Information processing method, method for obtaining analysis model, device and electronic equipment

Abstract

The embodiment of the present application provides an information processing method, including: obtaining a video frame to be analyzed in real time, wherein the video frame to be analyzed contains a target object; analyzing at least two video frames to be analyzed by a position information analysis model to obtain position change information of multiple specified parts of the target object; if the position change information of multiple specified parts of the target object meets the position change condition, it is determined that the target object has performed a preset movement, and counting processing is performed for the preset movement performed by the target object. This counting method avoids the use of manual counting, and the counting result is more accurate; at the same time, since the video frame to be analyzed can be analyzed only by the position information analysis model to obtain the position information of multiple specified parts of the target object in the video frame to be analyzed, thereby providing a basis for subsequently determining whether the target object has performed a preset movement and counting, compared with the counting method of the prior art, this counting method is more convenient.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and more particularly to an information processing method, a method for obtaining a position information analysis model, a device and an electronic device, and also to a method for determining an amount of exercise and another information processing method. Background Art

[0002] With the continuous improvement of modern living standards, people pay more and more attention to health issues. More and more people monitor their physical health by regularly conducting some physical test activities. When conducting physical test-related activities, it is necessary to monitor people's physical activities. Many existing applications can monitor some physical activities. For example, some smart wearable devices can monitor the wearer's heart rate and number of steps when running.

[0003] However, for some relatively complex pre-designed sports activities, such as rope skipping and pull-up sports, manual counting is required when monitoring the number of times such sports are completed. Some monitoring methods in the prior art use multiple relatively complex computing devices to count such sports. For example, when monitoring the number of times pull-up sports are completed, that is, when counting pull-up sports, a series of complex infrared counting devices are required to determine whether a standard pull-up action has been completed. Obviously, the counting method used in the above manner is relatively complex. Manual counting is time-consuming and labor-intensive and is prone to inaccurate counting. Summary of the invention

[0004] The embodiment of the present application provides an information processing method to solve the problems of complex, time-consuming, labor-intensive and inaccurate counting methods for sports in the prior art. The embodiment of the present application provides an information processing device, an electronic device and a computer storage medium. The embodiment of the present application also provides a method, a device, an electronic device and a computer storage medium for obtaining a position information analysis model. The embodiment of the present application also provides a method for determining the amount of exercise and another information processing method.

[0005] The present application provides an information processing method, including:

[0006] Obtaining a video frame to be analyzed in real time, wherein the video frame to be analyzed contains a target object;

[0007] Analyzing at least two of the video frames to be analyzed by a position information analysis model to obtain position change information of multiple specified parts of the target object; the position information analysis model is used to obtain position information of multiple specified parts of the target object in the video frames to be analyzed according to the video frames to be analyzed;

[0008] If the position change information of the multiple designated parts of the target object meets the position change condition, it is determined that the target object has performed a preset movement, and counting processing is performed on the preset movement performed by the target object.

[0009] Optionally, analyzing at least two of the to-be-analyzed video frames by using a position information analysis model to obtain position change information of multiple designated parts of the target object includes:

[0010] Using each of the at least two video frames as input data of the position information analysis model, obtaining position information of multiple designated parts of the target object in each of the at least two video frames;

[0011] Based on the position information of the multiple designated parts of the target object in each video frame, the position change information of the multiple designated parts of the target object is obtained.

[0012] Optionally, the location information analysis model is set on the mobile terminal.

[0013] Optionally, the position change information of the multiple designated parts of the target object satisfies the position change condition, including:

[0014] The difference between the local maximum value of coordinates and the local minimum value of coordinates corresponding to the position of each designated part of the multiple designated parts matches the position change condition.

[0015] Optionally, also include:

[0016] Determining whether the position change information of the plurality of designated parts of the target object meets the position change condition;

[0017] If the position change information of multiple specified parts of the target object does not meet the position change condition, it is determined that the target object has not performed the preset movement, and it is judged whether the position change information of multiple specified parts of the target object in the next video frame to be analyzed meets the position change condition.

[0018] Optionally, also include:

[0019] Determine whether the current time information has exceeded the counting time condition of the preset motion; if the current time information has exceeded the counting time condition of the preset motion, stop obtaining the video frame to be analyzed in real time, and obtain the counting result for the preset motion.

[0020] Optionally, the method further includes: displaying the counting result for the preset motion to the target object, and storing the counting result.

[0021] Optionally, also include:

[0022] Obtaining information about the degree of movement of the target object performing the preset movement;

[0023] The movement degree information is provided to a user so that the user can analyze the preset movement performed by the target object.

[0024] Optionally, obtaining the video frame to be analyzed in real time includes:

[0025] Using a mobile camera to take real-time video of the target object to obtain a video file containing the target object;

[0026] According to the video file containing the target object, a video frame to be analyzed is obtained.

[0027] Optionally, using each of the at least two video frames as input data of the position information analysis model to obtain position information of multiple designated parts of the target object in each of the at least two video frames includes:

[0028] Compressing each of the at least two video frames according to the input image data volume scale requirement of the position information analysis model to obtain at least two compressed video frames;

[0029] The at least two compressed video frames are respectively used as input data of the position information analysis model to obtain position information of multiple designated parts of the target object in each of the at least two video frames.

[0030] Optionally, the location information analysis model is obtained in the following manner:

[0031] Obtaining a key point position information analysis model running on a mobile terminal; wherein the key point position information analysis model running on a mobile terminal runs on a mobile terminal and can obtain key point position information of a target object contained in an image according to an input image;

[0032] Using the key point position information analysis model running on the mobile terminal as the initial position information analysis model;

[0033] The original convolution kernels in the initial position information analysis model framework are replaced with convolution kernels of preset size and preset number, and the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework are replaced with partial affinity field frameworks and heat map prediction frameworks of preset ratio and preset number to obtain an optimized position information analysis model, and the optimized position information analysis model is used as the position information analysis model.

[0034] Optionally, replacing the original convolution kernel in the initial position information analysis model framework with a convolution kernel of a preset size and a preset number includes:

[0035] The original 7x7 convolution kernel in the initial position information analysis model framework is replaced by three consecutive 3x3 convolution kernels;

[0036] The replacing the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with the partial affinity field framework and the heat map prediction framework with a preset ratio and a preset number includes:

[0037] In the initial position information analysis model framework, a partial affinity field framework is used to replace the original partial affinity field framework, and a heat map prediction framework is used to replace the original heat map prediction framework.

