Intelligent data processing method and system

Abstract

The present invention provides a data intelligent processing method and system, which relates to data processing technology, collects a customized background in a target scene, annotates the customized background with objects, generates an interactive image, and sends it to an acquisition end; determines the target object based on the selection information of the interactive image by the acquisition end; calls an initial posture model to associate with the target object, receives posture setting information of the initial posture model by the acquisition end, and generates a posture customization model; analyzes the relative posture relationship between the posture customization model and the target object to generate a customized posture relationship; obtains a real-time image based on a real-time signal from the acquisition end, and collects required data when the real-time posture relationship in the real-time image corresponds to the customized posture relationship, so as to facilitate the customization of image acquisition posture according to user needs, realize automatic acquisition of image data, and improve the convenience of user image acquisition.

Description

Technical Field

[0001] The present invention relates to data processing technology, and in particular to a data intelligent processing method and system. Background Art

[0002] In today's digital image acquisition environment, people's demand for personalized and precise image acquisition is growing. Ordinary users hope to easily obtain images that meet their creativity and expectations when taking selfies in daily life scenes.

[0003] Currently, traditional image acquisition methods have many limitations. In most cases, users can only seek help from others or manually set the timing, and lack the ability to pre-customize the shooting scene and their own posture. For example, when shooting landscape portraits, it is difficult for users to accurately know whether their position and posture in the picture can perfectly blend with the background. It often takes multiple attempts to shoot, wasting a lot of time and energy.

[0004] Therefore, how to customize the image acquisition posture according to user needs, realize automatic acquisition of image data, and improve the convenience of user image acquisition has become an urgent problem to be solved. Summary of the Invention

[0005] The present invention provides an intelligent data processing method and system, which can customize image acquisition posture according to user needs, realize automatic acquisition of image data, and improve the convenience of user image acquisition.

[0006] A first aspect of the present invention provides a method for intelligent data processing, comprising:

[0007] The customized background in the target scene is collected, and the objects in the customized background are annotated to generate an interactive image and sent to the collection end.

[0008] The target object is determined based on the selection information of the interactive image by the acquisition end.

[0009] The initial posture model is called to associate with the target object, the posture setting information of the initial posture model from the acquisition end is received, and a posture customization model is generated.

[0010] The relative posture relationship between the posture customization model and the target object is analyzed to generate a customized posture relationship.

[0011] A real-time image is acquired based on the real-time signal of the acquisition end, and when the real-time posture relationship in the real-time image corresponds to the customized posture relationship, required data is acquired.

[0012] Optionally, in a possible implementation of the first aspect, the initial posture model is a three-dimensional human virtual model, and the calling of the initial posture model to associate with the target object, receiving posture setting information of the initial posture model from the acquisition end, and generating the posture customization model include:

[0013] Calling a 3D human body virtual model from a preset model library and associating it with the target object to display it on the interactive image interface of the acquisition end;

[0014] Generating posture setting information according to the limb joint point setting of the three-dimensional human body virtual model by the acquisition end;

[0015] The posture of the three-dimensional human body virtual model is updated based on the posture setting information to generate a posture customized model.

[0016] Optionally, in a possible implementation of the first aspect, the method further includes:

[0017] Extracting a maximum bounding rectangle of the target object in the interactive image, and receiving an input direction selected by the acquisition end based on the maximum bounding rectangle and a physical size corresponding to the input direction;

[0018] Obtain the pixel size of the maximum circumscribed rectangle corresponding to the input direction in the interaction image;

[0019] The three-dimensional human body virtual model is proportionally adapted according to the ratio of the physical size and the pixel size in the same input direction.

[0020] Optionally, in a possible implementation of the first aspect, the analyzing the relative posture relationship between the customized posture model and the target object to generate a customized posture relationship includes:

[0021] Extracting limb joints of the posture customization model as feature points of the posture customization model through a human skeleton key point detection algorithm, wherein the limb joints include at least coordinate information of the hip joint, knee joint, and elbow joint in the interactive image;

[0022] An edge detection algorithm is used to extract the edge points of the target object as the feature points of the target object;

[0023] The relative position relationship between the characteristic points of the posture customized model and the characteristic points of the target object is determined, wherein the relative position relationship includes a relative angle, and a customized posture relationship is generated according to the relative position relationship.

[0024] Optionally, in a possible implementation of the first aspect, the method further includes:

[0025] Deleting the customized posture relationship;

[0026] receiving association information of the corresponding posture customized model feature points and the target object feature points from the acquisition end, and determining the corresponding posture customized model feature points and the target object feature points as an association group according to the association information;

[0027] The relative positional relationship between the characteristic points of the posture customization model and the characteristic points of the target object in each of the association groups is determined, and a customized posture relationship is generated according to the relative positional relationship.

[0028] Optionally, in a possible implementation of the first aspect, the method further includes:

[0029] When there is a spatial overlap between the posture customization model and the target object in the interactive image, receiving a click instruction from the user on the overlapped area of ​​the target object;

[0030] Rendering the limb part corresponding to the click instruction in the posture customization model as a hidden state, so that the overlapping area of ​​the posture customization model is displayed behind the target object, and marking the posture customization model feature points corresponding to the corresponding limb part as hidden feature points;

[0031] A visual constraint condition is generated according to the hidden feature point, and the customized posture relationship is updated based on the visual constraint condition.

[0032] Optionally, in a possible implementation of the first aspect, when the real-time posture relationship in the real-time image corresponds to the customized posture relationship, collecting the required data includes:

[0033] Analyzing the customized posture relationship to obtain posture customization model feature points and target object feature points;

[0034] The limb key points corresponding to the posture customized model feature points in the real-time image are used as real-time posture feature points through the human skeleton key point detection algorithm;

[0035] An edge detection algorithm is used to extract target feature points in the target object corresponding to the target object feature points as real-time target feature points;

[0036] A real-time posture relationship is determined according to the real-time posture feature points and the real-time target feature points, and when a matching degree between the real-time posture relationship and the customized posture relationship is greater than a preset matching degree, acquisition signal acquisition requirement data is generated.

