Motion recognition method and device based on privacy algorithm, equipment and medium

Abstract

The embodiment of the invention discloses an action recognition method, device and equipment based on a privacy algorithm and a medium, and the method comprises the steps: obtaining image data of a target object through edge computing equipment, and converting the image data into a first image through the edge computing equipment, the first image comprises at least one of a matchman image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data and a mixed image composed of the matchman image and the silhouette image; identifying actual action data of a real person in the image data based on the image data, and matching the actual action data with preset action data to obtain a matching result; and according to the matching result, updating the first image according to a preset rule to obtain a to-be-displayed image, and uploading the to-be-displayed image to a server side corresponding to the edge computing device.

Description

Technical Field

[0001] The present invention relates to the fields of privacy technology and image processing technology, and in particular to a privacy algorithm-based action recognition method, device, equipment and medium. Background Art

[0002] As an application combining traditional video technology with modern communication technology, video surveillance has attracted more and more attention at home and abroad. In recent years, with the popularization of broadband, the development of computer technology, and the improvement of image processing technology, video surveillance is becoming more and more widely used in various fields. With the popularization of broadband networks, video surveillance has gradually developed from local monitoring to remote monitoring, and remote network video monitoring systems represented by network video servers have emerged. In the field of intelligent maintenance, such as nursing homes and child care centers, video surveillance technology can provide real-time monitoring of the elderly, children and other groups that need special attention. By real-time monitoring and recording the dynamics of the guardian, abnormal situations can be discovered in time, providing a basis for intervention in emergency situations. In addition, the data recorded by video surveillance can also serve as the basis for subsequent analysis and research, helping medical staff to better understand the behavior patterns and health status of the guardian.

[0003] However, in existing video surveillance technologies, in order to protect the privacy of people in the video, video data is often processed by blurring and other means. However, the processed video data often blurs all information about the people in the video, making the information contained in the video data unclear. The processed video data is not conducive to obtaining information. Furthermore, in actual scenarios, for example, in scenarios where remote consultations are conducted based on video data, it is impossible to quickly and efficiently obtain consultation basis based on the processed video data, which makes the video data in actual applications contrary to the real-time requirements of video surveillance. Summary of the invention

[0004] In view of this, the present invention provides a privacy algorithm-based action recognition method, device, equipment and medium, which are used to solve the problem in the prior art that video data does not meet the real-time requirements of video surveillance in practical applications.

[0005] To achieve one, part, or all of the above purposes or other purposes, the present invention proposes a motion recognition method based on a privacy algorithm, comprising:

[0006] Acquire image data of a target object through an edge computing device, and convert the image data into a first image through the edge computing device, wherein the first image includes at least one of a stick figure image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the stick figure image and the silhouette image;

[0007] Using a preset recognition model, identifying actual action data of a real person in the image data according to the image data, and matching the actual action data with the preset action data to obtain a matching result;

[0008] According to the matching result, the first image is updated according to a preset rule to obtain an image to be displayed, and the image to be displayed is uploaded to a server corresponding to the edge computing device.

[0009] Optionally, the step of converting the image data into a first image by the edge computing device includes:

[0010] extracting a real-person image of the target object from the image data;

[0011] Extracting key points of a human body from the real person image based on the privacy algorithm to obtain key point data of the real person image;

[0012] A matchstick figure image in the first image corresponding to the real person in the image data is constructed according to the key point data.

[0013] Optionally, the step of converting the image data into a first image by the edge computing device further includes:

[0014] extracting a real-person image of the target object from the image data;

[0015] Acquire edge data of the real person image based on edge detection technology, and perform image segmentation on the real person image according to the edge data to obtain an initial image;

[0016] The initial image is filled according to a first preset filling rule to obtain a silhouette image in the first image corresponding to the real person in the image data.

[0017] Optionally, before the step of identifying actual action data of a real person in the image data according to the image data by using a preset recognition model, the step further includes:

[0018] Acquire a first preset data set, the first preset data set including first preset image data of a first preset group and first annotation information of the first preset image data, the first annotation information being used to characterize first action data of a real person in the first preset image data;

[0019] Acquire a second preset data set, the second preset data set including second preset image data of a second preset group and second annotation information of the second preset image data, the second annotation information being used to characterize second action data of a real person in the second preset image data;

[0020] A training set and a verification set are constructed according to the first preset data set and the second preset data set, and the convolutional neural network model is trained by the training set, and the trained convolutional neural network model is verified by the verification set to obtain the preset recognition model.