[0038] Optionally, also include:

[0039] For the maximum value of the heat map output by the heat map prediction framework in the optimized position information analysis model framework, a 3x3 pixel neighborhood is taken to perform pixel point refinement.

[0040] Optionally, if the input image contains multiple target objects, a bilinear interpolation mechanism is used to extract points corresponding to some affinity fields on the input image.

[0041] The present application also provides a method for obtaining a location information analysis model, including:

[0042] Obtaining a key point position information analysis model running on a mobile terminal; wherein the key point position information analysis model running on a mobile terminal runs on a mobile terminal and can obtain key point position information of a target object contained in an image according to an input image;

[0043] Using the key point position information analysis model running on the mobile terminal as the initial position information analysis model;

[0044] The original convolution kernel in the initial position information analysis model framework is replaced with a convolution kernel of a preset size and a preset number, and the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework are replaced with a partial affinity field framework and a preset number of heat map prediction frameworks to obtain the target position information analysis model.

[0045] The present application embodiment provides an information processing device, including:

[0046] A video frame obtaining unit, used for obtaining a video frame to be analyzed in real time, wherein the video frame to be analyzed contains a target object;

[0047] A position change information obtaining unit, configured to analyze at least two of the to-be-analyzed video frames through a position information analysis model to obtain position change information of a plurality of designated parts of the target object; the position information analysis model is configured to obtain position information of a plurality of designated parts of the target object in the to-be-analyzed video frames according to the to-be-analyzed video frames;

[0048] A counting processing unit is used to determine that the target object has performed a preset movement if the position change information of multiple designated parts of the target object meets the position change condition, and to perform counting processing on the preset movement performed by the target object.

[0049] The embodiment of the present application provides a device for obtaining a location information analysis model, including:

[0050] An initial analysis model obtaining unit is used to obtain a key point position information analysis model running on a mobile terminal; wherein the key point position information analysis model running on the mobile terminal runs on the mobile terminal and can obtain key point position information of a target object contained in an image according to an input image; and the key point position information analysis model running on the mobile terminal is used as an initial position information analysis model;

[0051] The target position information analysis model acquisition unit is used to replace the original convolution kernel in the initial position information analysis model framework with a convolution kernel of a preset size and a preset number, and replace the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with a preset ratio and a preset number of partial affinity field frameworks and heat map prediction frameworks to obtain the target position information analysis model.

[0052] The present application provides a method for determining an amount of exercise, including:

[0053] Receiving a video frame containing a moving object captured by a mobile terminal;

[0054] Performing real-time analysis on the video frames containing the moving object to obtain position change information of multiple specified parts of the moving object; the real-time analysis is performed by a position information analysis model set in the mobile terminal, and the position information analysis model is used to obtain position change information of multiple specified parts of the target object in the video frame based on the video frames containing the moving object collected by the mobile terminal;

[0055] If the position change information of the multiple designated parts of the moving object meets the position change condition, counting processing is performed on the movement of the moving object to obtain a counting result.

[0056] Optionally, also include:

[0057] Get the sports type information set by the user;

[0058] The position information analysis model is set using analysis parameters corresponding to the motion type information; in the step of performing real-time analysis on the video frame containing the motion object, the position information analysis model with the analysis parameters set is used for analysis.

[0059] Optionally, also include:

[0060] Obtaining motion parameters corresponding to the motion type information;

[0061] After obtaining the counting result, performing data processing on the counting result; and comparing the measurement result obtained after the data processing with the motion parameter to obtain a comparison result;

[0062] According to the comparison result, relevant movement prompts are issued through the mobile terminal.

[0063] The present application provides an information processing method, including:

[0064] receiving a video frame containing a moving object captured by a fitness device;

[0065] Performing real-time analysis on the video frames containing the moving object to obtain position change information of the designated part of the moving object; the real-time analysis is performed by a position information analysis model set in the fitness device, and the position information analysis model is used to obtain the position change information of the designated part of the target object in the video frame based on the video frames containing the moving object collected by the fitness device;

[0066] If the position change information of the designated part of the moving object meets the position change condition, it is determined that the target object has performed a preset movement, and counting processing is performed on the preset movement performed by the target object.

[0067] An embodiment of the present application provides an electronic device, which includes: a processor; a memory for storing a computer program, wherein the computer program is run by the processor to execute the methods described in the above-mentioned information processing method embodiment, the position information analysis model obtaining method embodiment, and the motion amount determination method embodiment.

[0068] An embodiment of the present application provides a computer storage medium storing a computer program. The computer program is executed by a processor to execute the methods described in the above-mentioned information processing method embodiment, position information analysis model obtaining method embodiment, and movement amount determination method embodiment.

[0069] Compared with the prior art, the embodiments of the present application have the following advantages:

[0070] The embodiment of the present application provides an information processing method, including: obtaining a video frame to be analyzed in real time, wherein the video frame to be analyzed contains a target object; analyzing at least two of the video frames to be analyzed by a position information analysis model to obtain position change information of multiple specified parts of the target object; the position information analysis model is used to obtain the position information of multiple specified parts of the target object in the video frame to be analyzed according to the video frame to be analyzed; if the position change information of the multiple specified parts of the target object meets the position change condition, it is determined that the target object has performed a preset movement, and the preset movement performed by the target object is counted. This counting method avoids the use of manual counting, and the counting result is more accurate; at the same time, since the video frame to be analyzed can be analyzed only by the position information analysis model to obtain the position information of multiple specified parts of the target object in the video frame to be analyzed, thereby providing a basis for subsequent determination of whether the target object has performed a preset movement and counting, compared with the counting method of the prior art, this counting method is more convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the embodiments of the present application 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 recorded in the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0072] Figure 1A This is a first schematic diagram of an application scenario embodiment of the information processing method provided in the present application.

[0073] Figure 1B This is a second schematic diagram of an application scenario embodiment of the information processing method provided in the present application.

[0074] Figure 2 This is a flowchart of an information processing method provided in the first embodiment of the present application.

[0075] Figure 3 This is a schematic diagram of an optimized location information analysis model provided in the first embodiment of the present application.

[0076] Figure 4 This is a flowchart of a method for obtaining a location information analysis model provided in the second embodiment of the present application.

[0077] Figure 5 This is a schematic diagram of an information processing device provided in the third embodiment of the present application.