[0037] Optionally, in a possible implementation of the first aspect, calculating the degree of matching between the real-time posture relationship and the customized posture relationship includes:

[0038] Obtaining angle differences corresponding to multiple groups of feature points in the real-time posture relationship and the customized posture relationship, marking feature points whose angle differences are within a preset range as matching feature points, and counting the number of the matching feature points as a first matching number;

[0039] Determining the number of feature point groups that need to be compared in the customized posture relationship to obtain a first total number, and obtaining a first matching degree based on a ratio of the first matching number to the first total number;

[0040] When the hidden feature points exist, counting a second total number of the hidden feature points and a second matching number of the hidden feature points that meet the hidden state;

[0041] A second matching degree is obtained according to a ratio of the second matching number to the second total number, and the matching degree is obtained based on the first matching degree and the second matching degree.

[0042] Optionally, in a possible implementation of the first aspect, the method further includes:

[0043] Load at least two initial pose models in the interaction image and associate them with different target objects respectively;

[0044] receiving the user's independent posture setting for each initial posture model and generating multiple posture customized models;

[0045] Analyze the relative posture relationship between all posture customization models and the corresponding target objects, and generate joint trigger conditions;

[0046] When the real-time posture relationship of all users detected in the real-time image matches the joint trigger condition, data collection is performed.

[0047] A second aspect of the present invention provides an intelligent data processing system, comprising:

[0048] A sending module is used to collect the customized background in the target scene, mark the objects in the customized background, generate an interactive image and send it to the collection end;

[0049] A determination module, configured to determine a target object based on the selection information of the interactive image by the acquisition end;

[0050] A receiving module, configured to call an initial posture model to associate with the target object, receive posture setting information of the initial posture model from the acquisition end, and generate a posture customization model;

[0051] A generation module, configured to analyze the relative posture relationship between the posture customization model and the target object to generate a customized posture relationship;

[0052] The acquisition module is used to obtain a real-time image based on the real-time signal of the acquisition end, and to acquire required data when the real-time posture relationship in the real-time image corresponds to the customized posture relationship.

[0053] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory, a processor, and a computer program, wherein the computer program is stored in the memory, and the processor runs the computer program to execute the first aspect of the present invention and various methods that may be involved in the first aspect.

[0054] According to a fourth aspect of the present invention, a readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, it is used to implement the first aspect of the present invention and various methods that may be involved in the first aspect.

[0055] The beneficial effects of the present invention are as follows:

[0056] 1. The present invention can customize the image acquisition posture according to user needs, realize automatic acquisition of image data, and improve the convenience of user image acquisition. First, the present invention can collect the customized background under the target scene, and mark the objects in the customized background to obtain an interactive image and send it to the acquisition end. The user can select the target object at the acquisition end, and then call the initial posture model and associate it with the target object. According to the user's posture setting information for the initial posture model, a posture customization model is generated, and the corresponding customized posture relationship is analyzed and compared with the real-time posture relationship in the real-time image acquired in real time, thereby realizing the automatic acquisition of required data and saving the user's shooting time.

[0057] 2. The present invention can personalize the posture and perform posture matching, thereby improving the accuracy and efficiency of image acquisition, so as to quickly obtain the captured image of the corresponding posture. First, the present invention can set a unique posture by adjusting the joint points of the model's limbs. The personalized customization method greatly satisfies the user's creative needs for shooting, gets rid of the limitations of traditional shooting methods, and allows users to construct shooting scenes and postures according to their own ideas. Secondly, the present invention can use the human skeleton key point detection algorithm and the edge detection algorithm to respectively extract the real-time posture feature points and real-time target feature points in the real-time image, and then determine the real-time posture relationship. When the matching degree between the real-time posture relationship and the customized posture relationship is greater than the preset matching degree, the acquisition signal acquisition requirement data is automatically generated. Thus, through the high-precision posture matching and automatic acquisition mechanism, the user avoids shooting errors caused by improper manual operation timing, ensures that the collected image fully meets the posture requirements pre-set by the user, greatly improves the accuracy and efficiency of image acquisition, saves the user's shooting time, and allows the user to easily obtain high-quality image data.

[0058] 3. The present invention can achieve multi-model collaboration and occlusion processing effects to improve the accuracy of user-customized acquisition postures. Among them, the present invention supports loading multiple initial posture models in the interactive image and associating different target objects respectively. The user sets the posture of each model independently to generate multiple posture customization models, analyzes the relative posture relationship between all posture customization models and the corresponding target objects to generate a joint trigger condition, and executes data collection when the real-time posture relationship of all users in the real-time image matches the joint trigger condition. At the same time, for the problem of spatial overlapping areas that may occur between the posture customization model and the target object, the present invention can render the corresponding limb parts in the posture customization model as a hidden state according to the user's click instruction on the overlapping area, and update the customized posture relationship. The present invention can comprehensively judge whether the posture of all people meets the joint trigger condition, realize collaborative shooting, and ensure the accuracy of image display effect and posture relationship when dealing with occlusion problems, thereby improving the feasibility and quality of image acquisition in complex scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 The present invention provides a flow chart of a method for intelligent data processing;

[0060] Figure 2 A schematic diagram of an overlapping area is provided for the present invention;

[0061] Figure 3 The present invention provides a structural diagram of a data intelligent processing system;

[0062] Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0063] The following specific embodiments are used to describe the technical solution of the present invention in detail. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0064] like Figure 1 As shown, the present invention provides a flow chart of a data intelligent processing method, which includes:

[0065] S1, collecting a customized background in a target scene, marking objects in the customized background, generating an interactive image, and sending it to a collection end.

[0066] It can be understood that a customized background is collected for the target scene, such as a park scene. The customized background is the environment in which the person will be photographed later. After the collection is completed, the objects in the background, such as benches, trees, grass, etc., are marked, and the specific information of these objects is marked. Through this marking, the background image has more information, forming an interactive image, and finally this interactive image is sent to the collection end. The collection end can be a device used by the user, such as a mobile phone, camera, etc., to provide basic image data for subsequent users to perform operations on the collection end.

[0067] Among them, the target scene is the regional scene for image capture and collection, for example, it can be a park, the customized background is the background of the person image customized and selected by the personnel when collecting images, for example, it can be a background with benches, trees, grass and lakes selected in the park, and the object annotation is to annotate the objects in the customized background with information so that subsequent personnel can select the corresponding objects for interaction, thereby realizing the creation of customized actions for image collection, so that the collection end can automatically match the actions for image collection, the interactive image is an image with annotated information, and the collection end is an information terminal for collecting images, which can be a mobile phone, camera, etc.