[0021] Optionally, the step of updating the first image according to a preset rule based on the matching result to obtain an image to be displayed includes:

[0022] When the first image is a match figure image corresponding to a real person in the image data, the match figure image is updated according to the deformation rule in the preset rule and the matching result to obtain a first display image in the images to be displayed;

[0023] When the first image is a silhouette image corresponding to a real person in the image data, the silhouette image is updated according to the matching result according to a second preset filling rule in the preset rules to obtain a second display image in the image to be displayed;

[0024] When the first image is a mixed image composed of the stickman image and the silhouette image, the first display image and the second display image are spliced ​​to obtain a third display image in the image to be displayed.

[0025] Optionally, the step of updating the match figure image according to the deformation rule in the preset rule and the matching result to obtain a first display image in the images to be displayed includes:

[0026] Determine a target key point and a target connecting line between the target key points in the key point data according to the matching result;

[0027] According to the deformation rules in the preset rules, the target key points and the target connecting lines are deformed to obtain deformation key points and deformation connecting lines;

[0028] The stick figure image is updated based on the deformation key points and the deformation connection lines to obtain a first display image among the images to be displayed.

[0029] Optionally, the step of uploading the image to be displayed to a server corresponding to the edge computing device includes:

[0030] Acquire basic data of the target object, wherein the basic data includes identity information of the target object and health monitoring data acquired according to the identity information;

[0031] Marking the image to be displayed according to the basic data to obtain a marked image to be displayed;

[0032] The step of marking the image to be displayed according to the basic data to obtain the marked image to be displayed includes:

[0033] Marking the basic data into the image to be displayed according to a preset marking rule to obtain a marked image to be displayed, wherein the preset marking rule at least includes a preset marking position of the basic data in the image to be displayed, a preset display method of the basic data, and a preset data form of the basic data in the image to be displayed;

[0034] The marked image to be displayed is uploaded to the server corresponding to the edge computing device.

[0035] On the other hand, the present application provides a motion recognition device based on a privacy algorithm, the device comprising:

[0036] a data acquisition module, configured to acquire image data of a target object through an edge computing device, and convert the image data into a first image through the edge computing device, wherein the first image includes at least one of a matchstick image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the matchstick image and the silhouette image;

[0037] An action recognition module, used to recognize actual action data of a real person in the image data according to the image data by using a preset recognition model, and match the actual action data with preset action data to obtain a matching result;

[0038] A data display module is used to update the first image according to a preset rule based on the matching result to obtain an image to be displayed, and upload the image to be displayed to a server corresponding to the edge computing device.

[0039] On the other hand, an embodiment of the present application provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the action recognition method based on the privacy algorithm as described above are performed.

[0040] On the other hand, an embodiment of the present application provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the action recognition method based on the privacy algorithm as described above are executed.

[0041] Implementing the embodiments of the present invention will have the following beneficial effects:

[0042] The image data of the target object is obtained through an edge computing device, and the image data is converted into a first image through the edge computing device, wherein the first image includes at least one of a stick figure image corresponding to a real person in the image data, a silhouette image corresponding to a real person in the image data, and a mixed image composed of the stick figure image and the silhouette image; the video data is processed by the edge computing device to avoid the transmission of video data with real person images, which can reduce the risk of video data leakage; the actual action data of the real person in the image data is identified according to the image data through a preset recognition model, and the actual action data is matched with the preset action data to obtain a matching result; the first image is updated according to the matching result according to a preset rule to obtain an image to be displayed, and the image to be displayed is uploaded to a server corresponding to the edge computing device; the first image is updated according to the preset rule, and the part that does not match the actual action data with the preset action data can be highlighted, that is, after the video data is privacy processed, the information contained in the video data is highlighted, which is convenient for data extraction according to the video data after privacy processing, and reduces the impact of the privacy of the video data on the information, so that the video data after privacy also meets the real-time requirements of video surveillance technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] in:

[0045] Figure 1 is a flowchart of a method for action recognition based on a privacy algorithm provided in an embodiment of the present application;

[0046] Figure 2 It is a structural diagram of a motion recognition device based on a privacy algorithm provided in an embodiment of the present application;

[0047] Figure 3 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application;

[0048] Figure 4 It is a structural schematic diagram of a computer-readable storage medium provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] like Figure 1 As shown, the embodiment of the present application provides an action recognition method based on a privacy algorithm, including:

[0051] S101, acquiring image data of a target object through an edge computing device, and converting the image data into a first image through the edge computing device, wherein the first image includes at least one of a matchstick image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the matchstick image and the silhouette image;

[0052] Exemplarily, the implementation forms of edge computing devices are diverse, including but not limited to cameras with embedded intelligent algorithms, local servers connected to multiple cameras, smartphones equipped with cameras, and any special devices or equipment that integrate camera technology and intelligent algorithms. The edge computing device at least includes an intelligent algorithm for converting the image data into a first image, that is, the edge computing device at least includes an intelligent algorithm for converting the image data into a stick figure image, an intelligent algorithm for converting the image data into a silhouette image, and an intelligent algorithm for merging the stick figure image with the silhouette image.

