates of the preset area sub-position, the area sub-position of the risk object in the monitoring sub-area can be determined, so as to determine the pre-generated security layer based on the corresponding area sub-position.

[0095] S22, retrieve a type of safety layer corresponding to the risk object based on the sub-location of the area, and overlay the type of safety layer on the monitoring data to obtain the display data.

[0096] It is understandable that when the corresponding sub-location of the area is obtained, a corresponding type of security layer can be superimposed on the monitoring data in order to hide or obscure objects that pose a risk of information leakage, thereby obtaining video frame data after information obscuring, reducing the risk of information leakage and improving information security. Among them, the type of security layer is a security layer with stable attributes corresponding to fixed risk objects.

[0097] For example, such as Figure 2As shown, this is a schematic diagram of data display provided by the present invention. When it is determined that the corresponding computer 1 is in an on state, a security layer corresponding to computer 1 is retrieved and superimposed on the image frame corresponding to the monitoring data, so that the computer screen in the on state is blocked, reducing the risk of information leakage.

[0098] S23, when the object attribute of the risk object is a variable attribute, obtain the displacement information corresponding to the risk object, and generate a second-class safety layer corresponding to the risk object based on the displacement information.

[0099] It is understandable that when the object attribute of a risky object is a dynamic attribute, it means that the corresponding risky object may move. In order to hide the information of the corresponding risky object and prevent information leakage during the movement of the risky object, it is necessary to obtain the displacement information of the corresponding risky object and generate the corresponding second-class security layer in order to improve the accuracy of information hiding and reduce the risk of information leakage.

[0100] Among them, the variable attribute is the object attribute that allows the risk object to change position, the displacement information is the information about the risk object's positional movement, and the second-class safety layer is the safety layer corresponding to the variable attribute of the risk object.

[0101] Based on the above embodiments, the specific implementation of step S23 (when the object attribute of the risk object is a variable attribute, obtain the displacement information corresponding to the risk object, and generate the second-class safety layer corresponding to the risk object based on the displacement information) can be as follows:

[0102] S231, the multiple image frames corresponding to the monitoring data are arranged in chronological order to obtain a monitoring sequence, and the image frames in the monitoring sequence are selected in turn.

[0103] It should be noted that risk objects with variable attributes may move in location within a certain period of time or remain stationary. When the location remains stationary, the security layer corresponding to the initial location can be overlaid on the remaining video frames without needing to re-process the data to generate a security layer based on the video frames, thus reducing the amount of data processing. When the location moves, the security layers corresponding to different video frames will change, thereby ensuring that the displayed data does not result in information leakage. Therefore, it is necessary to determine the displacement information of the corresponding risk object based on the video frames in order to obtain the corresponding second-class security layer for information hiding and to reduce the risk of information leakage.

[0104] The monitoring data is dynamically displayed by displaying multiple video frames based on time. This allows multiple image frames in the corresponding monitoring data to be arranged in chronological order to obtain a monitoring sequence. The selected frames are then identified based on the monitoring sequence to determine the displacement information, thereby generating the corresponding second-class security layer.

[0105] It is easy to understand that an image frame is the corresponding video frame in the monitoring data, the monitoring sequence is the image frame sequence after arranging the video frames in the monitoring data according to the time order, and the selected frame is the selected image frame, which is subsequently used to determine the displacement information of the risk object.

[0106] S232, obtain the second contour of the risk object in the current selected frame, determine the selected area in the next selected frame based on the positioning point corresponding to the second contour, and determine the displacement distance of the risk object in the adjacent selected frame based on the selected area.

[0107] Understandably, the second contour is the contour of a risk object with variable attributes, the positioning point is the location point that determines the displacement information of the risk object, the selected area is the area where the risk object may be located, and the displacement distance is the distance between the positioning points of the same risk object.

[0108] For example, when the risk object is a paper document and it is moved by personnel, the corresponding positioning point in the currently selected frame is in office area A. In the next moment, the personnel may move to area B, and the corresponding paper document may move. Therefore, the position of the positioning point will change in different image frames. Since the size and coordinate origin of each image frame are the same, the displacement distance of the risk object can be obtained based on the positioning point coordinates in adjacent image frames, so as to judge the movement of the risk object and generate the corresponding Class II safety layer.

[0109] It's easy to understand that since the movement of personnel is not very large, the movement span shown between adjacent frames is also not very large. For example, when the monitored area includes areas A, B, and C, if a person is initially in area A and not within the boundary area, that person will not appear in area B or C in the next moment. If a person is initially in the boundary area of ​​area A, that person may move to area B in the next moment. Consequently, the location of the corresponding non-paper document's positioning point may also change. Therefore, the location of the risk object in the next image frame can be predicted based on the positioning point location, thereby determining the selected area. Then, the risk object can be identified only in the selected area, and the corresponding positioning point location can be determined, which can reduce the amount of data processing and improve the efficiency of information hiding.