[0068] S2: Determine the target object based on the selection information of the interactive image by the acquisition end.

[0069] It is understandable that after the interactive image is sent to the acquisition end, the user can operate the interactive image at the acquisition end, that is, the user can select an object in the interactive image according to his or her own photography needs. For example, if the user wants to take a photo on a bench, he or she can select the bench in the interactive image. The acquisition end will record the user's selection information and then determine the target object based on the selection information, that is, the object that the user wants to interact with when taking a photo. By determining the target object, the position of the interactive object is clarified for subsequent posture setting and data collection operations.

[0070] The selection information is the operation information for triggering the selection of an object in the interactive image, and the target object is the interactive object selected by the person, for example, a bench, a tree, etc.

[0071] S3, calling the initial posture model to associate with the target object, receiving the posture setting information of the initial posture model from the acquisition end, and generating a posture customization model.

[0072] It should be noted that in order to facilitate the subsequent acquisition end to automatically identify and match the person's movements and postures, a corresponding posture customization model can be set according to user needs, so that when the subsequent acquisition end determines that the user's posture is similar to the set posture, automatic image capture and collection can be achieved, avoiding the user's unsightly movements due to time deviation, thereby affecting the image capture efficiency.

[0073] It can be understood that, on the basis of determining the target object, a suitable posture model is further set for taking pictures. First, the initial posture model, that is, the virtual image model corresponding to the character, is called and associated with the target object, so that the model and the target object are logically and visually connected. Then, the posture setting information of the initial posture model is received from the acquisition end. This information is the posture set by the user according to his or her own photography creativity or needs. Finally, a posture customization model is generated based on these posture setting information. This model is customized according to the user's specific needs and is used for subsequent recognition of photographic postures for automatic collection.

[0074] Among them, the initial posture model is a virtual model corresponding to the initial preset character posture, the posture setting information is the information of setting the posture, such as the posture information of raising the hand to make a "V" sign, and the posture customization model is a customized posture model.

[0075] In some embodiments, the specific implementation steps of step S3 (the initial posture model is a three-dimensional human body virtual model, the initial posture model is called to associate with the target object, the posture setting information of the initial posture model is received from the acquisition end, and the posture customization model is generated) include:

[0076] S31, calling a three-dimensional human body virtual model from a preset model library, and associating it with a target object and displaying it on an interactive image interface of the acquisition end.

[0077] It can be understood that a suitable three-dimensional human virtual model is selected from a pre-set model library. The model library stores multiple three-dimensional human virtual models of different types and postures for selection. After being called out, this three-dimensional human virtual model is associated with the previously determined target object. This association can be a correspondence in position, an association in action, etc. Finally, the associated three-dimensional human virtual model and the target object are displayed on the interactive image interface of the acquisition end, so that the user can intuitively see the relationship between the two, which is convenient for the user to perform subsequent operations.

[0078] The preset model library is a pre-set model database, and the three-dimensional human body virtual model is a three-dimensional virtual model corresponding to the human body.

[0079] S32, generating posture setting information according to the setting of the limb joint points of the three-dimensional human body virtual model by the acquisition end.

[0080] It is understandable that when the three-dimensional human virtual model is displayed on the interactive image interface of the acquisition end, the user can set the limb joints of the model, such as adjusting the position of the arms, the degree of bending of the legs, etc. The acquisition end will record the user's setting operations on these limb joints and generate posture setting information based on these operations. The posture setting information describes in detail the posture that the user wants the three-dimensional human virtual model to present, which is the key basis for the subsequent generation of posture customization models.

[0081] Among them, the limb joints are the joints corresponding to the corresponding limbs of the human body, such as ankle joints, knee joints, elbow joints, etc.

[0082] S33: updating the posture of the three-dimensional human body virtual model based on the posture setting information to generate a posture customized model.

[0083] It can be understood that the posture of the three-dimensional human virtual model is updated according to the generated posture setting information, and the limb joints of the model are adjusted according to the posture setting information so that the three-dimensional human virtual model presents the posture expected by the user. The updated three-dimensional human virtual model becomes a posture customization model, which is completely customized according to the specific needs of the user. Subsequently, this model can be used to determine the actual posture of the person when taking pictures, and perform related data collection operations.

[0084] In some embodiments, it further includes:

[0085] A1, extracting the maximum bounding rectangle of the target object in the interactive image, and receiving the input direction selected by the acquisition end based on the maximum bounding rectangle and the physical size corresponding to the input direction.

[0086] It can be understood that in order to obtain the relevant size information of the target object and the direction information specified by the user, so as to subsequently perform proportional adaptation on the three-dimensional human body virtual model, the maximum enclosing rectangle of the target object in the interactive image can be extracted through image processing technology. The rectangle can completely contain the target object and reflect the approximate outline and range of the target object to a certain extent. Then, the acquisition end receives the input direction selected by the user based on this maximum enclosing rectangle. For example, the user may select the horizontal direction, vertical direction or other specific direction of the target object. At the same time, the acquisition end also receives the physical size corresponding to the input direction, that is, the length of the target object along the input direction in the actual physical space and other size information. The information obtained facilitates subsequent proportional calculations, thereby adjusting the model proportionally and improving the model adaptability.

[0087] Among them, the maximum enclosing rectangle is the rectangle that surrounds the target object, the input direction is the directional boundary of the input size information, and the physical size is the size corresponding to the input direction, for example, it can be the maximum enclosing rectangle corresponding to the long side distance or the maximum enclosing rectangle corresponding to the wide side distance.

[0088] A2: Obtain the pixel size of the maximum circumscribed rectangle corresponding to the input direction in the interaction image.

[0089] It is understandable that after determining the input direction and the physical size of the target object, it is necessary to further obtain the pixel size in the corresponding input direction in the interactive image, that is, the interactive image can be analyzed to measure the number of pixels occupied by the largest circumscribed rectangle in the input direction to obtain the pixel size. The pixel size reflects the size ratio of the target object in the image. Combined with the obtained physical size, it is possible to establish a size correspondence between the image and the actual physical space, thereby preparing for the subsequent accurate proportional adaptation of the three-dimensional human virtual model.