[0053] Exemplarily, the silhouette image is used to represent the outline of a real person, and the real person's body parts are filled with a preset color (black, gray) to protect the real person's privacy. In order to more clearly display the edge information of the silhouette, the edge of the silhouette can be enhanced, such as using a Sobel operator, a Prewitt (cross differential) operator, etc. to sharpen the edge.

[0054] S102, identifying actual action data of a real person in the image data according to the image data by using a preset recognition model, and matching the actual action data with preset action data to obtain a matching result;

[0055] Exemplarily, the actual motion data is human body posture parameters, and the actual motion data can be understood as actual joint angle data, actual limb length data, etc. when the human body moves.

[0056] For example, since different joints and limbs are required to move when the human body performs different behaviors, the human body's behaviors can be identified through the movements of the human body's joints and limbs.

[0057] S103: Update the first image according to a preset rule based on the matching result to obtain an image to be displayed, and upload the image to be displayed to a server corresponding to the edge computing device.

[0058] Exemplarily, the first image is updated according to preset rules based on the matching results, so that the information in the video data can be highlighted. The preset rules include but are not limited to deformation rules and recoloring rules, that is, the information is highlighted through deformation and / or recoloring.

[0059] Exemplarily, the image to be displayed is uploaded to the server corresponding to the edge computing device, that is, the privacy protection processing is completed before the video data leaves the local environment, ensuring the absolute security of personal privacy information during transmission.

[0060] The image data of the target object is obtained through an edge computing device, and the image data is converted into a first image through the edge computing device, wherein the first image includes at least one of a stick figure image corresponding to a real person in the image data, a silhouette image corresponding to a real person in the image data, and a mixed image composed of the stick figure image and the silhouette image; the video data is processed by the edge computing device to avoid the transmission of video data with real person images, which can reduce the risk of video data leakage; the actual action data of the real person in the image data is identified according to the image data through a preset recognition model, and the actual action data is matched with the preset action data to obtain a matching result; the first image is updated according to the matching result according to a preset rule to obtain an image to be displayed, and the image to be displayed is uploaded to a server corresponding to the edge computing device; the first image is updated according to the preset rule, and the part that does not match the actual action data with the preset action data can be highlighted, that is, after the video data is privacy processed, the information contained in the video data is highlighted, which is convenient for data extraction according to the video data after privacy processing, and reduces the impact of the privacy of the video data on the information, so that the video data after privacy also meets the real-time requirements of video surveillance technology. By using edge computing devices to capture and convert individuals in the video into silhouettes or outlines, only key edge information is retained. This not only effectively protects personal privacy, but also significantly reduces the amount of video and image data, thereby reducing the bandwidth and storage costs required for data transmission, allowing artificial intelligence models to perform calculations and analysis more efficiently, significantly improving overall system efficiency and reducing operating costs.

[0061] In a possible implementation, the step of converting the image data into a first image by the edge computing device includes:

[0062] extracting a real-person image of the target object from the image data;

[0063] Extracting key points of a human body from the real person image based on the privacy algorithm to obtain key point data of the real person image;

[0064] A matchstick figure image in the first image corresponding to the real person in the image data is constructed according to the key point data.

[0065] Exemplarily, a real-life image containing the target person's private information (such as appearance, body shape, etc.) is replaced with a skeleton image (stickman image) to avoid the risk of leakage of the target person's private information. At the same time, the object of image recognition is converted into a skeleton image (stickman image). Compared with a real-life image, the image quality requirement when using a skeleton image for image recognition is lower than the image quality requirement when using a real-life image for image recognition, that is, the pixel requirement is reduced. As the pixel requirement is reduced, the computational complexity of the image recognition process is also reduced, thereby improving the efficiency of image recognition.

[0066] In a possible implementation, the step of converting the image data into a first image by the edge computing device further includes:

[0067] extracting a real-person image of the target object from the image data;

[0068] Acquire edge data of the real person image based on edge detection technology, and perform image segmentation on the real person image according to the edge data to obtain an initial image;

[0069] The initial image is filled according to a first preset filling rule to obtain a silhouette image in the first image corresponding to the real person in the image data.