[0110] Based on the above embodiments, step S232 (obtaining the second contour of the risk object in the current selected frame, determining the selected area in the next selected frame based on the positioning point corresponding to the second contour, and determining the displacement distance of the risk object in the adjacent selected frames based on the selected area) can be implemented as follows:

[0111] S2321, Extract the second contour corresponding to the risk object in the current selected frame, and determine the center point of the second contour as the positioning point.

[0112] It is understandable that the second contour corresponding to the risk object can be extracted using existing OpenCV image processing technology, and the center point of the second contour can be determined as the positioning point so that the displacement information of the risk object can be obtained subsequently based on the positioning points between adjacent selected frames.

[0113] S2322, obtain the shortest distance between the positioning point and the boundary of the selected area. When the shortest distance is greater than the critical distance threshold, obtain the regional position of the selected area in the current selected frame, and determine the monitoring sub-region corresponding to the regional position in the next selected frame as the selected area.

[0114] Understandably, the selected area boundary is the boundary of the monitoring sub-area where the positioning point is located. For example, when the positioning point is in area A, the distance from the positioning point to the boundary line of area A can be obtained based on the coordinates of the positioning point, thus determining the shortest distance. Since the risk object does not move a large distance in adjacent selected frames, the probability of the risk object being in the current area or the area closest to the boundary of the current area in the next selected frame is the highest. Therefore, when the shortest distance is greater than the critical distance threshold, the area position of the selected area in the current selected frame can be obtained, and the selected area in the next selected frame can be determined as the monitoring sub-area corresponding to the area position. For example, when the area position of the selected area in the current selected frame is obtained as area A, the monitoring sub-area corresponding to A in the current selected frame can be determined as the selected area in the next selected frame.

[0115] Among them, the critical distance threshold is the critical threshold for judging the distance of the risk object from the boundary of the area. When the shortest distance is greater than the critical distance threshold, it can be said that the risk object is far from the boundary of the area. When the shortest distance is less than the critical distance threshold, it can be said that the risk object is close to the boundary. In the next selected frame, the risk object may move to other areas. The area location is the location of the corresponding monitoring sub-area, such as the location A corresponding to the office area.

[0116] S2323, when the shortest distance is less than or equal to the critical distance threshold, the movement trajectory of the risk object is determined based on the location points corresponding to the risk object in multiple selected frames corresponding to the historical time period.

[0117] Understandably, when the shortest distance is less than or equal to the critical distance threshold, it indicates that the current risk object is located at the boundary of the area. Therefore, the corresponding movement trajectory can be obtained based on the location points of the risk object in multiple selected frames within the historical time period. When the location points in multiple selected frames are located at the same coordinate position, the corresponding movement trajectory is a single point, indicating that the corresponding risk object has not moved. When the movement trajectory is not a single point, it indicates that the risk object has moved, thus allowing the selection area corresponding to the next selected frame to be predicted, so that the corresponding displacement distance can be determined.

[0118] The historical time period is the time period before the time corresponding to the currently selected frame. For example, if the time corresponding to the currently selected frame is 10:00:01, then the historical time period can be 9:50:00-10:00:00. The movement trajectory is the trajectory of the risk object's position movement.

[0119] S2324, Obtain the location of the risk object corresponding to the unchanged movement trajectory, and determine the monitoring sub-region corresponding to the location in the next selected frame as the selected area.

[0120] Understandably, when the risk object corresponding to the movement trajectory does not change, the location of the risk object in the current selected frame can be obtained, and the selected area in the next selected frame can be determined as the monitoring sub-area corresponding to the location. For example, when the location corresponds to office area A, the selected area in the next selected frame is the office area corresponding to A.

[0121] S2325, determine the tangential direction corresponding to the endpoint of the changed action trajectory as the prediction direction, obtain the monitoring sub-region adjacent to the selected area and located in the prediction direction, as well as the area position of the selected area, and determine the monitoring sub-region corresponding to the area position in the next selected frame as the selected area.

[0122] It is understandable that the endpoint position is the position corresponding to the end of the movement trajectory, the tangential direction is the direction along the tangent of the movement trajectory at the endpoint position, and the prediction direction is the direction for predicting the selected area.

[0123] It is not difficult to understand that determining the location of the selected area and the monitoring sub-areas adjacent to the selected area and in the prediction direction, for example, Figure 3 The diagram shows a method for determining a selected area provided by the present invention. When the selected area is an office area, the corresponding area location is the area location corresponding to number A. When the prediction direction is as shown in the diagram, the monitoring sub-area adjacent to the selected area A and in the prediction direction can be determined to be the corridor area B. Then, the selected area corresponding to the next selected frame can be determined to be the corridor area corresponding to number B.

[0124] S2326, perform risk source identification on the selected area, and continue to obtain the second contour corresponding to the risk object in the selected area, and obtain the displacement distance based on the point distance between the positioning points corresponding to the same risk object in adjacent selected frames.

