A method and system for automatically detecting and warning screen shooting behavior

Abstract

The present invention discloses an automatic detection and alarm method and system for screen snapping behavior, wherein the automatic detection and alarm method includes a file detection step, a screen snapping detection step and a detection closing step; the present invention turns on an object detection model to detect the screen snapping behavior when a protected file is opened, and turns off the object detection model when the protected file is closed, thereby ensuring that only the behavior of snapping the screen of the protected file is detected, thereby not affecting the normal behavior of snapping the screen of other files; the present invention uses a trained object detection model to detect the screen snapping behavior, thereby improving the accuracy of the detection; the present invention obtains the real-time video of the PC user through a camera, detects the screen snapping behavior of the real-time video, and ensures the real-time nature of the detection.

Description

Technical Field

[0001] The present invention belongs to the field of computer vision technology, and more specifically, relates to an automatic detection and alarm method and system for screen shooting behavior. Background Art

[0002] Nowadays, there is a lot of protected information in the fields of national government affairs, security, science and technology, and military affairs. This information exists in various storage media in the form of files. When relevant personnel use PCs to process these files, if there is a lack of supervision, they may use mobile phones, cameras and other devices to shoot the screen. This behavior is relatively secretive and extremely harmful.

[0003] Therefore, it is necessary to design an automatic detection and alarm method, which can detect the screen capture of protected files by mobile phones, cameras and other devices, while not affecting the normal screen capture of other files by relevant personnel.

[0004] Currently, there is little research in this field. Relevant research mainly focuses on detecting the behavior of using mobile phones, and lacks targeted detection of the behavior of taking screenshots of protected files.

[0005] In order to facilitate the understanding of the present invention, the following explains relevant terms and related concepts:

[0006] Object detection model: A commonly used deep learning model for completing target object detection tasks. The input of the object detection model is an image, and the output image usually contains the following two types of detection frames: (1) Correct detection frame: The target object is successfully detected and marked with a detection frame; (2) False detection frame: The background is detected as the target object, and the content in the detection frame is the background. Commonly used object detection models include SSD, YOLO, etc.

[0007] Data enhancement: Common techniques in the image field, mainly including the following eight operations: flip transformation, random pruning, color jitter, translation transformation, scale transformation, contrast transformation, random noise, and affine transformation;

[0008] Recall rate: A commonly used evaluation indicator in deep learning. The calculation formula is: recall = TP / (TP+FN), where TP represents true positives, that is, the number of samples predicted by the model as positive examples and actually positive examples, and FN represents false negatives, that is, the number of samples predicted by the model as negative examples and actually positive examples;

[0009] Accuracy: A commonly used evaluation indicator in deep learning. The calculation formula is: accuracy = (TP + TN) / (P + N), where TN represents true negatives, that is, the number of samples predicted by the model to be negative examples and actually are negative examples, and P + N represents the total number of samples. Summary of the invention

[0010] In view of the defects of the prior art, the purpose of the present invention is to provide an automatic detection and alarm method and system for screen shooting behavior, aiming to solve the problem of being able to detect the behavior of shooting protected information on the screen.

[0011] To achieve the above-mentioned purpose, the present invention provides an automatic detection and alarm method for screen shooting behavior, which is used for communication between a PC end and a detection end, wherein the PC end stores protected files and the detection end installs a trained object detection model; and comprises the following steps:

[0012] (1) File detection steps:

[0013] When the PC detects that the protected file is opened and the object detection model is not started, it sends a model start instruction to the detection end and executes step (2); when it detects that the protected file is not opened and the object detection model is running, it sends a model close instruction to the detection end and executes step (3);

[0014] (2) Screen capture detection steps:

[0015] The detection end turns on the camera and at the same time turns on the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera, and when the screen shooting behavior is detected, an alarm message is sent to the PC;

[0016] (3) Detection and closure steps:

[0017] The detection end turns off the camera and the object detection model.

[0018] Preferably, the file detection step includes the following sub-steps:

[0019] (1-1) Initialization sub-step:

[0020] The PC side sets the value of the model status to No;

[0021] There are two values ​​of the model status: yes and no; yes means that the object detection model is turned on; no means that the object detection model is turned off;

[0022] (1-2) Detection sub-steps:

[0023] The PC side queries the file names of all open windows at a set time interval;

[0024] When it is found that the file name has a protection mark and the value of the model state is no, a model start instruction is sent to the detection end, and the value of the model state is set to yes, and then the screen shooting detection step is performed;

[0025] When it is found that the file name does not have a protection mark and the value of the model state is yes, a model closing instruction is sent to the detection end, and the value of the model state is set to no, and then the detection closing step is performed;

[0026] The protection mark is included in the file name of the protected file.