[0070] Exemplarily, a video stream (the image data) is captured in real time by a camera (edge ​​computing device), and the video stream is preprocessed, including denoising, color correction, etc., to obtain a preprocessed video stream (the image data); a pre-trained deep learning model is used, such as a real-time object detection algorithm (YOLO), a fast region-based convolutional neural network (Faster The invention relates to a method for detecting people in a video stream (the image data) preprocessed by R-CNN, etc., to identify the human body in the video stream, and to give a bounding box of the human body to obtain edge data; for continuous video frames, using tracking algorithms such as KCF (kernel correlation filter algorithm) and SORT (sorting algorithm) to track the detected human body to ensure the continuity of the human body in the video, and to obtain a series of edge data; constructing an arbitrary bounding box according to the edge data, and extracting the outline of the person in the detected human body bounding box using image segmentation and / or edge detection technology to obtain an initial image; filling the inside of the extracted outline (initial image) with a specific color (such as black, gray, etc.) according to a first preset filling rule to form a silhouette effect, and to obtain a silhouette image; at the same time, the edge of the silhouette image is consistent with the key edge information of the real person image (such as the outline of the head and limbs) for subsequent action and posture recognition.

[0071] For example, in order to more clearly display edge information of the silhouette, the edge of the silhouette may be enhanced, such as by using a Sobel operator, a Prewitt operator, or the like to perform edge sharpening.

[0072] In a possible implementation manner, before the step of identifying the actual action data of a real person in the image data according to the image data by using a preset recognition model, the step further includes:

[0073] Acquire a first preset data set, the first preset data set including first preset image data of a first preset group and first annotation information of the first preset image data, the first annotation information being used to characterize first action data of a real person in the first preset image data;

[0074] Acquire a second preset data set, the second preset data set including second preset image data of a second preset group and second annotation information of the second preset image data, the second annotation information being used to characterize second action data of a real person in the second preset image data;

[0075] A training set and a verification set are constructed according to the first preset data set and the second preset data set, and the convolutional neural network model is trained by the training set, and the trained convolutional neural network model is verified by the verification set to obtain the preset recognition model.

[0076] For example, monitoring technology is widely used in the field of intelligent monitoring. As a type of protected persons, disabled persons with physical disabilities often lack targeted attention due to their physical disabilities. That is, during the training process, existing recognition models are trained using real images of people with intact limbs as training sets. However, when people with intact limbs and disabled persons with physical disabilities do the same action, the body parameters of the healthy and disabled persons are often different. Therefore, when the recognition model trained using real images of people with intact limbs as training sets faces disabled persons with physical disabilities as protected persons, the recognition accuracy is greatly reduced, causing inconvenience to the monitoring process for disabled persons with physical disabilities.

[0077] Exemplarily, real-life images of physically disabled people are collected as first preset image data, and real-life images of people with healthy limbs are collected as second preset image data. The recognition model trained with the first preset image data and the second preset image data can accurately identify the body parameters of physically disabled people when facing physically disabled people as the guardians, thereby ensuring the applicability of this solution.

[0078] In a possible implementation manner, the step of updating the first image according to a preset rule based on the matching result to obtain an image to be displayed includes:

[0079] When the first image is a match figure image corresponding to a real person in the image data, the match figure image is updated according to the deformation rule in the preset rule and the matching result to obtain a first display image in the images to be displayed;

[0080] When the first image is a silhouette image corresponding to a real person in the image data, the silhouette image is updated according to the matching result according to a second preset filling rule in the preset rules to obtain a second display image in the image to be displayed;

[0081] When the first image is a mixed image composed of the stickman image and the silhouette image, the first display image and the second display image are spliced ​​to obtain a third display image in the image to be displayed.

[0082] Exemplarily, the step of updating the match figure image according to the deformation rule in the preset rule and the matching result to obtain the first display image in the image to be displayed includes:

[0083] Determine a target key point and a target connecting line between the target key points in the key point data according to the matching result;

[0084] According to the deformation rules in the preset rules, the target key points and the target connecting lines are deformed to obtain deformation key points and deformation connecting lines;

[0085] The stick figure image is updated based on the deformation key points and the deformation connection lines to obtain a first display image among the images to be displayed.

[0086] Exemplarily, the target key points and the target connecting lines are deformed according to the deformation rules in the preset rules, and the deformation processing includes but is not limited to bolding the target key points and the target connecting lines, performing graphic replacement processing on the target key points and the target connecting lines, performing color change processing on the target key points and the target connecting lines, etc.