[0078] Figure 6 This is a schematic diagram of a device for obtaining a location information analysis model provided in the fourth embodiment of the present application.

[0079] Figure 7 This is a schematic diagram of an electronic device provided in the seventh embodiment of the present application. DETAILED DESCRIPTION

[0080] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application. Therefore, the present application is not limited to the specific implementation disclosed below.

[0081] The present application provides an information processing method, device, electronic device and computer storage medium. The present application also provides a method, device, electronic device and computer storage medium for obtaining a position information analysis model. The present application also provides a method for determining the amount of movement and another information processing method. The following are specific embodiments.

[0082] In order to more clearly demonstrate the first information processing method provided by the present application, the application scenario of the first information processing method provided by the present application is first introduced. The first information processing method provided by the present application can be applied to the counting scenarios of processing some sports. The scenario embodiment of the first information processing method provided by the present application can be completed only on the mobile terminal. The mobile terminal can be an electronic device equipped with a camera device and loaded with a location information analysis model running on the mobile terminal, such as a smart phone, a PC (Personal Computer) terminal, a tablet computer, a smart speaker, and a smart set-top box. The above-mentioned smart phone, PC (Personal Computer) terminal, tablet computer, smart speaker, and smart set-top box collect motion video frames containing target objects through the camera device installed inside them, and analyze the collected motion video frames. In the scenario embodiment of the first information processing method provided by the present application, taking the mobile terminal as a smart phone as an example, taking the user participating in sports as the target object, the scenario embodiment of the first information processing method provided by the present application is described in detail.

[0083] Taking the use of a smart phone to count rope skipping as an example, before the user skips rope, turn on the mobile phone camera and fix the mobile phone, and ensure that the user's whole body is within the camera range of the camera during the process of skipping rope. When the user starts skipping rope, the camera of the mobile phone can collect the video of the user's skipping rope in real time, and transmit the real-time collected video to the position information analysis model running in real time on the mobile phone. In this scenario embodiment, the position information analysis model refers to a mobile terminal model that can obtain the position information of multiple specified parts of the user in the image based on the image. The multiple specified parts can be the user's body torso, hands, feet, head and other specified parts. In this scenario embodiment, the user's body torso, hands, and feet are used as the specified parts. Since the camera on the electronic device collects the video of the user skipping rope in real time, the real-time collected video frame can be used as an image.

[0084] The above-mentioned position information analysis model is a human key point position information analysis model based on a convolutional neural network and a mobile phone deep learning framework. The human key point position analysis detection model can detect the positions of the 14 human skeleton key points of the user in the video frame in real time according to the obtained video frame. The 14 human skeleton key points are 2D (2dimension, i.e., 2-dimensional) key points, and the position of each key point is represented by the coordinate value (x, y) at the current mobile phone screen resolution of the mobile phone. Specifically, the 14 human 2D key points are in order of 1-top of the head, 2-neck, 3-right shoulder, 4-right elbow, 5-right wrist, 6-left shoulder, 7-left elbow, 8-left wrist, 9-right hip, 10-right knee, 11-right ankle, 12-left hip, 13-left knee, 14-left ankle. In the rope skipping motion counting scenario, three groups of key points can be extracted from the 14 key points: human torso key points (1-top of the head, 2-neck, 9-right hip, 12-left hip), human wrist key points (5-right wrist, 8-left wrist), and human ankle key points (11-right ankle, 14-left ankle). The reason for this method of extraction is that the human torso is similar to a rigid body during rope skipping, and the deformation is not large, so it is easier to extract the change information of the coordinate value in the y-axis direction. The human wrist and human ankle are parts that are closely related to rope skipping, and they have regular movements themselves, which are suitable for judging a complete rope skipping action. For example, the wrist is in a relatively high position when taking off, and in a relatively low position at the highest point of the jump. The movement pattern of the ankle is just the opposite. It is on the ground when taking off, and in a relatively high position at the highest point of the jump. It is also easier to extract the change information of the coordinate value in the y-axis direction from the human wrist and human ankle. Of course, it is understandable that when establishing a coordinate system, a rectangular coordinate system can be established with the y-axis direction perpendicular to the ground and the x-axis direction parallel to the ground.

[0085] In this embodiment, according to the actual situation of rope skipping, the above three groups of key point position change information are used as the position change information of the designated part, and the above position change information are respectively the change information of the coordinate value in the y-axis direction, and the position change information of the designated part is used as the basis for rope skipping counting. The position information analysis model analyzes the video frames to be analyzed obtained in real time, and the three groups of key point position information of the user in the video frames can be obtained. Based on the three groups of key point position information of the user in at least two video frames, the three groups of key point position change information of the user can be obtained. Of course, it is also possible to analyze the three groups of key point position information of the user in multiple video frames. In this application, the three groups of key point position information of the user in multiple video frames are analyzed to obtain the three groups of key point position change information of the user.

[0086] After obtaining the three sets of key point position change information of the user, it is determined whether each set of key point position change information meets the preset counting conditions of the rope skipping movement. The preset counting conditions of the rope skipping movement are the preset position change conditions of the object parts of the movement. Assume that the preset counting conditions of the rope skipping movement are: the key point position change information of the human body trunk is required to be greater than a, the key point position change information of the human body wrist is required to be greater than b, and the key point position change information of the human body ankle is required to be greater than c. It can be that only when the three sets of key point position change information obtained all meet the above counting conditions can it be determined that the user has completed a complete rope skipping action, and the number of rope skipping is increased by one on the basis of the original number of rope skipping. Of course, it can also be that as long as one set of key point position change information or two sets of key point position change information of the above three sets of key point position change information meets the key point position change information calculation conditions of the group. For example, if the key point position change information of the human body ankle is greater than c, it can also be determined that the user has completed a complete rope skipping action, and the number of rope skipping is increased by one on the basis of the original number of rope skipping. Of course, the key points corresponding to the above three groups of key point position change information may also be other human key points, as long as the position change information corresponding to the selected key points during movement can be used to determine whether movement has occurred.