[0125] Specifically, risk sources are identified in selected areas of selected frames to obtain the second contour corresponding to the risk source, and then the displacement distance is obtained based on the coordinates of the positioning points of the same risk object in adjacent selected frames.

[0126] The point distance refers to the distance between the corresponding positioning points at different locations.

[0127] S233, determine the directional characteristics of the risk object based on the displacement distance, and generate two types of safety layers according to the directional characteristics and the second contour of the risk object, wherein the directional characteristics include dynamic characteristics and static characteristics.

[0128] It is understandable that the dynamic characteristics are the characteristics of the distance the risk object moves, including dynamic characteristics and static characteristics. Static characteristics mean that when the displacement distance of the risk object is less than a certain value, the risk object can be regarded as static. Dynamic characteristics mean that when the displacement distance of the risk object is greater than a certain value, the risk object can be regarded as dynamically moving.

[0129] It's easy to understand that when the movement of a risky object is small, such as when a person touches a paper document on a table while picking up an item, the paper document will move slightly, and the corresponding displacement distance will be very small. When the person moves the paper document, the corresponding displacement distance will increase. Due to the different displacement distances, the position in the image frame will also change. Thus, two types of security layers can be generated based on the movement characteristics and the second contour of the risky object, in order to reduce the risk of information leakage and improve information security.

[0130] Based on the above embodiments, step S233 (determining the dynamic characteristics of the risk object based on the displacement distance, and generating two types of safety layers according to the dynamic characteristics and the second contour of the risk object, wherein the dynamic characteristics include dynamic features and static features) can be specifically implemented as follows:

[0131] S2331, retrieve the initial layer corresponding to the monitoring data, generate an occlusion contour corresponding to the second contour in the initial layer, and retrieve the occlusion pixel value to update the occlusion contour to obtain the second type of security layer.

[0132] Understandably, the initial layer is a transparent layer with the same size as the monitoring image frame, the occlusion outline is the outline of the occluded risk object in the initial layer, and the occlusion pixel value is the pixel value used to occlude information, such as white corresponding to a blank paper document.

[0133] It is easy to understand that by updating the pixel values ​​in the occlusion outline to the occlusion pixel values, a second-class security layer can be obtained. This allows the image information of the risky object to be occluded and hidden when the corresponding second-class security layer is superimposed on the corresponding image frame.

[0134] S2332, determine the movement characteristics of the risk object whose displacement distance is less than the movement threshold as static characteristics, and determine the movement characteristics of the risk object whose displacement distance is greater than or equal to the movement threshold as dynamic characteristics.

[0135] It is understandable that when the displacement distance is less than the movement threshold, the movement characteristics of the corresponding risk object are static, and when the displacement distance is greater than or equal to the movement threshold, the movement characteristics of the corresponding risk object are dynamic.

[0136] Among them, the movement threshold is a threshold for the movement distance, which can be set in advance by humans to determine the value of the movement characteristics of the risky object.

[0137] S2333, when the risk object is a static feature, the two types of security layers corresponding to the risk object in adjacent selected frames are the same.

[0138] It is understandable that when the risk object is a static feature, it means that the positions of the risk objects corresponding to adjacent selected frames are approximately the same, and the corresponding Class II security layers will also be consistent. Therefore, there is no need to regenerate the Class II security layers, which reduces the amount of data processing and improves the hiding efficiency.

[0139] S2334, when the risk object is a dynamic feature, generate corresponding two types of security layers based on the second contour corresponding to each risk object in the adjacent selected frames.

[0140] It is understandable that when the risk object has a moving feature, the occlusion contour can be regenerated based on the second contour corresponding to each risk object in the adjacent selected frames, thereby obtaining the corresponding second-class security layer, so as to accurately cover and hide the risk object in the corresponding selected frame.

[0141] S24, the monitoring data is overlaid with the second type of security layer to obtain the display data, wherein the security layer includes a first type of security layer and a second type of security layer.

[0142] It is understandable that by overlaying the second-class security layer on the monitoring data, a displayable data can be obtained, thereby hiding information with potential security risks, reducing the risk of information leakage, and improving the security of information confidentiality.

[0143] S3. Perform a missing data analysis on the displayed data based on the safety calibration model to determine the calibration adjustment strategy corresponding to the displayed data.

[0144] It is understandable that when identifying risk sources in monitoring data based on pre-configured identification parameters, there may be instances where risk objects are missed or where there is a certain distance deviation in the displacement distance, resulting in the security layer not completely obscuring the corresponding risk object. Therefore, it is necessary to adjust the security layer in the corresponding displayed data in order to completely obscure and hide the risk object, reduce the risk of information leakage, and improve information security.

[0145] Among them, the security calibration model is the model of the security layer in the calibration display data.

[0146] Based on the above embodiments, step S3 (performing a missing data analysis on the displayed data according to the security calibration model to determine the calibration adjustment strategy corresponding to the displayed data) can be implemented as follows:

[0147] S31, based on the security calibration model, a preset number of image frames are randomly selected from the displayed data as calibration frames, the abnormal contours selected by the monitoring terminal in the calibration frames are obtained, and the image frames in the monitoring data corresponding to the calibration frames are determined as comparison frames.