[0087] Exemplarily, the silhouette image is updated according to the matching result according to the second preset filling rule among the preset rules, that is, the area where the target key points and the target connecting lines are located is filled with different colors, for example, the area where the first target point and the first target connecting line are located is filled with red, and the area where the second target point and the second target connecting line are located is filled with yellow, wherein the filled color is determined according to the matching result.

[0088] Exemplarily, all key points in the key point data whose matching results are failed are used as the target key points.

[0089] In a possible implementation, the step of uploading the image to be displayed to a server corresponding to the edge computing device includes:

[0090] Acquire basic data of the target object, wherein the basic data includes identity information of the target object and health monitoring data acquired according to the identity information;

[0091] Marking the image to be displayed according to the basic data to obtain a marked image to be displayed;

[0092] The step of marking the image to be displayed according to the basic data to obtain the marked image to be displayed includes:

[0093] Marking the basic data into the image to be displayed according to a preset marking rule to obtain a marked image to be displayed, wherein the preset marking rule at least includes a preset marking position of the basic data in the image to be displayed, a preset display method of the basic data, and a preset data form of the basic data in the image to be displayed;

[0094] The marked image to be displayed is uploaded to the server corresponding to the edge computing device.

[0095] For example, in traditional video surveillance, by observing the facial features and clothing of individuals, people can intuitively identify the specific person in the picture and his or her professional identity. For example, the patient's name can be known by his or her facial features, and he or she can be judged as a patient from the patient clothing he or she wears; similarly, the doctor's name can be recognized by his or her facial features, and his or her identity as a doctor can be confirmed based on the professional clothing he or she wears. However, after implementing privacy protection measures, such as using stickman skeleton deformation processing, it becomes difficult to directly identify the target object. Therefore, by marking the basic data (the identity information of the target object and the health monitoring data obtained based on the identity information) into the image to be displayed according to the preset marking rules, the marked image to be displayed can achieve effective identification of the target object and visual display of key information.

[0096] Specifically, when obtaining the identity information of the target object and the health monitoring data obtained based on the identity information, image recognition technology can be used, that is, face recognition is performed on the person in the video screen to determine the identity information of the person in the video screen (such as name, identity, etc.), and then the health monitoring data corresponding to the identity information (such as heart beat, heart rate and blood pressure) is screened out in the health monitoring database.

[0097] Exemplarily, through preset marking rules, at the preset marking position, the basic data (identity information of the target object and health monitoring data obtained based on the identity information) in a preset data form (text form, picture form, video form, etc.) is marked in the image to be displayed in a preset display mode (selective display mode, hidden display mode, interactive display mode, etc.). Specifically, the basic data displayed in an interactive display mode is displayed only when the user clicks on a target object with the mouse (such as hovering the mouse, clicking, long pressing or other gestures). In the process of visual management, unauthorized users can only see the stickman image, while users or administrators who are authorized to see specific information can see the basic data authorized to view when clicking. It ensures that efficient and accurate on-site management effects can be achieved while ensuring privacy and security.

[0098] Exemplarily, when displaying, the server side corresponding to the edge computing device can integrate the marked images to be displayed with more relevant monitoring data into a unified monitoring interface, which is convenient for users to conduct real-time monitoring and data analysis, thereby greatly improving the efficiency and intuitiveness of monitoring management, while reducing storage and transmission costs by reducing the amount of original data, and enhancing privacy protection features.

[0099] Exemplarily, the process of marking the image to be displayed can be confirmed by using auxiliary text or logo, the text information of the stick figure skeleton (target key point and target connection line) marked as abnormal behavior changes, and the auxiliary icon arrow or other directional icons are used for guiding marking. The marking method can be to change the video display area method, focus the screen on the stick figure skeleton screen area marked as abnormal behavior, including but not limited to zooming in or out the monitored video screen, marking with lines or graphics, etc.

[0100] Exemplary, the image to be displayed can be a stick figure skeleton that shields all normal behaviors, and only includes the image of the stick figure skeleton (target key point and target connecting line) of abnormal behavior. The image to be displayed can be a stick figure skeleton that is highlighted as abnormal behavior. Relative to the stick figure skeleton of normal behavior, the stick figure skeleton that is highlighted is more easily identified. This method of highlighting includes but is not limited to improving the brightness of the stick figure skeleton marked as abnormal behavior, showing more differentiated colors, increasing or decreasing the volume or area, using the outer contour light or stroke and other modes. The image to be displayed can be a stick figure skeleton that is highlighted as abnormal behavior, such as increasing the transparency of the stick figure skeleton of normal behavior, dimming the brightness, blurring, deforming filter processing, etc. The image to be displayed can deform the stick figure skeleton marked as abnormal behavior, including but not limited to replacing with virtual characters such as: patients, dangerous elements and other virtual characters.