[0090] The pixel size is the length distance corresponding to the input direction in the interaction image.

[0091] A3: performing proportional adaptation on the three-dimensional human body virtual model according to the ratio of the physical size and the pixel size in the same input direction.

[0092] It can be understood that after obtaining the physical size and pixel size of the target object in the same input direction, the proportional relationship between the two is calculated, which represents the conversion relationship between the pixels in the image and the actual physical size. Then, the three-dimensional human body virtual model is adjusted according to the proportional relationship so that it matches the target object in the actual physical space in size and proportion.

[0093] For example, when the target object is actually large, but the corresponding pixel size in the image is relatively small, the three-dimensional human body virtual model will be appropriately enlarged according to the calculated ratio, and vice versa. By adjusting the model size proportionally, the three-dimensional human body virtual model can more realistically reflect the size and proportion of the target object in the actual scene, thereby improving the accuracy and authenticity of subsequent data collection and photography.

[0094] It is worth mentioning that when there are multiple input directions, the difference in proportions in multiple directions can be determined. If the difference is not large, the average can be calculated based on the proportions corresponding to multiple directions to obtain the average proportion. If the difference is too large, it means that uneven proportions will occur when adjusting the model. In this case, the three-dimensional human body virtual model can be adjusted automatically and manually to achieve adaptation of the model proportions.

[0095] S4, analyzing the relative posture relationship between the posture customization model and the target object to generate a customized posture relationship.

[0096] It is understandable that in order to determine the posture model that meets the user's needs, the posture customization model can be analyzed, and the relative posture relationship with the target object can be determined, and then a customized posture relationship can be generated, which provides a standard basis for subsequent judgment of whether the posture during actual photography meets the requirements. By accurately analyzing the relative posture relationship between the two, it can be ensured that when data is collected, the posture of the person and the target object is presented in accordance with the expected effect.

[0097] Among them, the relative posture relationship is the relative position relationship between the posture customization model and the target object. For example, when you want a person to stand behind a tree, the corresponding posture customization model is behind the target object, so the relative posture relationship between the two can be obtained as the model is placed behind the target object, thereby obtaining a customized posture relationship. The customized posture relationship is the posture relationship customized when collecting images.

[0098] It is not difficult to understand that different relative posture relationships will affect the recognition of the collection posture by the acquisition end. For example, when the posture customization model is in front of the target object, the posture customization model will not be blocked by the target object. When the posture customization model is behind the target object, the posture customization model will be blocked by the target object. The image content of the subsequent character recognition by the acquisition end is also different. Therefore, the customized posture relationship can be generated by parsing the relative posture relationship, so that the accuracy of image recognition can be improved subsequently.

[0099] In some embodiments, the specific implementation steps of step S4 (analyzing the relative posture relationship between the customized posture model and the target object to generate a customized posture relationship) include:

[0100] S41, extracting limb joint points of the posture customization model as posture customization model feature points through a human skeleton key point detection algorithm, wherein the limb joint points at least include coordinate information of hip joints, knee joints, and elbow joints in the interactive image.

[0101] It is understandable that in order to meet user needs, the posture customization model is customized, so that the joint points of the user posture customization model can be determined in advance, so that the corresponding joint points can be adjusted later to achieve customized actions that meet the needs.

[0102] Among them, the human skeleton key point detection algorithm is an algorithm that mainly detects some key points of the human body, such as DeepPose, Top-Down, Convolutional Pose Machines and other algorithms. This algorithm can accurately identify the limb joints of the model, such as hip joints, knee joints, elbow joints and other key parts. By obtaining the coordinate information of these joints in the interactive image, they are used as feature points of the posture customization model. These feature points can effectively characterize the posture of the posture customization model. Subsequently, by analyzing the relationship between these feature points and the feature points of the target object, the relative posture relationship between the two can be obtained, laying the foundation for generating customized posture relationships.

[0103] It is not difficult to understand that the limb joint points are the joint points of the limbs and torso corresponding to the posture customization model, including hip joints, knee joints, elbow joints, etc. The posture customization model feature points are the feature points for action customization of the posture customization model, that is, the limb joint points of the posture customization model.

[0104] S42: Using an edge detection algorithm to extract edge points of the target object as feature points of the target object.

[0105] It can be understood that the edge detection algorithm can effectively identify the boundary contour of the target object, thereby extracting the edge points of the object. These edge points, as the feature points of the target object, can reflect the shape and position information of the target object. By extracting the feature points of the target object, it provides key data support for the subsequent determination of the relative position relationship between the target object and the posture customization model, making it possible to more accurately analyze the posture relationship between the two.

[0106] The object edge point is an edge contour point corresponding to the target object, and the target object feature point is a feature point corresponding to the target object, that is, an object edge point of the target object.

[0107] It is not difficult to understand that the edge detection algorithm is an algorithm for detecting and analyzing the edges of an image, such as the Roberts operator, the Prewitt operator, the Sobel operator, and the like.

[0108] S43 , determining a relative position relationship between the characteristic points of the posture customized model and the characteristic points of the target object, wherein the relative position relationship includes a relative angle, and generating a customized posture relationship according to the relative position relationship.

[0109] It is understandable that after obtaining the feature points of the posture customization model and the feature points of the target object, it is necessary to further analyze the relative position relationship between the two. The relative position relationship not only includes the distance relationship between the feature points, but also covers information such as relative angles. By calculating and analyzing the relative position relationship, the relative posture between the posture customization model and the target object can be clearly determined. Finally, a customized posture relationship is generated based on the obtained relative position relationship. The customized posture relationship specifically describes the accurate posture of the posture customization model relative to the target object, and is an important criterion for subsequent judgment of whether the real-time posture relationship meets the requirements.

[0110] The relative position relationship is the corresponding position relationship between the feature points of the posture customization model and the feature points of the target object, and the relative angle is the inclination angle of the feature points of the posture customization model and the feature points of the target object relative to the horizontal plane.

[0111] In some embodiments, further comprising:

[0112] B1. When there is a spatial overlap between the posture customization model and the target object in the interactive image, a click instruction from the user on the overlapped area of ​​the target object is received.