[0087] Specifically, see Figure 1A , which shows the first schematic diagram of the embodiment of the scenario of the present application. The process of counting rope skipping in this way is as follows: first, turn on the camera device of the mobile phone. After turning on the camera device, the camera device can collect video frames of rope skipping movement in real time, and start counting rope skipping at the same time. When counting, the currently collected video frames are used as input data of the position information analysis model in advance, so as to obtain the 14 2D key point position information of the user in the video frame in real time. Afterwards, three groups of key point position information are obtained respectively. Based on multiple video frames obtained in real time, three groups of key point position change information can be obtained. In this embodiment, each group of key points uses position change curve information to represent the position change information.

[0088] After obtaining the three groups of key point position change curve information, it is determined whether the difference between the local maximum and the local minimum of each group of key point position change curves matches the specified counting threshold condition. The local maximum and the local minimum can be the y values ​​corresponding to the adjacent local highest point and the local lowest point on the curve. If it is determined that the difference between the local maximum and the local minimum of each group of key point position change curves matches the specified counting condition, the skipping rope count is added by 1. At the same time, it is also necessary to determine whether the skipping rope counting time has expired. If the judgment result is no, continue to obtain the video frames of the skipping rope movement in real time, and continue to use the collected video frames as input data for the position information analysis model. If the skipping rope counting time has expired, no more video frames will be collected, the skipping rope counting ends, and the final counting result of the skipping rope is obtained.

[0089] In this embodiment, for the counting scenario of motion, counting can be achieved based on the location information analysis model running on the mobile phone, such as Figure 1B As shown, it is a second schematic diagram of an application scenario embodiment of the information processing method provided by the present application. A video frame containing a target object is sent to a mobile phone terminal, and the mobile phone terminal analyzes the video frame through a position information analysis model running by itself, thereby obtaining the position information of multiple specified parts of the target object in the video frame to be analyzed. If the position change information of multiple specified parts of the target object meets the position change condition, it is determined that the target object has performed a preset movement, and counting processing is performed on the preset movement performed by the target object. Of course, it is understandable that it is also necessary to obtain the position change information of multiple specified parts of the target object, and the video frame containing the target object must be multiple video frames.

[0090] In fact, the information processing method of the present application can also be applied in fitness scenes. For example, a fitness device placed in a gym or at home and equipped with a camera device can be used to collect video frames containing moving objects and analyze the video frames. Specifically, the analysis process is implemented by a position information analysis model installed on a mobile terminal or a fitness device. The position information of multiple specified parts of the moving object in the video frame can be obtained through the position information analysis model. In practice, the multiple specified parts can be key points of the human body. After obtaining the position information of multiple specified parts of the moving object in multiple video frames, the position change information of the multiple specified parts of the moving object can be obtained, thereby obtaining the movement degree information of the moving object. For example, the counting result of the moving object's movement. Through the above process, it can be achieved in the gym or at home, only the mobile terminal or fitness equipment needs to detect the movement degree information of the moving object, and at the same time, it is also convenient for the moving object to grasp the movement degree information of itself.

[0091] Of course, the sports of the present application are not limited to rope skipping, and the information processing method of the present application is only described by rope skipping as an example. The sports are not specifically limited in the embodiments of the present application, and the information processing method of the present application can also be applied to pull-up sports, badminton sports, and other sports involving changes in designated parts of sports objects. At the same time, the sports degree information of the present application is not limited to sports counting results, but can also be sports posture information of sports objects.

[0092] The application scenario of the above-mentioned information processing method for counting rope skipping movements is only one embodiment of the application scenario of the information processing method provided by the present application. The purpose of providing this application scenario embodiment is to facilitate understanding of the information processing method provided by the present application, and is not used to limit the information processing method provided by the present application. Other application scenarios of the information processing method provided in the embodiments of the present application are not described one by one.

[0093] First embodiment

[0094] The first embodiment of the present application provides an information processing method, as follows: Figure 2-Figure 3 Provide explanation.

[0095] Please refer to Figure 2 , which is a flowchart of an information processing method provided in the first embodiment of the present application.

[0096] In step S201, a video frame to be analyzed is obtained in real time, and the video frame to be analyzed contains a target object.

[0097] First, the video frames to be analyzed are obtained in real time. In this embodiment, multiple video frames containing the target object are collected in real time as the video frames to be analyzed, and the multiple video frames are analyzed to obtain the position change information of multiple specified parts of the target object in the multiple video frames. Through the position change information of the multiple specified parts of the target object, it can be determined whether the target object has performed a preset movement, and the preset movement that has been performed can be counted.

[0098] Multiple video frames refer to video frames that all contain a target object and are used to count and analyze preset movements. The target object may refer to a user performing a preset movement, and the preset movement may refer to some sports, including rope skipping, pull-ups, and the like. Since it is based on the position change information of multiple designated parts of the target object to determine whether the target object has performed a preset movement, and the preset movement that has been performed is counted, the multiple video frames collected in real time include the target object. For example, the multiple video frames collected may be multiple continuous video frames that all contain user A performing rope skipping, or multiple continuous video frames that all contain user B performing pull-ups.

[0099] As one of the ways to obtain multiple video frames in real time, it can be: first, use the camera device on the mobile terminal to take real-time video of the target object to obtain a video file containing the target object. Then, based on the video file containing the target object, multiple video frames are obtained. As for obtaining multiple video frames based on the video file, the existing more mature method of extracting video frames through video files can be used, which will not be described in detail here. Since the camera device on the mobile terminal takes real-time video of the target object, the multiple video frames must contain the target object.

[0100] In step S202, at least two video frames to be analyzed are analyzed by using a position information analysis model to obtain position change information of multiple designated parts of the target object.

[0101] After acquiring multiple video frames in step S201, the multiple video frames are analyzed based on the position information analysis model to obtain position change information of multiple designated parts of the target object. It should be noted that the multiple video frames are at least two video frames.

[0102] In step S201, it has been explained that this embodiment uses the position change information of multiple designated parts of the target object in multiple videos to determine whether the target object has performed a preset movement. The position change information of the multiple designated parts is necessarily related to the preset movement. When skipping is used as the preset movement, the multiple designated parts of the target object can be a human body trunk, a human body wrist, and a human body ankle. The position change information of the designated parts can refer to the position coordinate change information of the designated parts of the target object at different times caused by the process of performing the preset movement. Since the human body trunk will change in the vertical direction during skipping, and the human body wrist and human body ankle are parts closely related to the skipping movement, and they will have regular movements themselves, the above parts are suitable for determining whether the user has performed a complete skipping action. For example, the user's wrist is at a relatively high position when taking off, and is at a relatively low position at the highest point of the jump. The movement law of the ankle is just the opposite. The user's ankle is at the ground when taking off, and is at a relatively high position at the highest point of the jump. Through the position change information of these designated parts of the target object, it can be determined whether the target object has performed a preset movement.