[0101] In a possible implementation, Figure 2 As shown, the present application provides a motion recognition device based on a privacy algorithm, the device comprising:

[0102] A data acquisition module 201 is used to obtain image data of a target object through an edge computing device, and convert the image data into a first image through the edge computing device, wherein the first image includes at least one of a matchstick image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the matchstick image and the silhouette image;

[0103] The action recognition module 202 is used to recognize actual action data of a real person in the image data according to the image data by using a preset recognition model, and match the actual action data with preset action data to obtain a matching result;

[0104] The data display module 203 is used to update the first image according to the matching result according to a preset rule to obtain an image to be displayed, and upload the image to be displayed to a server corresponding to the edge computing device.

[0105] In a possible implementation, the data acquisition module 201 is used to:

[0106] extracting a real-person image of the target object from the image data;

[0107] Extracting key points of a human body from the real person image based on the privacy algorithm to obtain key point data of the real person image;

[0108] A matchstick figure image in the first image corresponding to the real person in the image data is constructed according to the key point data.

[0109] In a possible implementation, the data acquisition module 201 is used to:

[0110] extracting a real-person image of the target object from the image data;

[0111] Acquire edge data of the real person image based on edge detection technology, and perform image segmentation on the real person image according to the edge data to obtain an initial image;

[0112] The initial image is filled according to a first preset filling rule to obtain a silhouette image in the first image corresponding to the real person in the image data.

[0113] In a possible implementation, the data acquisition module 201 is used to:

[0114] Acquire a first preset data set, the first preset data set including first preset image data of a first preset group and first annotation information of the first preset image data, the first annotation information being used to characterize first action data of a real person in the first preset image data;

[0115] Acquire a second preset data set, the second preset data set including second preset image data of a second preset group and second annotation information of the second preset image data, the second annotation information being used to characterize second action data of a real person in the second preset image data;

[0116] A training set and a verification set are constructed according to the first preset data set and the second preset data set, and the convolutional neural network model is trained by the training set, and the trained convolutional neural network model is verified by the verification set to obtain the preset recognition model.

[0117] In a possible implementation, the action recognition module 202 is used to:

[0118] When the first image is a match figure image corresponding to a real person in the image data, the match figure image is updated according to the deformation rule in the preset rule and the matching result to obtain a first display image in the images to be displayed;

[0119] When the first image is a silhouette image corresponding to a real person in the image data, the silhouette image is updated according to the matching result according to a second preset filling rule in the preset rules to obtain a second display image in the image to be displayed;

[0120] When the first image is a mixed image composed of the stickman image and the silhouette image, the first display image and the second display image are spliced ​​to obtain a third display image in the image to be displayed.

[0121] In a possible implementation, the data display module 203 is used to:

[0122] Determine a target key point and a target connecting line between the target key points in the key point data according to the matching result;

[0123] According to the deformation rules in the preset rules, the target key points and the target connecting lines are deformed to obtain deformation key points and deformation connecting lines;

[0124] The stick figure image is updated based on the deformation key points and the deformation connection lines to obtain a first display image among the images to be displayed.

[0125] In a possible implementation, the data display module 203 is used to:

[0126] Acquire basic data of the target object, wherein the basic data includes identity information of the target object and health monitoring data acquired according to the identity information;

[0127] Marking the image to be displayed according to the basic data to obtain a marked image to be displayed;

[0128] The step of marking the image to be displayed according to the basic data to obtain the marked image to be displayed includes:

[0129] Marking the basic data into the image to be displayed according to a preset marking rule to obtain a marked image to be displayed, wherein the preset marking rule at least includes a preset marking position of the basic data in the image to be displayed, a preset display method of the basic data, and a preset data form of the basic data in the image to be displayed;

[0130] The marked image to be displayed is uploaded to the server corresponding to the edge computing device.