[0148] It is understandable that the preset quantity is the number of image frames selected in advance, the calibration frame is the image frame used to calibrate the hiding of risk object information, the abnormal contour is the contour of the risk object that shows occlusion abnormality, and the comparison frame is the image frame in the monitoring data that corresponds to the calibration frame without the superimposed security layer.

[0149] It's easy to understand that regulators can mark abnormal contours on the calibration frame. For example, some paper documents might not be completely obscured, or some paper documents might not be identified as risky items and therefore not obscured. The corresponding not-completely-obscured risky item contours can be marked and selected as abnormal contours. This allows for the extraction of comparison frames from the monitoring data that correspond to the calibration frame time, enabling subsequent similarity comparisons and the adoption of appropriate strategies for adjustments. This ensures accurate obscuring and hiding of risky items, reducing the risk of information leakage.

[0150] S32, determine a reference contour in the comparison frame that has the same positioning point as the abnormal contour, and perform a similarity comparison between the reference contour and the abnormal contour.

[0151] It is understandable that the reference contour is the risk object contour in the comparison frame that has the same location point as the abnormal contour. The similarity comparison between the reference contour and the abnormal contour can be used to determine the corresponding abnormal situation, which is the cause of occlusion missing or incomplete occlusion. In this way, the corresponding strategy can be determined for calibration and adjustment. The similarity comparison is the comparison of the degree of similarity between the reference contour and the abnormal contour.

[0152] It is worth mentioning that the parameters for similarity comparison include at least pixel values ​​and dimensions. The pixel values ​​in the reference contour are compared with the pixel values ​​in the abnormal contour, and the dimensions of the reference contour are compared with the dimensions of the abnormal contour to obtain pixel value comparison values ​​and dimension comparison values. The two are combined to calculate the similarity, so as to determine the type of abnormal contour in the subsequent process.

[0153] S33, abnormal contours with a similarity greater than or equal to the similarity threshold are identified as unoccluded contours, and abnormal contours with a similarity less than the similarity threshold are identified as occluded or incomplete contours.

[0154] Understandably, when the similarity is greater than or equal to the similarity threshold, it means that the corresponding reference contour is consistent with the abnormal contour, and the pixel value similarity is also similar. This means that the corresponding abnormal contour in the calibration frame can be identified as consistent with the reference contour of the actual risk object. The abnormal contour of the risk object in the calibration frame is not occluded, so it can be determined as an unoccluded contour. When the contour similarity is less than the similarity threshold, it means that the risk object in the corresponding calibration frame is partially occluded, but not completely occluded. Therefore, it can be determined as an occluded incomplete contour, so that corresponding calibration methods can be adopted according to different types of abnormal contours to improve the accuracy of information occlusion and reduce the risk of information leakage.

[0155] Among them, the similarity threshold is the threshold corresponding to the similarity between the reference contour and the comparison contour. It can be set in advance by humans. The unoccluded contour is the risk object contour that is not occluded, and the occluded incomplete contour is the risk object contour that is not completely occluded.

[0156] S34, determine the completion adjustment strategy corresponding to the display data that only has the unobstructed contour, determine the coverage adjustment strategy corresponding to the display data that only has the obstructed and incomplete contour, and determine the comprehensive adjustment strategy corresponding to the display data that has both the unobstructed contour and the obstructed and incomplete contour.

[0157] The calibration adjustment strategies include a completion adjustment strategy, a coverage adjustment strategy, and a comprehensive adjustment strategy.

[0158] Understandably, when the displayed data only contains unobstructed outlines, calibration and adjustment can be performed using corresponding completion adjustment strategies. For example, the sensitivity of the corresponding recognition parameters can be adjusted to make it more sensitive and more accurately identify the corresponding risk objects, so as to generate corresponding obstruction outlines to cover the risk objects. When the displayed data only contains obstructed or incomplete outlines, the displayed data can be calibrated and adjusted using coverage adjustment strategies. For example, the parameters of the corresponding obstruction outline can be increased to make the obstruction outline larger and cover the unobstructed part. When the displayed data contains both unobstructed outlines and obstructed or incomplete outlines, calibration can be performed using a comprehensive adjustment strategy to hide the corresponding risk objects and reduce the risk of information leakage.

[0159] It is easy to understand that by determining the corresponding calibration and adjustment strategies based on the abnormal contours of different situations, targeted adjustments can be made subsequently to make the adjusted parameters more accurate, thereby improving the accuracy of information occlusion and reducing the risk of information leakage.

[0160] S4, Based on the calibration adjustment strategy, adjust the layer attribute parameters corresponding to the displayed data to obtain the safety adjustment data of the displayed data.