[0113] It is understandable that in the interactive image, the user's corresponding posture customization model may be placed behind the target object, such as a tree or bench, that is, the posture customization model is placed behind the target object. As a result, the posture customization model and the target object may overlap in space, which will affect the visual effect of the image and the accuracy of data collection. At this time, the system will wait for the user's operation. When the user clicks on the overlapping area of ​​the target object, the system receives the click instruction. This instruction is issued by the user based on actual needs and visual perception. It is used to solve the problems caused by the overlapping area and provide a basis for subsequent adjustment of the display status of the posture customization model.

[0114] The target object overlapping area is an area where the posture customization model and the target object overlap, and the click instruction is operation instruction information for clicking the overlapping area.

[0115] It is not difficult to understand that when the user needs to take a pose in front of the target object, there is no need to click on the overlapping area when calling and placing the pose customization model. Figure 2 As shown, when the user needs to show a pose behind the target object to take a photo, the posture customization model is called and placed. Since the layer corresponding to the posture customization model is above the target object, it is necessary to click on the overlapping area to move the layer corresponding to the posture customization model downward to obtain the customized posture required by the user.

[0116] B2, rendering the limb part corresponding to the click instruction in the posture customization model as a hidden state, so that the overlapping area of ​​the posture customization model is displayed behind the target object, and marking the posture customization model feature points corresponding to the corresponding limb part as hidden feature points.

[0117] It can be understood that after receiving the user's click command, the limb part corresponding to the click position in the posture customization model can be located, and this limb part can be rendered as a hidden state. In this way, visually, the overlapping part of the posture customization model will be displayed behind the target object, avoiding the visual confusion caused by the overlap. At the same time, for the convenience of subsequent processing, the feature points of the posture customization model corresponding to the limb part are marked as hidden feature points. These hidden feature points will play a special role in the subsequent matching degree calculation and posture relationship adjustment.

[0118] B3: Generate visual constraints according to the hidden feature points, and update the customized posture relationship based on the visual constraints.

[0119] It can be understood that after the hidden feature points are marked, visual constraints can be generated based on the hidden feature points. The visual constraints specify the visibility rules that the posture customization model needs to meet when displayed, ensuring that the display relationship between the posture customization model and the target object meets the user's expectations. Then, the customized posture relationship is updated based on the visual constraints.

[0120] It is not difficult to understand that the updated customized posture relationship not only takes into account the relative position and angle of the posture customization model and the target object, but also considers the visibility of the overlapping area, making the customized posture relationship more complete and accurate, and providing a more reliable standard for subsequent data collection and posture matching.

[0121] In some embodiments, it further includes:

[0122] C1, delete the customized posture relationship.

[0123] It is understandable that since users need to make more detailed posture settings for the corresponding models and target objects, the original customized posture relationship needs to be deleted so that a customized posture relationship that better meets actual needs can be regenerated later to avoid interference of old information on subsequent operations.

[0124] C2, receiving association information of the corresponding posture customized model feature points and the target object feature points from the acquisition end, and determining the corresponding posture customized model feature points and the target object feature points as an association group according to the association information.

[0125] It is understandable that when the old customized posture relationship is deleted, the posture relationship needs to be re-established. The acquisition end will receive the user's association operation information on the posture customization model feature points and the target object feature points. For example, the selected posture customization model feature points and target object feature points can be triggered and clicked to associate the corresponding posture customization model feature points and target object feature points into the same group, thereby improving the accuracy and efficiency of the customized posture.

[0126] Among them, the association information is the operation information of associating the feature points of the posture customization model and the feature points of the target object. For example, it can be the operation information triggered by continuous clicks in a short period of time, so as to associate the clicked feature points. The association group is the feature point combination after associating the feature points of the posture customization model and the feature points of the target object.

[0127] It is not difficult to understand that due to the large number of posture customization model feature points and target object feature points, for example, when the arm in the model is set to support the backrest of the sub-target object chair, the node corresponding to the hand and the multiple target object feature points on the chair back have a certain angle. Therefore, in order to quickly perform posture comparison during subsequent posture recognition, the selected posture customization model feature points and target object feature points can be associated, so that the nodes corresponding to the associated group can be quickly compared subsequently to reduce the amount of data processing.

[0128] In some embodiments, it further includes:

[0129] The spacing distance between the characteristic points of the posture customization model and the characteristic points of each target object is determined, and the characteristic points of the target object are arranged in descending order according to the spacing distance to obtain a characteristic point sequence.

[0130] It should be noted that when selecting posture customization model feature points and target object feature points for association, the target object feature points closest to the posture customization model feature points can be selected for association. Therefore, the target object feature points can be arranged according to the interval distance so as to subsequently determine the target object feature points associated with the posture customization model feature points.

[0131] The spacing distance is the distance between the feature points of the posture customization model and the feature points of the target object, and the feature point sequence is the sequence of the feature points of the target object arranged in descending order.

[0132] The first target object feature point in the feature sequence is selected as the associated feature point.

[0133] It can be understood that the associated feature point is the target object feature point associated with the posture customization model feature point, that is, the first target object feature point in the feature sequence.

[0134] C3, determining the relative position relationship between the posture customization model feature points and the target object feature points in each of the association groups, and generating a customized posture relationship according to the relative position relationship.

[0135] It can be understood that after obtaining multiple association groups, the posture customization model feature points and target object feature points in each association group are analyzed to determine the relative position relationship between them, including relative distance, relative angle, etc. The relative position relationship can accurately describe the specific posture situation between the posture customization model and the target object. Then, based on the relative position relationship determined in all association groups, the customized posture relationship is regenerated. The newly generated customized posture relationship is more in line with the user's latest needs and settings, and provides an accurate standard for subsequent judgment of whether the real-time posture relationship meets the requirements.

[0136] S5, acquiring a real-time image based on the real-time signal of the acquisition end, and acquiring required data when the real-time posture relationship in the real-time image corresponds to the customized posture relationship.

[0137] It can be understood that the current real-time image is obtained through the real-time signal of the acquisition end. The real-time signal is the signal recorded in real time by the acquisition end. Then, the real-time posture relationship in the real-time image (that is, the relationship between the posture of the person in the image and the target object) is compared with the previously generated customized posture relationship (pre-set posture relationship that meets the requirements). When the real-time posture relationship meets the customized posture relationship, the required data is collected, that is, the person is automatically photographed and collected.