[0131] The image data of the target object is obtained through an edge computing device, and the image data is converted into a first image through the edge computing device, wherein the first image includes at least one of a stick figure image corresponding to the real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the stick figure image and the silhouette image; the video data is processed through the edge computing device, the face information in the video is identified and matched with the system personnel database, the corresponding personnel information is retrieved, and the name, identity, occupation and other details are marked around it according to the custom settings, and the individual's physiological indicators (such as heartbeat, heart rate and blood pressure) are displayed at the same time, so as to achieve efficient and accurate on-site management under the premise of ensuring privacy and security; avoiding the transmission of video data with real-person images can reduce the risk of video data leakage. risk; through a preset recognition model, the actual action data of a real person in the image data is identified according to the image data, and the actual action data is matched with the preset action data to obtain a matching result; according to the matching result, the first image is updated according to a preset rule to obtain an image to be displayed, and the image to be displayed is uploaded to a server corresponding to the edge computing device; by updating the first image according to the preset rule, the part where the actual action data does not match the preset action data can be highlighted, that is, after the video data is privacy processed, the information contained in the video data is highlighted, which is convenient for data extraction according to the video data after privacy processing, and reduces the impact of video data privacy on information, so that the video data meets the real-time requirements of video surveillance technology after privacy.

[0132] In a possible implementation, Figure 3 As shown, an embodiment of the present application provides an electronic device 300, including: a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, it implements the following steps: obtaining image data of a target object through an edge computing device, and converting the image data into a first image through the edge computing device, wherein the first image includes at least one of a matchstick image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the matchstick image and the silhouette image; identifying actual action data of a real person in the image data according to the image data through a preset recognition model, and matching the actual action data with the preset action data to obtain a matching result; updating the first image according to a preset rule according to the matching result to obtain an image to be displayed, and uploading the image to be displayed to a server corresponding to the edge computing device.

[0133] The image data of the target object is obtained through an edge computing device, and the image data is converted into a first image through the edge computing device, wherein the first image includes at least one of a stick figure image corresponding to a real person in the image data, a silhouette image corresponding to a real person in the image data, and a mixed image composed of the stick figure image and the silhouette image; the video data is processed by the edge computing device to avoid the transmission of video data with real person images, which can reduce the risk of video data leakage; the actual action data of the real person in the image data is identified according to the image data through a preset recognition model, and the actual action data is matched with the preset action data to obtain a matching result; the first image is updated according to the matching result according to a preset rule to obtain an image to be displayed, and the image to be displayed is uploaded to a server corresponding to the edge computing device; the first image is updated according to the preset rule, and the part that does not match the actual action data with the preset action data can be highlighted, that is, after the video data is privacy processed, the information contained in the video data is highlighted, which is convenient for data extraction according to the video data after privacy processing, and reduces the impact of the privacy of the video data on the information, so that the video data after privacy also meets the real-time requirements of video surveillance technology.

[0134] In a possible implementation, Figure 4 As shown, an embodiment of the present application provides a computer-readable storage medium 400, on which a computer program 411 is stored, and when the computer program 411 is executed by a processor, it implements: obtaining image data of a target object through an edge computing device, and converting the image data into a first image through the edge computing device, the first image including at least one of a matchstick image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the matchstick image and the silhouette image; identifying actual action data of the real person in the image data according to the image data through a preset recognition model, and matching the actual action data with the preset action data to obtain a matching result; updating the first image according to a preset rule according to the matching result to obtain an image to be displayed, and uploading the image to be displayed to a server corresponding to the edge computing device.

[0135] The image data of the target object is obtained through an edge computing device, and the image data is converted into a first image through the edge computing device, wherein the first image includes at least one of a stick figure image corresponding to a real person in the image data, a silhouette image corresponding to a real person in the image data, and a mixed image composed of the stick figure image and the silhouette image; the video data is processed by the edge computing device to avoid the transmission of video data with real person images, which can reduce the risk of video data leakage; the actual action data of the real person in the image data is identified according to the image data through a preset recognition model, and the actual action data is matched with the preset action data to obtain a matching result; the first image is updated according to the matching result according to a preset rule to obtain an image to be displayed, and the image to be displayed is uploaded to a server corresponding to the edge computing device; the first image is updated according to the preset rule, and the part that does not match the actual action data with the preset action data can be highlighted, that is, after the video data is privacy processed, the information contained in the video data is highlighted, which is convenient for data extraction according to the video data after privacy processing, and reduces the impact of the privacy of the video data on the information, so that the video data after privacy also meets the real-time requirements of video surveillance technology.

[0136] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0137] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0138] The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, etc., or any suitable combination of the foregoing.

[0139] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0140] It should be understood by those skilled in the art that the modules or steps of the present invention described above can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, optionally, they can be implemented by a program code executable by a computer device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0141] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described above, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

[0142] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for action recognition based on a privacy algorithm, characterized in that: include: Acquire image data of a target object through an edge computing device, and convert the image data into a first image through the edge computing device, wherein the first image includes at least one of a stick figure image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the stick figure image and the silhouette image; Using a preset recognition model, identifying actual action data of a real person in the image data according to the image data, and matching the actual action data with the preset action data to obtain a matching result; According to the matching result, the first image is updated according to a preset rule to obtain an image to be displayed, and the image to be displayed is uploaded to a server corresponding to the edge computing device.