[0161] Specifically, the layer attribute parameters corresponding to the displayed data are adjusted according to the calibration and adjustment strategy to obtain security adjustment data, so as to adjust the security layer. This can mask abnormal outlines in the corresponding displayed data, thereby improving the accuracy of information masking and reducing the risk of information leakage.

[0162] Among them, the layer attribute parameters are the parameters for generating the safety layer, and the safety adjustment data are the data for adjusting the safety layer corresponding to the abnormal outline in the displayed data.

[0163] Based on the above embodiments, step S4 (adjusting the layer attribute parameters corresponding to the display data based on the calibration adjustment strategy to obtain the safety adjustment data of the display data) can be implemented as follows:

[0164] S41, count the first number of unobstructed contours and the second number of obstructed or incomplete contours in each monitoring sub-region of the displayed data.

[0165] Understandably, since the risk objects differ in different monitoring sub-regions, parameters can be adjusted based on the number of corresponding abnormal contours in each sub-region, combined with appropriate adjustment strategies. This can improve adjustment efficiency while ensuring the accuracy of the safety layer, thereby enhancing the accuracy of risk object occlusion.

[0166] The first quantity is the number of unobstructed contours in each monitoring sub-region, and the second quantity is the number of obstructed or incomplete contours in each monitoring sub-region.

[0167] Through the above implementation methods, the present invention can determine a first quantity and a second quantity, so as to subsequently determine the adjustment parameters of the corresponding monitoring sub-region based on the first quantity and the second quantity, so as to accurately cover the corresponding abnormal contours and improve the security of information.

[0168] It is easy to understand that when there are only unoccluded outlines in the monitored sub-region, the corresponding completion adjustment strategy will only have the first quantity. When there are only occluded and incomplete outlines in the monitored sub-region, the corresponding coverage adjustment strategy will only refer to the second quantity for parameter adjustment. When there are both unoccluded outlines and occluded and incomplete outlines in the monitored sub-region, the corresponding monitored sub-region will have both the first and second quantities, and the adjustment parameters will be determined based on the first and second quantities.

[0169] S42, a sensitivity adjustment coefficient is obtained based on the ratio of the first quantity to the reference sensitivity value, and an expansion adjustment coefficient is obtained based on the ratio of the second quantity to the reference expansion value.

[0170] Understandably, the baseline sensitivity value is a preset baseline value for sensitivity. Under standard conditions, it represents the number of risk objects that cannot be identified. The higher the first number, the more risk objects that cannot be identified. The higher the sensitivity adjustment coefficient obtained by the ratio of the first number to the baseline sensitivity value, the more risk objects that cannot be identified, and the greater the adjustment range required. Similarly, the higher the second number, the more risk objects that are not completely covered, and the greater the proportion of the outline that needs to be expanded to completely cover the missing outline.

[0171] Among them, the sensitivity adjustment coefficient is the coefficient for adjusting the sensitivity of identifying risk objects, that is, the ratio of the first quantity to the baseline sensitivity value; the baseline amplification value is the preset baseline value of contour amplification, that is, the parameter value of the occluded contour under normal conditions; and the amplification adjustment coefficient is the coefficient for adjusting the contour amplification ratio, that is, the ratio of the second quantity to the baseline amplification value.

[0172] S43, based on the sensitivity adjustment coefficient, traverse the sensitivity adjustment table to determine the preset sensitivity parameter corresponding to the preset coefficient range where the sensitivity adjustment coefficient is located as the first adjustment parameter.

[0173] Understandably, by iterating through the sensitivity adjustment table using the determined sensitivity adjustment coefficient, a preset coefficient range in which the sensitivity adjustment coefficient falls is selected. The preset sensitivity parameter corresponding to the preset coefficient range is then used as the first adjustment parameter. This allows the sensitivity parameter of the corresponding monitoring sub-area to be adjusted to the first adjustment parameter, making the identification of risk objects more accurate, avoiding the omission of risk objects, thereby improving the accuracy of information obstruction and reducing the risk of information leakage.

[0174] The sensitivity adjustment table is a reference table for adjusting sensitivity parameters. It includes preset coefficient ranges and corresponding preset sensitivity parameters. The preset coefficient ranges are pre-set sensitivity coefficient ranges, and the preset sensitivity parameters are preset sensitivity parameters. The preset coefficient ranges and preset sensitivity parameters are one-to-one correspondences. The first adjustment parameter is the preset sensitivity parameter corresponding to the preset coefficient range in which the sensitivity adjustment coefficient is located.

[0175] It is easy to understand that the first adjustment parameter is determined based on the sensitivity adjustment coefficient, so that the parameters of the corresponding monitoring sub-area can be adjusted in the future to improve the accuracy of data hiding and reduce the risk of information leakage.

[0176] The sensitivity adjustment table can be obtained through the following steps:

[0177] S431, obtain multiple identification levels configured by the monitoring terminal, each identification level is configured with a corresponding preset sensitivity parameter, wherein the higher the identification level, the more preset sensitivity parameters are corresponding to the identification level.