[0138] Among them, the real-time signal is the trigger signal recorded in real time by the acquisition end, the real-time image is the image information collected in real time by the acquisition end, for example, it can be an image photo collected by a camera, and the demand data is the photo image data required by the user.

[0139] Through the above implementation, the present invention can compare the user's current photographing posture with the preset customized posture relationship in real time, thereby automatically capturing images for the user to meet the user's needs and save the user's photographing time.

[0140] In some embodiments, the specific implementation steps of step S5 (collecting required data when the real-time posture relationship in the real-time image corresponds to the customized posture relationship) include:

[0141] S51 , analyzing the customized posture relationship to obtain posture customized model feature points and target object feature points.

[0142] It is understandable that before comparing the real-time posture relationship with the customized posture relationship, the customized posture relationship needs to be parsed first. Since the customized posture relationship is generated based on the previously extracted posture customization model feature points and target object feature points, these feature points can be obtained again by parsing the customized posture relationship. The parsed feature points are an important basis for subsequently judging whether the real-time posture relationship matches the customized posture relationship. Only when the feature points in the customized posture relationship are clarified can they be accurately compared with the feature points in the real-time image.

[0143] S52, using a human skeleton key point detection algorithm to detect limb key points in the real-time image that correspond to the gesture customized model feature points as real-time gesture feature points.

[0144] It can be understood that after obtaining the posture customization model feature points in the customized posture relationship, the real-time image is processed, and the human skeleton key point detection algorithm is used to find the limb key points corresponding to the posture customization model feature points in the real-time image. For example, the hip joint, knee joint, elbow joint, etc. are determined as the posture customization model feature points in the customized posture relationship, then the corresponding limb joint points of the character are detected in the real-time image through the algorithm, and used as the real-time posture feature points. These real-time posture feature points reflect the posture of the character in the real-time image and provide key information for determining the real-time posture relationship.

[0145] Among them, the limb key points are the nodes corresponding to the human torso, and the real-time posture feature points are the key points of the human torso in the real-time image.

[0146] S53: Using an edge detection algorithm, extract target feature points in the target object corresponding to the target object feature points as real-time target feature points.

[0147] It can be understood that, similarly, for the target object, an edge detection algorithm is used to extract points corresponding to the previously determined target object feature points in the real-time image. The edge detection algorithm can accurately identify the edge of the target object, thereby finding the real-time target feature points corresponding to the target object feature points. The real-time target feature points reflect the shape and position information of the target object in the real-time image, and together with the real-time posture feature points, are used to determine the real-time posture relationship.

[0148] S54 , determining a real-time posture relationship according to the real-time posture feature points and the real-time target feature points, and generating acquisition signal acquisition requirement data when a matching degree between the real-time posture relationship and the customized posture relationship is greater than a preset matching degree.

[0149] It can be understood that after obtaining the real-time posture feature points and the real-time target feature points, the relationship between them is analyzed to determine the real-time posture relationship, and then the real-time posture relationship is compared with the customized posture relationship to calculate the matching degree between the two. When the matching degree is greater than the preset matching degree threshold, it means that the real-time posture relationship is consistent with the customized posture relationship. At this time, a collection signal is generated to trigger the data collection operation and collect the required data. This ensures that the collected data meets the pre-set posture requirements and improves the quality and availability of the data.

[0150] The preset matching degree is the similarity between the preset real-time posture relationship and the customized posture relationship, and the acquisition signal is the instruction signal for performing image acquisition.

[0151] In some embodiments, the specific implementation steps for calculating the matching degree between the real-time posture relationship and the customized posture relationship before step S54 include:

[0152] D1, obtaining angle differences corresponding to multiple groups of feature points in the real-time posture relationship and the customized posture relationship, marking feature points whose angle differences are within a preset range as matching feature points, and counting the number of the matching feature points as a first matching number.

[0153] It can be understood that, first, the angles corresponding to multiple groups of feature points in the real-time posture relationship and the customized posture relationship are compared. Since the relative angle between the feature points is an important indicator for describing the posture relationship, by calculating the angle difference between them, it is possible to determine whether the postures represented by the two groups of feature points are similar. When the angle difference is within a pre-set range, it indicates that these feature points are matched in posture, and they are marked as matching feature points. Finally, the number of matching feature points is counted. This number is the first matching number, which preliminarily reflects the matching situation between the real-time posture and the customized posture in terms of feature point angles.

[0154] Among them, the angle difference is the angle difference between the relative angle of the feature point in the real-time posture relationship and the relative angle of the feature point in the customized posture relationship. The preset range is the pre-set angle difference range, such as ±5°. The matching feature point is the feature point with matching angles, and the first matching number is the number of matching feature points.

[0155] It is worth mentioning that since the posture customization model may be located behind the target object, the target object will block the posture customization model, and the corresponding posture customization model feature points will be blocked. Therefore, when obtaining multiple groups of feature points, the associated groups corresponding to the blocked and hidden posture customization model feature points will not be angle matched.

[0156] D2, determining the number of feature point groups that need to be compared in the customized posture relationship to obtain a first total number, and obtaining a first matching degree according to a ratio of the first matching number to the first total number.

[0157] It is understandable that after obtaining the first matching number, it is also necessary to determine the total number of feature point groups that need to be compared in the customized posture relationship, that is, the first total number. The first total number represents the size of all feature point groups participating in the comparison. Then, the first matching number is divided by the first total number. The ratio obtained is the first matching degree. The first matching degree reflects the degree of matching between the real-time posture relationship and the customized posture relationship in terms of ordinary feature point angles, and is an important part of the matching degree calculation.

[0158] Among them, the feature point group is the associated group corresponding to the feature points of the posture customization model and the feature points of the target object, the first total number is the number of feature point groups that need to be compared in the customized posture relationship, that is, the number of feature point groups corresponding to the hidden posture customization model feature points excluding the first matching number, and the first matching degree is the ratio of the first matching number to the first total number.

[0159] D3: When the hidden feature points exist, counting a second total number of the hidden feature points and a second matching number of the hidden feature points that meet the hidden state.