[0103] Before obtaining the position change information of multiple specified parts of the target object, the position information of multiple specified parts of the target object must first be obtained. The position information of multiple specified parts of the target object is obtained through a position information analysis model. The position information analysis model in this embodiment refers to a model for obtaining the position information of multiple specified parts of the target object in the video frame to be analyzed based on the video frame to be analyzed. The position information analysis model in this embodiment can run on a mobile terminal.

[0104] Regarding the position information analysis model in this embodiment, compared with the existing human body key point position information analysis model running on the mobile terminal, the overall calculation amount on the mobile terminal is reduced, and the model accuracy is significantly improved.

[0105] The existing human key point position information analysis model running on a mobile terminal is a model that can run on a mobile terminal and can obtain the key point position information of the human body contained in the image according to the input image, and can be a human skeleton key point positioning model. The key point position information of the human body is used to locate the positions of the head, neck, shoulders, elbows, hands, hips, knees, feet and other parts of the human body in the image. In order to further improve the operating efficiency and processing performance of the existing human key point position information analysis model running on a mobile terminal, it is optimized as an initial position information analysis model.

[0106] Specifically, the optimization of the initial position information analysis model is divided into two aspects, namely, optimization at the model framework level and optimization at the post-processing level.

[0107] First, the optimization at the model framework level includes: replacing the original convolution kernels in the initial position information analysis model framework with convolution kernels of preset size and preset number, and replacing the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with partial affinity field frameworks and heat map prediction frameworks with preset ratios and preset numbers.

[0108] More specifically, the original convolution kernel in the initial position information analysis model framework is replaced with a convolution kernel of a preset size and a preset number, including: replacing the original 7x7 convolution kernel in the initial position information analysis model framework with three consecutive 3x3 convolution kernels; replacing the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with a preset ratio and a preset number of partial affinity field frameworks and heat map prediction frameworks, including: in the initial position information analysis model framework, a partial affinity field framework (PAF Stage) and a heat map prediction framework (Heatmap Stage) are used to replace the original partial affinity field framework and the original heat map prediction framework. That is: a partial affinity field framework is used to replace the original partial affinity field framework, and a heat map prediction framework is used to replace the original heat map prediction framework.

[0109] In addition, the optimization at the model framework level also includes: optimization at the refining stage level of the partial affinity field framework and the heat map prediction framework. In the initial position information analysis model framework, the original partial affinity field framework and the original heat map prediction framework are refined through multiple refining stages to improve the accuracy of the key points of the analysis model. However, the original partial affinity field framework and the original heat map prediction framework are refined separately, but at the end of each refining stage, the refined results of the original partial affinity field framework and the refined results of the original heat map prediction framework are merged together as input data for the next refining stage. In this embodiment, the partial affinity field framework and the heat map prediction framework are not refined separately, but a cascade method is used to optimize the partial affinity field framework first, and then optimize the heat map prediction framework. This can reduce the amount of parameters and improve the accuracy at the same time. Because while optimizing the partial affinity field framework, the accuracy of the heat map prediction framework will also be improved.

[0110] like Figure 3 As shown in FIG. , it is a schematic diagram of the location information analysis model after optimization in this embodiment. Figure 3 The location information analysis model shown in the figure contains three consecutive 3x3 convolution kernels, a partial affinity field framework (PAF Stage) and a heatmap prediction framework (Heatmap Stagge). Since the model runs on the mobile terminal, it is optimized into a smaller model, and the number of channels in the optimized model is reduced, which reduces the amount of calculation of the model. When the optimized small model running on the mobile terminal processes the image to obtain the key points of the human body, it can achieve the same calculation accuracy as the large model running on the server.

[0111] Secondly, the optimization at the post-processing level includes: when the processed image is in single-person mode, that is, when the number of target objects in the video frame to be analyzed is one, the maximum value of the heat map output by the heat map prediction framework in the location information analysis model framework is extracted by taking a 3x3 pixel neighborhood. At the same time, when the number of target objects in the input image of the location information analysis model is multiple, for example, when the input image contains multiple people, a bilinear interpolation mechanism is used to extract points corresponding to some affinity fields on the input image.

[0112] When using the position information analysis model of this embodiment, an image is input. First, the image is converted into high-dimensional features through the backbone network. Then, the heat map prediction framework is used to predict the heat map of each skeletal key point of the human body in the image. The (x, y) coordinates of the local maximum value on the heat map are selected as candidate key points, and then the x-axis PAF value and y-axis PAF value between each pair of candidate key points are predicted through the partial affinity field framework (the PAF value represents the confidence level of whether there is a connection between the key point pairs). Five positions are sampled between each pair of key points (x1, y1) and (x2, y2), and the PAF values ​​of these five positions are obtained on the PAF graph. If the PAF values ​​of these five positions are all greater than a certain specified threshold, it is considered that the pair of key points is established and belongs to the same person. Finally, the torsos that have been combined into key point pairs can be combined into a complete person. The above process introduces how to use when the processed image contains multiple people. Figure 3 The analysis model shown obtains the key point location information of a certain person.

[0113] By optimizing the initial position information analysis model at the model framework level and optimizing it at the post-processing level, the position information analysis model in this embodiment can be obtained. After obtaining the position information analysis model, the position information analysis model can be used to analyze multiple video frames. The position information analysis model is the position information analysis model running on the mobile terminal.

[0114] As analyzing multiple video frames through the position information analysis model to obtain the position change information of multiple designated parts of the target object, it can be: first, each video frame in the multiple video frames is used as input data of the position information analysis model to obtain the position information of the multiple designated parts of the target object in each video frame in the multiple video frames. Then, based on the position information of the multiple designated parts of the target object in each video frame in the multiple video frames, the position change information of the multiple designated parts of the target object is obtained.

[0115] Specifically, based on the position information of multiple designated parts of the target object in each of the multiple video frames, the position change information of multiple designated parts of the target object is obtained, which can be: first, based on the position information of multiple designated parts of the target object in each of the multiple video frames, the position information of each designated part of the target object in the multiple video frames is obtained respectively. Then, based on the position information of each designated part, the position change information of each designated part in the multiple video frames is obtained. Finally, based on the position change information of each designated part in the multiple video frames, the position change information of multiple designated parts of the target object is obtained.