[0178] It is understandable that the identification level is the level of risk object identification. The higher the level, the more preset sensitivity parameters there are, and thus the higher the sensitivity, which can more accurately identify risk objects in the monitoring sub-area in order to hide information.

[0179] S432, arrange the preset sensitivity parameters from largest to smallest according to the identification level to obtain a parameter sequence, and obtain the preset coefficient range configured by the monitoring terminal for each preset sensitivity parameter to obtain a sensitivity adjustment table. The higher the identification level, the larger the preset coefficient range corresponding to the preset sensitivity parameter.

[0180] It is understandable that the parameter sequence is a numerical sequence of preset sensitivity parameters, that is, a numerical sequence after sorting the preset sensitivity parameters according to the size order of the identification level. Thus, it can be arranged sequentially according to the preset sensitivity parameters set by the regulatory end and the corresponding preset coefficient range to obtain a sensitivity adjustment table. Therefore, the larger the preset coefficient range in which the first adjustment parameter is located, the higher the identification level of the corresponding preset sensitivity parameter, the more accurate the identification of risky objects, and thus improve the accuracy of information occlusion of risky objects and reduce the risk of information leakage.

[0181] S44, according to the expansion adjustment coefficient, the expansion correspondence table is traversed to determine the preset expansion parameter corresponding to the preset coefficient interval where the expansion adjustment coefficient is located as the second adjustment parameter.

[0182] Understandably, by traversing the determined augmentation adjustment coefficient in the augmentation correspondence table, the preset coefficient range in which the augmentation adjustment coefficient is located is selected. The preset augmentation parameter corresponding to the preset coefficient range is used as the second adjustment parameter so that the augmentation parameter of the corresponding monitoring sub-region can be adjusted to the second adjustment parameter in the future. This allows the occlusion contour to completely cover the risk object, thereby improving the integrity of information occlusion and reducing the risk of information leakage.

[0183] The expansion correspondence table is a reference table for adjusting expansion parameters, which includes preset coefficient ranges and corresponding preset expansion parameters. The preset coefficient range is a pre-set coefficient range, and the preset expansion parameter is a preset expansion parameter. The preset coefficient range and the preset expansion parameter are in one-to-one correspondence. The second adjustment parameter is the preset expansion parameter corresponding to the preset coefficient range where the expansion adjustment coefficient is located.

[0184] It is easy to understand that the corresponding second adjustment parameter is determined based on the augmentation adjustment coefficient, so that the parameters of the corresponding monitoring sub-area can be adjusted in the future, thereby improving the accuracy of data hiding and reducing the risk of information leakage.

[0185] The expanded correspondence table can be obtained through the following steps:

[0186] S441, acquire multiple augmentation levels configured by the monitoring terminal, each augmentation level is configured with a corresponding augmentation ratio, wherein the higher the augmentation level, the greater the augmentation ratio corresponding to the augmentation level.

[0187] It is understandable that the augmentation level is the level of the risk object's occlusion outline size. The higher the level, the greater the augmentation ratio, and thus the larger the corresponding occlusion outline, which can completely occlude the risk object in the monitoring sub-area for information hiding.

[0188] The augmentation ratio is the ratio by which the occluded contour is enlarged.

[0189] S442, Arrange the expansion ratios from largest to smallest according to the expansion level to obtain an expansion sequence, and obtain the preset coefficient range configured by the regulatory end for each expansion ratio to obtain an expansion correspondence table.

[0190] The preset expansion parameters include the expansion ratio. The larger the expansion level, the larger the preset coefficient range corresponding to the expansion ratio.

[0191] It is understandable that the expansion sequence is a sequence after sorting the expansion ratios, that is, sorting each expansion ratio in descending order of expansion level. This allows the preset coefficient range configured by the regulatory end for each expansion ratio to be displayed at each expansion level, thereby obtaining the corresponding expansion correspondence table. This table is then used to determine the second adjustment parameter corresponding to each monitoring sub-region.

[0192] S45, based on the first adjustment parameter and / or the second adjustment parameter, obtain the layer adjustment parameter, and bind the layer adjustment parameter and the corresponding monitoring sub-region to obtain safety adjustment data.

[0193] It is understandable that the layer adjustment parameters are the parameters for adjusting the security layer, including the first adjustment parameter and the second adjustment parameter. The layer adjustment parameters are bound to the corresponding monitoring sub-area to obtain security adjustment data, so that risky objects in each monitoring sub-area can be accurately identified and blocked, reducing the risk of information leakage and improving information security.

[0194] See Figure 4 This is a schematic diagram of the structure of a security monitoring system based on image recognition provided in an embodiment of the present invention. The data processing system of the security monitoring system based on image recognition includes:

[0195] The identification module is used to determine multiple monitoring sub-regions corresponding to the monitoring data based on the regional configuration information of the regulatory end, identify risk sources in each monitoring sub-region, and determine the risk objects in the monitoring sub-region.