[0160] It is understandable that when the posture customization model in the customized posture relationship is behind the target object, there are joint points that are obscured, and the obscured key points are hidden feature points. For example, when there is a spatial overlapping area between the posture customization model and the target object in the interactive image, some posture customization model feature points may be marked as hidden feature points. At this time, it is necessary to count the total number of hidden feature points (i.e., the second total number) and the number of hidden feature points that meet the hidden state requirements (i.e., the second matching number) separately. Meeting the hidden state means that the performance of these hidden feature points in the real-time image meets the pre-set hidden conditions, such as being displayed behind the target object.

[0161] The hidden feature points are key points of the human body blocked by the target object, the second total number is the number of hidden feature points, and the second matching number is the number of hidden feature points that meet the hidden state.

[0162] D4: Obtain a second matching degree according to a ratio of the second matching number to the second total number, and obtain the matching degree based on the first matching degree and the second matching degree.

[0163] It can be understood that the ratio obtained by dividing the second matching number by the second total number is the second matching degree, which reflects the matching status of the hidden feature points in the real-time posture and the customized posture. Finally, the first matching degree and the second matching degree are combined to obtain the final matching degree.

[0164] The matching degree may be the sum of the first matching degree and the second matching degree.

[0165] It is not difficult to understand that the first matching degree and the second matching degree are combined to obtain the matching degree of the final real-time image. Among them, when calculating the matching degree of the real-time posture relationship in the real-time image based on the first matching degree and the second matching degree, the weight of the limb node can be increased to increase the importance of the corresponding node, so as to ensure that the matching degree can comprehensively and accurately reflect the overall matching degree of the real-time posture relationship and the customized posture relationship, and provide a reference basis for the subsequent judgment of whether to collect the required data.

[0166] In some embodiments, it further includes:

[0167] E1, load at least two initial posture models in the interaction image and associate them with different target objects respectively.

[0168] It is understandable that when multiple users need to take images, in order to achieve a combined scene setting of multiple characters or multiple postures with different target objects, based on the interactive image, at least two initial posture models (such as different character virtual image models) are loaded from a preset model library. These models can represent different users or roles. Then, each initial posture model is associated with a different target object, and a corresponding relationship between the posture model and the target object is established.

[0169] For example, one initial posture model is associated with a bench in the interactive image, and another is associated with the tree next to it. The association setting makes it easier for subsequent users to set postures for different models and realize multi-model collaborative data collection.

[0170] E2 receives the user's independent posture setting for each initial posture model and generates multiple posture customized models.

[0171] It can be understood that after multiple initial posture models are associated with different target objects, the user can perform independent posture setting operations for each initial posture model. The user adjusts the limb joints, postures, etc. of each model according to his or her own needs and creativity. The acquisition end receives the user's posture setting information and generates multiple posture customization models according to the setting information corresponding to each initial posture model.

[0172] Among them, each posture customization model is formed according to the user's specific settings for the corresponding initial posture model, reflecting the user's personalized posture requirements for different characters or models.

[0173] E3, analyzes the relative posture relationship between all posture customization models and the corresponding target objects, and generates joint trigger conditions.

[0174] It can be understood that after obtaining multiple posture customization models, the relative posture relationship between each posture customization model and its corresponding target object is analyzed, which includes extracting the limb joint feature points of the posture customization model through the human skeleton key point detection algorithm, and using the edge detection algorithm to extract the edge feature points of the target object, and then determining the relative position relationship between them (such as relative angle, relative distance, etc.), and integrating the relative posture relationships of all posture customization models and the corresponding target objects to generate a joint trigger condition.

[0175] Among them, the joint trigger condition is a comprehensive judgment criterion, which stipulates the combination of conditions that must be met for the posture relationship between all posture customization models and the target object. Only when these conditions are met will the subsequent data collection operation be triggered.

[0176] E4: When the real-time posture relationship of all users detected in the real-time image matches the joint trigger condition, data collection is performed.

[0177] It can be understood that the real-time posture relationship is compared with the previously generated joint trigger conditions. When it is detected that the real-time posture relationships of all users match the joint trigger conditions, it means that the current real-time scene meets the pre-set posture requirements. At this time, the data acquisition operation is performed to automatically collect the required image data to meet user needs.

[0178] like Figure 3 As shown, the present invention provides a data intelligent processing system, comprising:

[0179] The sending module is used to collect the customized background in the target scene, mark the objects in the customized background, generate an interactive image and send it to the collection end.

[0180] The determination module is used to determine the target object based on the selection information of the interactive image by the acquisition end.

[0181] The receiving module is used to call the initial posture model to associate with the target object, receive the posture setting information of the initial posture model from the acquisition end, and generate a posture customization model.

[0182] A generation module is used to analyze the relative posture relationship between the posture customization model and the target object to generate a customized posture relationship.

[0183] The acquisition module is used to obtain a real-time image based on the real-time signal of the acquisition end, and to acquire required data when the real-time posture relationship in the real-time image corresponds to the customized posture relationship.

[0184] See also Figure 4, is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention, wherein the electronic device 40 includes: a processor 41, a memory 42 and a computer program;

[0185] The memory 42 is used to store the computer program, which may also be a flash memory. The computer program is, for example, an application program or a functional module for implementing the above method.

[0186] The processor 41 is configured to execute the computer program stored in the memory to implement the various steps performed by the device in the above method. For details, please refer to the relevant description in the above method embodiment.

[0187] Optionally, the memory 42 may be independent or integrated with the processor 41 .

[0188] When the memory 42 is a device independent of the processor 41, the device may further include:

[0189] The bus 43 is used to connect the memory 42 and the processor 41 .

[0190] The present invention also provides a readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, it is used to implement the methods provided in the various embodiments described above.

[0191] The readable storage medium may be a computer storage medium or a communication medium. Communication media include any medium that facilitates the transfer of computer programs from one location to another. Computer storage media may be any available medium that can be accessed by a general-purpose or special-purpose computer. For example, a readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium may also be an integral part of the processor. The processor and the readable storage medium may be located in an application-specific integrated circuit (ASIC). In addition, the ASIC may be located in a user device. Of course, the processor and the readable storage medium may also exist as discrete components in a communication device. The readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0192] The present invention also provides a program product, which includes execution instructions stored in a readable storage medium. At least one processor of a device can read the execution instructions from the readable storage medium, and at least one processor executes the execution instructions so that the device implements the methods provided in the various embodiments described above.