[0027] Preferably, the screen capture detection step further includes the following steps: after the PC receives the alarm information, the PC closes the opened protected file or performs a screen lock operation.

[0028] Preferably, the object detection model training method comprises the following steps:

[0029] (1) Dataset division steps:

[0030] Performing data enhancement on the original picture set to obtain an enhanced picture set; the original picture set is composed of pictures, and the pictures include a shooting device and its camera;

[0031] Divide the enhanced image set into the original training set and the original test set according to the set ratio;

[0032] Adding the annotation file corresponding to each picture in the enhanced picture set to the original training set or the original test set where the picture is located to form a training set or a test set; the annotation file annotates the coordinate information of the shooting device in the corresponding picture;

[0033] Mark the object detection model as a model to be trained;

[0034] (2) Training steps:

[0035] Inputting the training set into the model to be trained, performing iterative training according to the set batch input amount and number of iterations, to obtain a training model;

[0036] (3) Test steps:

[0037] The training model is tested using the test set. When both the accuracy and the recall are not lower than the set values, the training model is determined to be a trained object detection model, and the process ends.

[0038] When the accuracy is lower than a set value, increasing the number of images in the training set;

[0039] When the recall rate is lower than the set value, one or more of the following operations are selected for execution: increasing the number of images in the training set, performing data enhancement on the training set, and increasing the number of iterations;

[0040] The training model is marked as a model to be trained, and steps (2) to (3) are repeatedly performed in sequence until a trained object detection model is obtained.

[0041] The present invention provides a model state detection method, which is used on a PC, and comprises the following steps:

[0042] (1) Model state initialization steps:

[0043] Set the value of the model state to No;

[0044] (2) Model status detection steps:

[0045] Query the file names of all open windows at set time intervals;

[0046] When it is found that the file name has a protection mark and the value of the model state is no, a model opening instruction is sent to the detection end, and the value of the model state is set to yes;

[0047] When it is found that the file name does not have a protection mark and the value of the model state is yes, a model closing instruction is sent to the detection end, and the value of the model state is set to no;

[0048] The protection mark is included in the file name of the protected file.

[0049] Preferably, the model status detection step also includes the following operation: after receiving the alarm information, closing the opened protected file or performing a screen lock operation.

[0050] The present invention provides a video real-time detection method, which is used for a detection end and comprises the following steps:

[0051] (1) Instruction receiving steps:

[0052] Receive a command from the PC. If it is a model opening command, execute step (2); if it is a model closing command, execute step (3);

[0053] (2) Detection steps:

[0054] Turn on the camera, and at the same time turn on the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera, and send an alarm message to the PC when the screen shooting behavior is detected;

[0055] (3) Closing steps:

[0056] Turn off the camera and the object detection model.

[0057] The present invention provides an automatic detection and alarm system for screen-shooting behavior, which is used for communication between a PC terminal and a detection terminal, wherein the PC terminal stores protected files and the detection terminal installs a trained object detection model; the system includes the following modules:

[0058] File detection module:

[0059] When the PC detects that the protected file is opened and the object detection model is not started, the model start instruction is sent to the detection end to execute the operation in the screen capture detection module; when the PC detects that the protected file is not opened and the object detection model is running, the model close instruction is sent to the detection end to execute the operation in the detection close module;

[0060] Screen capture detection module:

[0061] Used to turn on the camera at the detection end, and at the same time turn on the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera, and send an alarm message to the PC when the screen shooting behavior is detected;

[0062] Detect shutdown module:

[0063] Used to turn off the camera and the object detection model at the detection end.

[0064] The present invention provides a model state detection system for a PC, comprising the following modules:

[0065] Model state initialization module:

[0066] Used to set the value of the model state to No;

[0067] Model state detection module:

[0068] Used to query the file names of all open windows at a set time interval;

[0069] When it is found that the file name has a protection mark and the value of the model state is no, a model opening instruction is sent to the detection end, and the value of the model state is set to yes;

[0070] When it is found that the file name does not have a protection mark and the value of the model state is yes, a model closing instruction is sent to the detection end, and the value of the model state is set to no;

[0071] The protection mark is included in the file name of the protected file.