[0196] The hiding module is used to obtain the object attributes of the risk object, generate a security layer corresponding to the risk object based on the object attributes, and hide the monitoring data based on the security layer to obtain the display data.

[0197] The analysis module is used to perform missing data analysis on the displayed data based on the security calibration model, and to determine the calibration adjustment strategy corresponding to the displayed data.

[0198] The adjustment module is used to adjust the layer attribute parameters corresponding to the displayed data based on the calibration adjustment strategy to obtain the safety adjustment data of the displayed data.

[0199] The adjustment module is used to adjust the layer attribute parameters corresponding to the displayed data based on the calibration adjustment strategy to obtain the safety adjustment data of the displayed data.

[0200] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A security monitoring method based on image recognition, characterized in that, include: Based on the regional configuration information of the regulatory end, multiple monitoring sub-regions corresponding to the monitoring data are determined, and risk sources are identified in each monitoring sub-region to determine the risk objects in the monitoring sub-region; Obtain the object attributes of the risky object, generate a security layer corresponding to the risky object based on the object attributes, and perform risk hiding on the monitoring data based on the security layer to obtain the displayed data, including: When the object attribute of the risk object is a stable attribute, the operation status of the risk object is identified, and the sub-location of the region corresponding to the risk object whose operation status is on is obtained. Based on the sub-location of the area, a type of security layer corresponding to the risk object is retrieved, and the monitoring data is overlaid with the type of security layer to obtain the displayed data; or... When the object attribute of the risk object is a variable attribute, obtain the displacement information corresponding to the risk object, and generate a second-class safety layer corresponding to the risk object based on the displacement information. The monitoring data is overlaid with the second type of security layer to obtain the displayed data, wherein the security layer includes a first type of security layer and a second type of security layer; Based on the safety calibration model, a missing data analysis is performed on the displayed data to determine the corresponding calibration adjustment strategy. Based on the calibration adjustment strategy, the layer attribute parameters corresponding to the displayed data are adjusted to obtain the safety adjustment data of the displayed data.

2. The method according to claim 1, characterized in that, When the object attribute of the risky object is stable, the operational status of the risky object is identified, and the sub-location of the region corresponding to the risky object whose operational status is "on" is obtained, including: Extract the first contour of the risk object, and determine the first pixel point whose pixel value is located within the black screen pixel range in the first contour; The black screen ratio corresponding to the risk object is determined based on the ratio of the number of the first pixel to the total number of pixels in the first contour. The operating state corresponding to the risk object whose black screen ratio is less than the black screen threshold is determined to be the on state, and the sub-location of the risk object in the monitoring sub-region is obtained.

3. The method according to claim 1, characterized in that, When the object attribute of the risky object is a variable attribute, the displacement information corresponding to the risky object is obtained, and a second-class safety layer corresponding to the risky object is generated based on the displacement information, including: Multiple image frames corresponding to the monitoring data are arranged in chronological order to obtain a monitoring sequence, and the image frames in the monitoring sequence are selected sequentially. Obtain the second contour of the risk object in the current selected frame, determine the selected area in the next selected frame based on the positioning point corresponding to the second contour, and determine the displacement distance of the risk object in the adjacent selected frame based on the selected area; The movement characteristics of the risk object are determined based on the displacement distance, and two types of safety layers are generated based on the movement characteristics and the second contour of the risk object. The movement characteristics include dynamic features and static features.

4. The method according to claim 3, characterized in that, Obtain the second contour of the risk object in the current selected frame, determine the selected area in the next selected frame based on the positioning point corresponding to the second contour, and determine the displacement distance of the risk object in the adjacent selected frames based on the selected area, including: Extract the second contour corresponding to the risk object in the current selected frame, and determine the center point of the second contour as the positioning point; Obtain the shortest distance between the positioning point and the boundary of the selected area. When the shortest distance is greater than the critical distance threshold, obtain the regional position of the selected area in the current selected frame, and determine the monitoring sub-region corresponding to the regional position in the next selected frame as the selected area. When the shortest distance is less than or equal to the critical distance threshold, the movement trajectory of the risk object is determined based on the location points corresponding to the risk object in multiple selected frames corresponding to the historical time period. Obtain the location of the risk object corresponding to the unchanged movement trajectory, and determine the monitoring sub-region corresponding to the location in the next selected frame as the selected area; The tangential direction corresponding to the endpoint of the changed action trajectory is determined as the prediction direction. The monitoring sub-regions adjacent to the selected area and located in the prediction direction, as well as the area position of the selected area, are obtained. The monitoring sub-region corresponding to the area position in the next selected frame is determined as the selected area. Risk sources are identified in the selected area, and the second contour corresponding to the risk object in the selected area is obtained. The displacement distance is obtained based on the point distance between the positioning points corresponding to the same risk object in adjacent selected frames.