2. The action recognition method based on the privacy algorithm as claimed in claim 1, characterized in that: The step of converting the image data into a first image by the edge computing device comprises: extracting a real-person image of the target object from the image data; Extracting key points of a human body from the real person image based on the privacy algorithm to obtain key point data of the real person image; A matchstick figure image in the first image corresponding to the real person in the image data is constructed according to the key point data.

3. The action recognition method based on privacy algorithm as claimed in claim 1, characterized in that: The step of converting the image data into a first image by the edge computing device further includes: extracting a real-person image of the target object from the image data; Acquire edge data of the real person image based on edge detection technology, and perform image segmentation on the real person image according to the edge data to obtain an initial image; The initial image is filled according to a first preset filling rule to obtain a silhouette image in the first image corresponding to the real person in the image data.

4. The action recognition method based on the privacy algorithm as claimed in claim 1, characterized in that: Before the step of identifying the actual action data of a real person in the image data according to the image data by using a preset recognition model, the method further includes: Acquire a first preset data set, the first preset data set including first preset image data of a first preset group and first annotation information of the first preset image data, the first annotation information being used to characterize first action data of a real person in the first preset image data; Acquire a second preset data set, the second preset data set including second preset image data of a second preset group and second annotation information of the second preset image data, the second annotation information being used to characterize second action data of a real person in the second preset image data; A training set and a verification set are constructed according to the first preset data set and the second preset data set, and the convolutional neural network model is trained by the training set, and the trained convolutional neural network model is verified by the verification set to obtain the preset recognition model.

5. The action recognition method based on privacy algorithm as claimed in claim 1, characterized in that: The step of updating the first image according to a preset rule based on the matching result to obtain an image to be displayed includes: When the first image is a match figure image corresponding to a real person in the image data, the match figure image is updated according to the deformation rule in the preset rule and the matching result to obtain a first display image in the images to be displayed; When the first image is a silhouette image corresponding to a real person in the image data, the silhouette image is updated according to the matching result according to a second preset filling rule in the preset rules to obtain a second display image in the image to be displayed; When the first image is a mixed image composed of the stickman image and the silhouette image, the first display image and the second display image are spliced ​​to obtain a third display image in the image to be displayed.

6. The action recognition method based on the privacy algorithm as claimed in claim 2, characterized in that: The step of updating the match man image according to the deformation rule in the preset rule and the matching result to obtain the first display image in the image to be displayed includes: Determine a target key point and a target connecting line between the target key points in the key point data according to the matching result; According to the deformation rules in the preset rules, the target key points and the target connecting lines are deformed to obtain deformation key points and deformation connecting lines; The stick figure image is updated based on the deformation key points and the deformation connection lines to obtain a first display image among the images to be displayed.

7. The action recognition method based on privacy algorithm as claimed in claim 1, characterized in that: The step of uploading the image to be displayed to a server corresponding to the edge computing device further includes: Acquire basic data of the target object, wherein the basic data includes identity information of the target object and health monitoring data acquired according to the identity information; Marking the image to be displayed according to the basic data to obtain a marked image to be displayed; The step of marking the image to be displayed according to the basic data to obtain the marked image to be displayed includes: Marking the basic data into the image to be displayed according to a preset marking rule to obtain a marked image to be displayed, wherein the preset marking rule at least includes a preset marking position of the basic data in the image to be displayed, a preset display method of the basic data, and a preset data form of the basic data in the image to be displayed; The marked image to be displayed is uploaded to the server corresponding to the edge computing device.

8. A motion recognition device based on a privacy algorithm, characterized in that: The device comprises: a data acquisition module, configured to acquire image data of a target object through an edge computing device, and convert the image data into a first image through the edge computing device, wherein the first image includes at least one of a matchstick image corresponding to a real person in the image data, a silhouette image corresponding to the real person in the image data, and a mixed image composed of the matchstick image and the silhouette image; An action recognition module, used to recognize actual action data of a real person in the image data according to the image data by using a preset recognition model, and match the actual action data with preset action data to obtain a matching result; A data display module is used to update the first image according to a preset rule based on the matching result to obtain an image to be displayed, and upload the image to be displayed to a server corresponding to the edge computing device.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the action recognition method based on the privacy algorithm as described in any one of claims 1 to 7 are performed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the action recognition method based on a privacy algorithm as described in any one of claims 1 to 7 are executed.

Images