[0116] It should be noted that before using multiple video frames as input data of the position information analysis model to obtain the position information of multiple specified parts of the target object in each of the multiple video frames, each of the multiple video frames is compressed according to the requirement conditions of the input image data scale of the position information analysis model to obtain multiple compressed video frames.

[0117] After obtaining a plurality of compressed video frames, the compressed plurality of video frames are used as input data of a position information analysis model to obtain position information of a plurality of designated parts of a target object in each of the plurality of video frames.

[0118] In step S203, if the position change information of the plurality of designated parts of the target object meets the position change condition, it is determined that the target object has performed a preset movement, and counting processing is performed on the preset movement performed by the target object.

[0119] After obtaining the position change information of the multiple designated parts of the target object in step S203, it is determined whether the position change information of the multiple designated parts of the target object meets the position change condition. The position change condition refers to a pre-set standard condition for determining whether the object performs a preset movement based on the position change information of the multiple designated parts of the target object.

[0120] If the position change information of multiple specified parts of the target object meets the position change condition, it is determined that the target object has performed a preset movement, and the preset movement performed by the target object is counted. At the same time, it is determined whether the current time information has exceeded the counting time condition of the preset movement; if the current time information has exceeded the counting time condition of the preset movement, the real-time acquisition of the video frame to be analyzed is stopped, and the counting result of the preset movement performed by the target object is obtained, that is, the counting result for the preset movement. For example: when the preset movement is rope skipping, the counting result obtained for the preset movement is the number of rope skipping by the target user. If the current time information does not exceed the counting time condition of the preset movement, it is continued to be determined whether the position change information of multiple specified parts of the target object in the next video frame to be analyzed meets the position change condition.

[0121] Of course, if the position change information of multiple specified parts of the target object does not meet the position change conditions of the object parts of the preset movement, it can be determined that the target object has not performed the preset movement, and when the current time information does not exceed the counting time conditions of the preset movement, continue to judge whether the position change information of multiple specified parts of the target object in the next video frame to be analyzed meets the position change conditions.

[0122] Specifically, the position change information of the multiple designated parts of the target object meeting the position change condition may mean that the difference between the local maximum value of the coordinates and the local minimum value of the coordinates corresponding to the position of each designated part in the multiple designated parts matches the position change condition. In the scenario embodiment, it has been described in detail to establish a coordinate system and obtain the coordinates corresponding to the position of each designated part in the multiple designated parts of the target object. Based on multiple video frames, a coordinate curve graph corresponding to the position of each designated part can be obtained, and then the local maximum value of the coordinates and the local minimum value of the coordinates corresponding to the position of each designated part in the multiple designated parts can be obtained.

[0123] After obtaining the counting result of the preset motion of the target object, the counting result of the preset motion of the target object is displayed to the target object, and the counting result is stored.

[0124] Of course, the above counting result can be considered as a kind of motion degree information, and other forms of motion degree information can also be stored or provided to the user for the user to analyze the preset motion of the target object. The motion degree information is introduced in detail in the scenario embodiment, and the relevant parts can refer to the scenario embodiment, which will not be repeated here.

[0125] The information processing method of this embodiment obtains the video frame to be analyzed in real time, and the video frame to be analyzed contains the target object; analyzes at least two video frames to be analyzed by the position information analysis model to obtain the position change information of multiple specified parts of the target object; if the position change information of multiple specified parts of the target object meets the position change condition, it is determined that the target object has performed a preset movement, and the preset movement of the target object is counted. This counting method avoids manual counting, and the counting result is more accurate; at the same time, since the video frame to be analyzed can be analyzed only by the position information analysis model to obtain the position information of multiple specified parts of the target object in the video frame to be analyzed, it provides a basis for subsequent determination of whether the target object has performed a preset movement and counting, and compared with the counting method of the prior art, this counting method is more convenient.

[0126] Second embodiment

[0127] In the above-mentioned first embodiment, an information processing method is provided, and the information processing method mainly uses a location information analysis model to perform information processing. Correspondingly, the second embodiment of the present application provides a method for obtaining a location information analysis model. Since the method for obtaining the location information analysis model has been described in detail in the first embodiment, please refer to the description of the first embodiment for the relevant points, and no further description will be given here.

[0128] This embodiment provides a method for obtaining a location information analysis model. Figure 4, which is a flowchart of a method for obtaining a location information analysis model provided in the second embodiment of the present application.

[0129] In step S401, a key point position information analysis model running on the mobile terminal is obtained; and the key point position information analysis model running on the mobile terminal is used as an initial position information analysis model.

[0130] In this embodiment, the key point position information analysis model running on the mobile terminal runs on the mobile terminal, and can obtain the key point position information of the target object contained in the image according to the input image.

[0131] In step S402, the original convolution kernel in the initial position information analysis model framework is replaced with a convolution kernel of a preset size and a preset number, and the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework are replaced with a partial affinity field framework and a heat map prediction framework with a preset ratio and a preset number to obtain the target position information analysis model.

[0132] Third embodiment

[0133] Corresponding to the information processing method application scenario embodiment provided by the present application and the information processing method provided by the first embodiment, the third embodiment of the present application also provides an information processing device. Since the device embodiment is basically similar to the application scenario embodiment and the first embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the application scenario embodiment and the first embodiment. The device embodiment described below is only illustrative.

[0134] Please refer to Figure 5 , which is a schematic diagram of an information processing device provided in the third embodiment of the present application.

[0135] The information processing device comprises:

[0136] The video frame obtaining unit 501 is used to obtain the video frame to be analyzed in real time, wherein the video frame to be analyzed contains the target object;

[0137] The position change information obtaining unit 502 is used to analyze the video frame to be analyzed by using a position information analysis model to obtain position change information of multiple specified parts of the target object; the position information analysis model is used to obtain position information of multiple specified parts of the target object in the video frame to be analyzed according to the video frame to be analyzed;

[0138] The counting processing unit 503 is configured to determine that the target object has performed a preset movement if the position change information of the multiple designated parts of the target object meets the position change condition, and perform counting processing on the preset movement performed by the target object.

[0139] Optionally, the position change information obtaining unit is specifically used to:

[0140] Using each of the at least two video frames as input data of the position information analysis model, obtaining position information of multiple designated parts of the target object in each of the at least two video frames;

[0141] Based on the position information of the multiple designated parts of the target object in each video frame, the position change information of the multiple designated parts of the target object is obtained.