5. The method according to claim 3, characterized in that, Based on the displacement distance, the dynamic characteristics of the hazardous object are determined. Two types of safety layers are generated based on the dynamic characteristics and the second contour of the hazardous object. The dynamic characteristics include dynamic features and static features, including: The initial layer corresponding to the monitoring data is retrieved, and an occlusion contour corresponding to the second contour is generated in the initial layer. The occlusion pixel value is retrieved to update the occlusion contour to obtain the second type of security layer. The movement characteristics of a risk object whose displacement distance is less than the movement threshold are determined to be static characteristics, and the movement characteristics of a risk object whose displacement distance is greater than or equal to the movement threshold are determined to be dynamic characteristics. When the risk object is a static feature, the two types of security layers corresponding to the risk object in adjacent selected frames are the same; When the risk object is a dynamic feature, two types of security layers are generated based on the second contour corresponding to each risk object in the adjacent selected frames.

6. The method according to claim 1, characterized in that, Based on the security calibration model, a missing data analysis is performed on the displayed data to determine the corresponding calibration adjustment strategy, including: Based on the security calibration model, a preset number of image frames are randomly selected from the displayed data as calibration frames. The abnormal contours selected by the monitoring terminal in the calibration frames are obtained, and the image frames in the monitoring data corresponding to the calibration frames are determined as comparison frames. Determine a reference contour in the comparison frame that has the same location point as the abnormal contour, and compare the similarity between the reference contour and the abnormal contour. Abnormal contours with a similarity greater than or equal to the similarity threshold are identified as unoccluded contours, while abnormal contours with a similarity less than the similarity threshold are identified as occluded or incomplete contours. The display data that only contains the unobscured outline corresponds to the completion adjustment strategy; the display data that only contains the obscured and incomplete outline corresponds to the coverage adjustment strategy; and the display data that contains both the unobscured outline and the obscured and incomplete outline corresponds to the comprehensive adjustment strategy. The calibration adjustment strategies include a completion adjustment strategy, a coverage adjustment strategy, and a comprehensive adjustment strategy.

7. The method according to claim 6, characterized in that, Based on the calibration adjustment strategy, the layer attribute parameters corresponding to the displayed data are adjusted to obtain the safety adjustment data of the displayed data, including: The first number of unoccluded contours and the second number of occluded or incomplete contours in each monitoring sub-region of the displayed data are counted. The sensitivity adjustment coefficient is obtained based on the ratio of the first quantity to the baseline sensitivity value, and the augmentation adjustment coefficient is obtained based on the ratio of the second quantity to the baseline augmentation value. Based on the sensitivity adjustment coefficient, the sensitivity adjustment table is traversed to determine the preset sensitivity parameter corresponding to the preset coefficient range where the sensitivity adjustment coefficient is located as the first adjustment parameter; The expansion correspondence table is traversed according to the expansion adjustment coefficient to determine the preset expansion parameter corresponding to the preset coefficient interval where the expansion adjustment coefficient is located as the second adjustment parameter. Layer adjustment parameters are obtained based on the first adjustment parameter and / or the second adjustment parameter, and the layer adjustment parameters and the corresponding monitoring sub-regions are bound to obtain safety adjustment data.

8. The method according to claim 7, characterized in that, The sensitivity adjustment table and the expanded correspondence table are obtained through the following steps: The system acquires multiple identification levels configured by the monitoring terminal. Each identification level is configured with a corresponding preset sensitivity parameter. The higher the identification level, the more preset sensitivity parameters are corresponding to that identification level. Arrange the preset sensitivity parameters from largest to smallest according to the identification level to obtain a parameter sequence, and obtain the preset coefficient range configured by the monitoring terminal for each preset sensitivity parameter to obtain a sensitivity adjustment table. The higher the identification level, the larger the preset coefficient range corresponding to the preset sensitivity parameter. The system acquires multiple expansion levels configured by the regulatory terminal, each expansion level being configured with a corresponding expansion ratio. The higher the expansion level, the greater the corresponding expansion ratio. The expansion sequence is obtained by arranging the expansion ratios from largest to smallest according to the expansion level, and the expansion correspondence table is obtained by obtaining the preset coefficient range configured by the regulatory end for each expansion ratio. The preset expansion parameters include the expansion ratio. The larger the expansion level, the larger the preset coefficient range corresponding to the expansion ratio.

9. A security monitoring system based on image recognition implementing the method of claim 1, characterized in that, include: The identification module is used to determine multiple monitoring sub-regions corresponding to the monitoring data based on the regional configuration information of the regulatory end, identify risk sources in each monitoring sub-region, and determine the risk objects in the monitoring sub-regions. The hidden module is used to obtain the object attributes of the risk object, generate a security layer corresponding to the risk object based on the object attributes, and hide the monitoring data based on the security layer to obtain the display data; The analysis module is used to perform a missing data analysis on the displayed data based on the safety calibration model, and to determine the calibration adjustment strategy corresponding to the displayed data; The adjustment module is used to adjust the layer attribute parameters corresponding to the displayed data based on the calibration adjustment strategy to obtain the safety adjustment data of the displayed data.

Images