[0072] The present invention provides a real-time video detection system for a detection terminal, comprising the following modules:

[0073] Instruction receiving module:

[0074] Used to receive instructions from the PC. If it is a model opening instruction, it executes the operations in the detection module; if it is a model closing instruction, it executes the operations in the closing module;

[0075] Detection module:

[0076] Used to turn on the camera and at the same time turn on the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera, and send an alarm message to the PC when the screen shooting behavior is detected;

[0077] To close a module:

[0078] Used to turn off the camera and the object detection model.

[0079] The above technical scheme conceived by the present invention, compared with the prior art, because the present invention turns on the object detection model when the protected file is opened, and turns off the object detection model when the protected file is closed, thereby ensuring that only the behavior of taking a screen shot of the protected file is detected, thereby not affecting the normal behavior of taking a screen shot of other files; the present invention uses a trained object detection model to detect the screen shooting behavior, thereby improving the accuracy of the detection; the present invention obtains the real-time video of the PC user through the camera, detects the screen shooting behavior of the real-time video, and ensures the real-time nature of the detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 is a flow chart of an automatic detection and alarm method for screen-shooting behavior provided by an embodiment of the present invention;

[0081] Figure 2 It is a flowchart of a file detection step in an automatic detection and alarm method for screen-shooting behavior provided by an embodiment of the present invention;

[0082] Figure 3 It is a flow chart of a method for training an object detection model in an automatic detection and alarm method for screen-shooting behavior provided by an embodiment of the present invention;

[0083] Figure 4 is a flow chart of a model state detection method provided by an embodiment of the present invention;

[0084] Figure 5 It is a flow chart of a real-time video detection method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0085] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0086] like Figure 1As shown, this embodiment provides an automatic detection and alarm method for screen shooting behavior, which is used for communication between a PC end and a detection end, the PC end is a Windows or Linux platform; the detection end is hi3516DV300; the PC end stores protected files, and the detection end installs a trained object detection model; the following steps are included:

[0087] (1) File detection steps, such as Figure 2 As shown, it includes the following sub-steps:

[0088] (1-1) Initialization sub-step:

[0089] The PC side sets the value of the model status to No;

[0090] The model state is a variable with two values: yes and no; yes means the object detection model is turned on; no means the object detection model is turned off;

[0091] (1-2) Detection sub-steps:

[0092] The PC side queries the file names of all open windows at a set time interval; the set time interval is 5 seconds; the query method is: using the API provided by Windows or Linux to traverse all open windows and perform character matching operations. When a protection mark is matched, the traversal operation is terminated.

[0093] When it is found that a protection mark appears in the file name and the value of the model status is no, a model opening instruction is sent to the detection end, and the value of the model status is set to yes, and then step (2) is executed; when it is found that no protection mark appears in the file name and the value of the model status is yes, a model closing instruction is sent to the detection end, and the value of the model status is set to no, and then step (3) is executed; the protection mark is added to the file name of the protected file by the program on the PC side.

[0094] After receiving the alarm information, the PC closes the opened protected file or locks the screen; and sends a screenshot of the alarm information to the server connected to the PC.

[0095] (2) Screen capture detection steps:

[0096] The detection end turns on the camera and the object detection model at the same time to perform real-time detection on the real-time video stream of the PC user collected by the camera. When the screen-shooting behavior is detected, an alarm message is sent to the PC. There are 30 frames in 1 second of video. The object detection model uses a frame-by-frame detection method and can detect up to 33 frames in 1 second, thus ensuring the real-time detection.

[0097] (3) Detection and closure steps:

[0098] The detection end turns off the camera and object detection model.

[0099] Since the present invention starts the object detection model when the protected file is opened, and closes the object detection model when the protected file is closed, it ensures that only the behavior of taking a screenshot of the protected file is detected, thereby not affecting the normal behavior of taking a screenshot of other files;

[0100] The present invention obtains the real-time video of the PC user through the camera, detects the screen shooting behavior of the real-time video, and ensures the real-time nature of the detection.