[0142] Optionally, the location information analysis model is set on the mobile terminal.

[0143] Optionally, the position change information of the multiple designated parts of the target object satisfies the position change condition, including:

[0144] The difference between the local maximum value of coordinates and the local minimum value of coordinates corresponding to the position of each designated part of the multiple designated parts matches the position change condition.

[0145] Optionally, the device further includes a first judgment unit, the first judgment unit being configured to judge whether the position change information of the plurality of designated parts of the target object satisfies a position change condition;

[0146] If the position change information of multiple specified parts of the target object does not meet the position change condition, it is determined that the target object has not performed the preset movement, and it is judged whether the position change information of multiple specified parts of the target object in the next video frame to be analyzed meets the position change condition.

[0147] Optionally, the device also includes a second judgment unit, which is used to judge whether the current time information has exceeded the counting time condition of the preset motion; if the current time information has exceeded the counting time condition of the preset motion, stop obtaining the video frame to be analyzed in real time, and obtain the counting result for the preset motion.

[0148] Optionally, the device further includes a display and storage unit, which is used to display the counting result for the preset motion to the target object and store the calculation result.

[0149] Optionally, the device further comprises: an exercise degree information obtaining unit;

[0150] The exercise degree information obtaining unit is used to:

[0151] Obtaining information about the degree of movement of the target object performing the preset movement;

[0152] The movement degree information is provided to a user so that the user can analyze the preset movement performed by the target object.

[0153] Optionally, the video frame obtaining unit is specifically used to:

[0154] Using a mobile camera to take real-time video of the target object to obtain a video file containing the target object;

[0155] According to the video file containing the target object, a video frame to be analyzed is obtained.

[0156] Optionally, the device further comprises a compression unit, configured to compress each of the at least two video frames according to a requirement of an input image data volume scale of the position information analysis model to obtain at least two compressed video frames;

[0157] The position change information obtaining unit is specifically used for:

[0158] The at least two compressed video frames are respectively used as input data of the position information analysis model to obtain position information of multiple designated parts of the target object in each of the at least two video frames.

[0159] Optionally, the device further includes an analysis model obtaining unit, wherein the analysis model obtaining unit is specifically configured to:

[0160] Obtaining a key point position information analysis model running on a mobile terminal; wherein the key point position information analysis model running on a mobile terminal runs on a mobile terminal and can obtain key point position information of a target object contained in an image according to an input image;

[0161] Using the key point position information analysis model running on the mobile terminal as the initial position information analysis model;

[0162] The original convolution kernels in the initial position information analysis model framework are replaced with convolution kernels of preset size and preset number, and the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework are replaced with partial affinity field frameworks and heat map prediction frameworks of preset ratio and preset number to obtain an optimized position information analysis model, and the optimized position information analysis model is used as the position information analysis model.

[0163] Optionally, the analysis model obtaining unit is specifically used to:

[0164] The original 7x7 convolution kernel in the initial position information analysis model framework is replaced by three consecutive 3x3 convolution kernels;

[0165] In the initial position information analysis model framework, a partial affinity field framework is used to replace the original partial affinity field framework, and a heat map prediction framework is used to replace the original heat map prediction framework.

[0166] Optionally, the device also includes a first optimization unit, which is used to extract pixel points by taking a 3x3 pixel neighborhood for the maximum value of the heat map output by the heat map prediction framework in the optimized location information analysis model framework.

[0167] Optionally, the device further includes a second optimization unit, wherein the second optimization unit is configured to extract points corresponding to some affinity fields on the input image by using a bilinear interpolation mechanism if the input image contains multiple target objects.

[0168] Fourth embodiment

[0169] Corresponding to the method for obtaining the location information analysis model provided in the second embodiment of the present application, the fourth embodiment of the present application also provides a device for obtaining the location information analysis model. Since the device embodiment is basically similar to the second embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the second embodiment. The device embodiment described below is only exemplary.

[0170] Please refer to Figure 6 , which is a schematic diagram of a device for obtaining a location information analysis model provided in the fourth embodiment of the present application.

[0171] The device for obtaining the position information analysis model includes:

[0172] The initial analysis model obtaining unit 601 is used to obtain a human body key point position information analysis model running on a mobile terminal; wherein the human body key point position information analysis model running on a mobile terminal runs on a mobile terminal and can obtain key point position information of a human body contained in an image according to an input image; the human body key point position information analysis model running on a mobile terminal is used as an initial position information analysis model;

[0173] The target position information analysis model acquisition unit 602 is used to replace the original convolution kernel in the initial position information analysis model framework with a convolution kernel of a preset size and a preset number, and replace the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with a preset ratio and a preset number of partial affinity field frameworks and heat map prediction frameworks to obtain the target position information analysis model.

[0174] Fifth embodiment

[0175] In the above-mentioned first embodiment, an information processing method is provided. The information processing method mainly uses a position information analysis model to perform information processing, and its application scenario can be used to determine the amount of exercise. The fifth embodiment of the present application provides a method for determining the amount of exercise. Since part of the content of this method has been described in detail in the first embodiment, please refer to the description of the first embodiment for relevant matters, which will not be repeated here.

[0176] This embodiment provides a method for determining an amount of exercise, including:

[0177] First, a video frame containing a moving object captured by a mobile terminal is received.

[0178] Afterwards, the video frames containing the moving objects are analyzed in real time to obtain the position change information of multiple specified parts of the moving objects. The real-time analysis is undertaken by a position information analysis model set in the mobile terminal, which is used to obtain the position change information of multiple specified parts of the target object in the video frame based on the video frames containing the moving objects collected by the mobile terminal.

[0179] If the position change information of multiple designated parts of the moving object meets the position change condition, counting processing is performed on the movement of the moving object to obtain a counting result.

[0180] In a further solution of the method, the motion type information set by the user can also be obtained, and the position information analysis model is set based on the analysis parameters corresponding to the motion type information. In the step of performing real-time analysis on the video frame containing the moving object, the position information analysis model with the analysis parameters set is used for analysis. In this embodiment, the user can set the motion type information related to the motion through the client.