[0193] In the embodiments of the above-mentioned devices, it should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASICs), etc. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention may be directly executed by a hardware processor or by a combination of hardware and software modules within the processor.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent data processing, characterized in that: include: Collecting the customized background in the target scene, annotating the objects in the customized background to generate an interactive image and sending it to the collection end; Determining a target object based on the selected information of the interactive image by the acquisition end; Calling the initial posture model to associate with the target object, receiving the posture setting information of the initial posture model from the acquisition end, and generating a posture customization model; Analyzing the relative posture relationship between the posture customization model and the target object to generate a customized posture relationship; Acquire a real-time image based on the real-time signal of the acquisition terminal, and collect required data when the real-time posture relationship in the real-time image corresponds to the customized posture relationship; Also includes: When there is a spatial overlap between the posture customization model and the target object in the interactive image, receiving a click instruction from the user on the overlapped area of ​​the target object; Rendering the limb part corresponding to the click instruction in the posture customization model as a hidden state, so that the overlapping area of ​​the posture customization model is displayed behind the target object, and marking the posture customization model feature points corresponding to the corresponding limb part as hidden feature points; A visual constraint condition is generated according to the hidden feature point, and the customized posture relationship is updated based on the visual constraint condition.

2. The data intelligent processing method according to claim 1, characterized in that: The initial posture model is a three-dimensional human body virtual model. The initial posture model is called to associate with the target object, posture setting information of the initial posture model is received from the acquisition end, and a posture customization model is generated, including: Calling a 3D human body virtual model from a preset model library and associating it with the target object to display it on the interactive image interface of the acquisition end; Generating posture setting information according to the limb joint point setting of the three-dimensional human body virtual model by the acquisition end; The posture of the three-dimensional human body virtual model is updated based on the posture setting information to generate a posture customized model.

3. The data intelligent processing method according to claim 2, characterized in that: Also includes: Extracting a maximum bounding rectangle of the target object in the interactive image, and receiving an input direction selected by the acquisition end based on the maximum bounding rectangle and a physical size corresponding to the input direction; Obtain the pixel size of the maximum circumscribed rectangle corresponding to the input direction in the interaction image; The three-dimensional human body virtual model is proportionally adapted according to the ratio of the physical size and the pixel size in the same input direction.

4. The data intelligent processing method according to claim 1, characterized in that: The step of analyzing the relative posture relationship between the posture customization model and the target object to generate a customized posture relationship includes: Extracting limb joints of the posture customization model as feature points of the posture customization model through a human skeleton key point detection algorithm, wherein the limb joints include at least coordinate information of the hip joint, knee joint, and elbow joint in the interactive image; An edge detection algorithm is used to extract the edge points of the target object as the feature points of the target object; The relative position relationship between the characteristic points of the posture customized model and the characteristic points of the target object is determined, wherein the relative position relationship includes a relative angle, and a customized posture relationship is generated according to the relative position relationship.

5. The data intelligent processing method according to claim 4, characterized in that: Also includes: Deleting the customized posture relationship; receiving association information of the corresponding posture customized model feature points and the target object feature points from the acquisition end, and determining the corresponding posture customized model feature points and the target object feature points as an association group according to the association information; The relative positional relationship between the characteristic points of the posture customization model and the characteristic points of the target object in each of the association groups is determined, and a customized posture relationship is generated according to the relative positional relationship.

6. The data intelligent processing method according to claim 5, characterized in that: When the real-time posture relationship in the real-time image corresponds to the customized posture relationship, collecting required data includes: Analyzing the customized posture relationship to obtain posture customization model feature points and target object feature points; The limb key points corresponding to the posture customized model feature points in the real-time image are used as real-time posture feature points through the human skeleton key point detection algorithm; An edge detection algorithm is used to extract target feature points in the target object corresponding to the target object feature points as real-time target feature points; A real-time posture relationship is determined according to the real-time posture feature points and the real-time target feature points, and when a matching degree between the real-time posture relationship and the customized posture relationship is greater than a preset matching degree, acquisition signal acquisition requirement data is generated.

7. The data intelligent processing method according to claim 6, characterized in that: Calculate the matching degree between the real-time posture relationship and the customized posture relationship, including: Obtaining angle differences corresponding to multiple groups of feature points in the real-time posture relationship and the customized posture relationship, marking feature points whose angle differences are within a preset range as matching feature points, and counting the number of the matching feature points as a first matching number; Determining the number of feature point groups that need to be compared in the customized posture relationship to obtain a first total number, and obtaining a first matching degree based on a ratio of the first matching number to the first total number; When the hidden feature points exist, counting a second total number of the hidden feature points and a second matching number of the hidden feature points that meet the hidden state; A second matching degree is obtained according to a ratio of the second matching number to the second total number, and the matching degree is obtained based on the first matching degree and the second matching degree.

8. The data intelligent processing method according to claim 1, characterized in that: Also includes: Load at least two initial pose models in the interaction image and associate them with different target objects respectively; receiving the user's independent posture setting for each initial posture model and generating multiple posture customized models; Analyze the relative posture relationship between all posture customization models and the corresponding target objects, and generate joint trigger conditions; When the real-time posture relationship of all users detected in the real-time image matches the joint trigger condition, data collection is performed.

9. A data intelligent processing system according to the data intelligent processing method of claim 1, characterized in that: include: A sending module is used to collect the customized background in the target scene, mark the objects in the customized background, generate an interactive image and send it to the collection end; A determination module, configured to determine a target object based on the selection information of the interactive image by the acquisition end; A receiving module, configured to call an initial posture model to associate with the target object, receive posture setting information of the initial posture model from the acquisition end, and generate a posture customization model; A generation module, configured to analyze the relative posture relationship between the posture customization model and the target object to generate a customized posture relationship; The acquisition module is used to obtain a real-time image based on the real-time signal of the acquisition end, and to acquire required data when the real-time posture relationship in the real-time image corresponds to the customized posture relationship.

Images