[0101] like Figure 3 As shown, the training method of the object detection model includes the following steps:

[0102] (1) Dataset division steps:

[0103] Performing data enhancement on the original picture set, one data enhancement operation or multiple data enhancement operations may be performed to obtain an enhanced picture set; the original picture set is composed of pictures, and the pictures include a shooting device and its camera;

[0104] The enhanced image set is divided into an original training set and an original test set according to a set ratio; wherein the set ratio is 3:1;

[0105] The annotation file corresponding to each image in the enhanced image set is added to the original training set or the original test set where the image is located to form a training set or a test set; the coordinate information of the shooting device in the corresponding image is annotated in the annotation file; among them, the training set has 50,000 images and the test set has 10,000 images;

[0106] Mark the object detection model as the model to be trained; the object detection model is yolov3;

[0107] (2) Training steps:

[0108] Input the training set into the model to be trained, perform iterative training according to the set batch input size and number of iterations, and obtain the training model; wherein, an image and its corresponding annotation file are called a pair of files, the batch input size is 32 pairs of files, and the number of iterations is 30,000 times;

[0109] (3) Test steps:

[0110] Use the test set to test the training model. When the accuracy and recall are not lower than the set value, the training model is determined to be a trained object detection model and the process ends. The set value is 0.9.

[0111] When the accuracy is lower than the set value, increase the number of images in the training set, and add the images that produce false positives and their corresponding annotation files to the training set;

[0112] When the recall rate is lower than the set value, select one or more of the following operations to perform: increase the number of images in the training set, perform data enhancement on the training set, and increase the number of iterations; among them, one data enhancement operation can be performed or multiple operations can be performed.

[0113] Mark the training model as the model to be trained, and repeat steps (2) to (3) in sequence until a trained object detection model is obtained.

[0114] The present invention uses a trained object detection model to detect screen-shooting behavior, thereby improving the accuracy of detection.

[0115] like Figure 4 As shown, this embodiment provides a model state detection method for a PC, comprising the following steps:

[0116] (1) Model state initialization step:

[0117] Set the value of the model state to No;

[0118] (2) Model status detection steps:

[0119] The file names of all open windows are queried at the set time interval. When a protection mark is found in the file name and the value of the model status is no, a model opening instruction is sent to the detection end, and the value of the model status is set to yes; when no protection mark is found in the file name and the value of the model status is yes, a model closing instruction is sent to the detection end, and the value of the model status is set to no; the protection mark is included in the file name of the protected file.

[0120] After receiving the alarm information, close the opened protected file or lock the screen; and send a screenshot of the alarm information to the server connected to the PC.

[0121] like Figure 5 As shown, this embodiment provides a real-time video detection method, which is used for a detection terminal and includes the following steps:

[0122] (1) Instruction receiving steps:

[0123] Receive a command from the PC. If it is a model opening command, execute step (2); if it is a model closing command, execute step (3);

[0124] (2) Detection steps:

[0125] Turn on the camera and the object detection model to detect the real-time video stream of the PC user collected by the camera. When the screen capture behavior is detected, an alarm message is sent to the PC.

[0126] (3) Closing steps:

[0127] Turn off the camera and object detection model.

[0128] This embodiment provides an automatic detection and alarm system for screen-shooting behavior, which is used for communication between a PC and a detection terminal. The PC stores protected files and the detection terminal installs a trained object detection model. The system includes the following modules:

[0129] File detection module:

[0130] When the PC detects that the protected file is opened and the object detection model is not started, it sends a model start command to the detection end to execute the operation in the screen capture detection module; when it detects that the protected file is not opened and the object detection model is running, it sends a model close command to the detection end to execute the operation in the detection close module;

[0131] After receiving the alarm information, the PC closes the opened protected file or locks the screen; and sends a screenshot of the alarm information to the server connected to the PC.

[0132] Screen capture detection module:

[0133] It is used to open the camera at the detection end and start the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera. When the screen shooting behavior is detected, an alarm message is sent to the PC.

[0134] Detect shutdown module:

[0135] Used to turn off the camera and object detection model on the detection side.

[0136] Among them, the file detection module includes the following sub-modules:

[0137] Initialize the submodule:

[0138] Used to set the value of the model status to No on the PC side;

[0139] There are two values ​​for the model status: yes and no; yes means the object detection model is turned on; no means the object detection model is turned off;

[0140] Detection submodule:

[0141] Used to query the file names of all open windows on the PC at a set time interval;

[0142] When it is found that the file name has a protection mark and the value of the model status is no, a model start instruction is sent to the detection end, and the value of the model status is set to yes, and then the operation in the screen capture detection module is executed;

[0143] When it is found that the protection mark does not appear in the file name and the value of the model status is yes, a model closing instruction is sent to the detection end, and the value of the model status is set to no, and then the operation in the detection closing module is executed; the protection mark is included in the file name of the protected file.

[0144] After receiving the alarm information, the PC closes the opened protected file or locks the screen; and sends a screenshot of the alarm information to the server connected to the PC.

[0145] Among them, the training system of the object detection model includes the following modules:

[0146] Dataset partitioning module:

[0147] Used to perform data enhancement on the original picture set to obtain an enhanced picture set; the original picture set is composed of pictures, and the pictures include shooting devices and their cameras;

[0148] Divide the enhanced image set into the original training set and the original test set according to the set ratio;

[0149] The annotation file corresponding to each image in the enhanced image set is added to the original training set or the original test set where the image is located to form a training set or a test set; the annotation file is annotated with coordinate information of the shooting device in the corresponding image;

[0150] Mark the object detection model as a model to be trained;

[0151] Training Modules:

[0152] It is used to input the training set into the model to be trained, perform iterative training according to the set batch input amount and number of iterations, and obtain the training model;

[0153] Test Module:

[0154] Used to test the training model using the test set. When the accuracy and recall rates are not lower than the set values, the training model is determined to be a trained object detection model and the process ends.

[0155] When the accuracy is lower than the set value, increase the number of images in the training set;

[0156] When the recall rate is lower than the set value, select one or more of the following operations to perform: increase the number of images in the training set, perform data enhancement on the training set, and increase the number of iterations;

[0157] Mark the training model as the model to be trained, and repeat the operations in the training module and the testing module in sequence until a trained object detection model is obtained.

[0158] This embodiment provides a model state detection system for a PC, including the following modules:

[0159] Model state initialization module:

[0160] Used to set the value of the model state to No;

[0161] Model state detection module:

[0162] Used to query the file names of all open windows at a set time interval;

[0163] When it is found that the file name has a protection mark and the value of the model state is no, a model opening instruction is sent to the detection end, and the value of the model state is set to yes;

[0164] When it is found that the protection mark does not appear in the file name and the value of the model state is yes, a model closing instruction is sent to the detection end, and the value of the model state is set to no; the protection mark is included in the file name of the protected file.

[0165] After receiving the alarm information, close the opened protected file or lock the screen; and send a screenshot of the alarm information to the server connected to the PC.

[0166] This embodiment provides a real-time video detection system for a detection terminal, including the following modules:

[0167] Instruction receiving module:

[0168] Used to receive instructions from the PC. If it is a model opening instruction, it executes the operations in the detection module; if it is a model closing instruction, it executes the operations in the closing module;

[0169] Detection module:

[0170] It is used to turn on the camera and the object detection model to detect the real-time video stream of the PC user collected by the camera. When the screen shooting behavior is detected, an alarm message is sent to the PC.

[0171] To close a module:

[0172] Used to turn off the camera and object detection model.

[0173] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An automatic detection and alarm method for screen-shooting behavior, used for communication between a PC and a detection terminal, the PC storing protected files, and the detection terminal installing a trained object detection model; characterized in that: The following steps are involved: (1) File detection steps: When the PC detects that the protected file is opened and the object detection model is not started, it sends a model start instruction to the detection end and executes step (2); when it detects that the protected file is not opened and the object detection model is running, it sends a model close instruction to the detection end and executes step (3); (2) Screen capture detection steps: The detection end turns on the camera and at the same time turns on the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera, and when the screen shooting behavior is detected, an alarm message is sent to the PC; (3) Detection and shutdown steps: The detection end turns off the camera and the object detection model; The file detection step includes the following sub-steps: (1-1) Initialization sub-step: The PC side sets the value of the model status to No; There are two values ​​of the model status: yes and no; yes means that the object detection model is turned on; no means that the object detection model is turned off; (1-2) Detection sub-steps: The PC side queries the file names of all open windows at a set time interval; When it is found that the file name has a protection mark and the value of the model state is no, a model start instruction is sent to the detection end, and the value of the model state is set to yes, and then the screen shooting detection step is performed; When it is found that the file name does not have a protection mark and the value of the model state is yes, a model closing instruction is sent to the detection end, and the value of the model state is set to no, and then the detection closing step is performed; The protection mark is included in the file name of the protected file.

2. The automatic detection and alarm method for screen-shooting behavior according to claim 1, characterized in that: The screen capture detection step also includes the following steps: after the PC receives the alarm information, the opened protected file is closed or the screen is locked.

3. The automatic detection and alarm method for screen-shooting behavior according to claim 1, characterized in that: The training method of the object detection model includes the following data set division steps, training steps and testing steps: Steps to divide the data set: Perform data enhancement on the original picture set to obtain an enhanced picture set; The original image set consists of images, and the images include the shooting device and its camera; Divide the enhanced image set into the original training set and the original test set according to the set ratio; Adding the annotation file corresponding to each picture in the enhanced picture set to the original training set or the original test set where the picture is located to form a training set or a test set; the annotation file annotates the coordinate information of the shooting device in the corresponding picture; Mark the object detection model as a model to be trained; Training steps: Inputting the training set into the model to be trained, performing iterative training according to the set batch input amount and number of iterations, to obtain a training model; Test steps: The training model is tested using the test set. When both the accuracy and the recall are not lower than the set values, the training model is determined to be a trained object detection model, and the process ends. When the accuracy is lower than a set value, increasing the number of images in the training set; When the recall rate is lower than the set value, one or more of the following operations are selected for execution: increasing the number of images in the training set, performing data enhancement on the training set, and increasing the number of iterations; The training model is marked as a model to be trained, and steps (2) to (3) are repeatedly performed in sequence until a trained object detection model is obtained.

4. A model state detection method, used on a PC, where a protected file is stored, characterized in that: The following steps are involved: (1) Model state initialization step: Set the value of the model state to No; The model is specifically an object detection model; there are two values ​​for the model status: yes and no; yes means the object detection model is turned on; no means the object detection model is turned off; the trained object detection model is installed on the detection end; (2) Model status detection steps: Query the file names of all open windows at set time intervals; When it is found that the file name has a protection mark and the value of the model state is no, a model opening instruction is sent to the detection end, and the value of the model state is set to yes; When it is found that the file name does not have a protection mark and the value of the model state is yes, a model closing instruction is sent to the detection end, and the value of the model state is set to no; The protection mark is included in the file name of the protected file; The model start instruction is used to instruct the detection end to turn on the camera and start the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera. When the screen capture behavior is detected, an alarm message is sent to the PC. The model close command is used to instruct the detection end to close the camera and object detection model.

5. The model state detection method according to claim 4, characterized in that: The model status detection step also includes the following operations: after receiving the alarm information, closing the opened protected file or performing a screen lock operation.

6. An automatic detection and alarm system for screen-shooting behavior, used for communication between a PC and a detection terminal, the PC storing protected files, and the detection terminal installing a trained object detection model; characterized in that: Includes the following modules: File detection module: When the PC detects that the protected file is opened and the object detection model is not started, it sends a model start instruction to the detection end to execute the operation in the screen capture detection module; When it is detected that the protected file is not opened and the object detection model is running, a model closing instruction is sent to the detection end to execute the operation in the detection closing module; Screen capture detection module: Used to turn on the camera at the detection end, and at the same time turn on the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera, and send an alarm message to the PC when the screen shooting behavior is detected; Detect shutdown module: Used to turn off the camera and the object detection model at the detection end; The file detection module is specifically used to perform the following sub-steps: (1-1) Initialization sub-step: The PC side sets the value of the model status to No; There are two values ​​of the model status: yes and no; yes means that the object detection model is turned on; no means that the object detection model is turned off; (1-2) Detection sub-steps: The PC side queries the file names of all open windows at a set time interval; When it is found that the file name has a protection mark and the value of the model status is no, a model start instruction is sent to the detection end, and the value of the model status is set to yes, and then the screen capture detection step is performed; When it is found that the file name does not have a protection mark and the value of the model state is yes, a model closing instruction is sent to the detection end, and the value of the model state is set to no, and then the detection closing step is executed; The protection mark is included in the file name of the protected file.

7. A model status detection system, used on a PC, which stores protected files, characterized in that: Includes the following modules: Model state initialization module: Used to set the value of the model state to No; The model is specifically an object detection model; there are two values ​​for the model status: yes and no; yes means the object detection model is turned on; no means the object detection model is turned off; the trained object detection model is installed on the detection end; Model state detection module: Used to query the file names of all open windows at a set time interval; When it is found that the file name has a protection mark and the value of the model state is no, a model opening instruction is sent to the detection end, and the value of the model state is set to yes; When it is found that the file name does not have a protection mark and the value of the model state is yes, a model closing instruction is sent to the detection end, and the value of the model state is set to no; The protection mark is included in the file name of the protected file; The model start instruction is used to instruct the detection end to turn on the camera and start the object detection model to perform real-time detection on the real-time video stream of the PC user collected by the camera. When the screen capture behavior is detected, an alarm message is sent to the PC. The model close command is used to instruct the detection end to close the camera and object detection model.

Images