[0181] At the same time, the motion parameters corresponding to the motion type information can also be obtained; after obtaining the counting result, the counting result is processed; and the measurement result obtained after the data processing is compared with the motion parameter to obtain a comparison result. According to the comparison result, the mobile terminal issues a related motion prompt. For example, the related motion prompt can be whether the heart rate of the exercise subject is in a normal state and the exercise rate during the exercise.

[0182] Sixth embodiment

[0183] Corresponding to part of the content of the scenario embodiment, this embodiment provides an information processing method, including the following steps:

[0184] First: receiving a video frame containing a moving object captured by a fitness device.

[0185] Afterwards, the video frames containing the moving objects are analyzed in real time to obtain the position change information of the specified parts of the moving objects; the real-time analysis is undertaken by the position information analysis model set in the fitness equipment, and the position information analysis model is used to obtain the position change information of the specified parts of the target object in the video frames based on the video frames containing the moving objects collected by the fitness equipment.

[0186] If the position change information of the designated part of the moving object meets the position change condition, it is finally determined that the target object has performed a preset movement, and counting processing is performed on the preset movement performed by the target object.

[0187] Seventh embodiment

[0188] Corresponding to the methods provided in the first to second embodiments and the fifth to sixth embodiments of the present application, the seventh embodiment of the present application further provides an electronic device.

[0189] As shown in 7, Figure 7 This is a schematic diagram of an electronic device provided in the seventh embodiment of the present application.

[0190] The electronic device includes:

[0191] Processor 701;

[0192] The memory 702 is used to store a computer program. The computer program is executed by the processor to perform the methods described in the first embodiment to the second embodiment and the fifth embodiment to the sixth embodiment.

[0193] Eighth embodiment

[0194] Corresponding to the methods provided in the first to second embodiments and the fifth to sixth embodiments of the present application, the eighth embodiment of the present application further provides a computer storage medium, wherein the computer storage medium stores a computer program, which is executed by a processor to execute the methods described in the first to second embodiments and the fifth to sixth embodiments.

[0195] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

[0196] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory. The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0197] 1. Computer readable media includes permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in this article, computer readable media does not include non-transitory computer readable storage media (non-transitorycomputer readable storage media), such as modulated data signals and carrier waves.

[0198] 2. Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

Claims

1. An information processing method, characterized in that: include: Obtaining a video frame to be analyzed in real time, wherein the video frame to be analyzed contains a target object; Analyzing at least two of the video frames to be analyzed by a position information analysis model to obtain position change information of multiple specified parts of the target object; the position information analysis model is used to obtain position information of multiple specified parts of the target object in the video frames to be analyzed according to the video frames to be analyzed; wherein the position information analysis model is obtained by: Obtain a key point position information analysis model running on a mobile terminal; wherein the key point position information analysis model running on the mobile terminal runs on the mobile terminal and can obtain the key point position information of the target object contained in the image according to the input image; use the key point position information analysis model running on the mobile terminal as the initial position information analysis model; replace the original convolution kernel in the initial position information analysis model framework with a convolution kernel of a preset size and a preset number, and replace the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with a preset ratio and a preset number of partial affinity field frameworks and heat map prediction frameworks to obtain an optimized position information analysis model, and use the optimized position information analysis model as the position information analysis model; If the position change information of the multiple designated parts of the target object meets the position change condition, it is determined that the target object has performed a preset movement, and counting processing is performed on the preset movement performed by the target object.

2. The method according to claim 1, characterized in that The step of analyzing at least two of the to-be-analyzed video frames using a position information analysis model to obtain position change information of multiple designated parts of the target object includes: Using each of the at least two video frames as input data of the position information analysis model, obtaining position information of multiple designated parts of the target object in each of the at least two video frames; Based on the position information of the multiple designated parts of the target object in each video frame, the position change information of the multiple designated parts of the target object is obtained.

3. The method according to claim 1, characterized in that The location information analysis model is set on the mobile terminal.

4. The method according to claim 1, characterized in that: The position change information of the multiple designated parts of the target object meets the position change condition, including: The difference between the local maximum value of coordinates and the local minimum value of coordinates corresponding to the position of each designated part of the multiple designated parts matches the position change condition.

5. The method according to claim 1, characterized in that Also includes: Determining whether the position change information of the plurality of designated parts of the target object meets the position change condition; If the position change information of multiple specified parts of the target object does not meet the position change condition, it is determined that the target object has not performed the preset movement, and it is judged whether the position change information of multiple specified parts of the target object in the next video frame to be analyzed meets the position change condition.

6. The method according to claim 1, characterized in that Also includes: Determine whether the current time information has exceeded the preset motion counting time condition; If the current time information exceeds the counting time condition of the preset motion, the real-time acquisition of the video frame to be analyzed is stopped, and the counting result for the preset motion is obtained.

7. The method according to claim 6, characterized in that Also includes: The counting result for the preset motion is presented to the target object, and the counting result is stored.

8. The method according to claim 1, characterized in that: Also includes: Obtaining information about the degree of movement of the target object performing the preset movement; The movement degree information is provided to a user so that the user can analyze the preset movement performed by the target object.

9. The method according to claim 1, characterized in that: The real-time acquisition of the video frame to be analyzed includes: Using a mobile camera to take real-time video of the target object to obtain a video file containing the target object; According to the video file containing the target object, a video frame to be analyzed is obtained.

10. The method according to claim 2, characterized in that The step of using each of the at least two video frames as input data of the position information analysis model to obtain position information of multiple designated parts of the target object in each of the at least two video frames includes: Compressing each of the at least two video frames according to the input image data volume scale requirement of the position information analysis model to obtain at least two compressed video frames; The at least two compressed video frames are respectively used as input data of the position information analysis model to obtain position information of multiple designated parts of the target object in each of the at least two video frames.

11. The method according to claim 1, characterized in that: The replacing the original convolution kernel in the initial position information analysis model framework with a convolution kernel of a preset size and a preset number includes: The original 7x7 convolution kernel in the initial position information analysis model framework is replaced by three consecutive 3x3 convolution kernels; The replacing the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with the partial affinity field framework and the heat map prediction framework with a preset ratio and a preset number includes: In the initial position information analysis model framework, a partial affinity field framework is used to replace the original partial affinity field framework, and a heat map prediction framework is used to replace the original heat map prediction framework.

12. The method according to claim 1, characterized in that Also includes: For the maximum value of the heat map output by the heat map prediction framework in the optimized position information analysis model framework, a 3x3 pixel neighborhood is taken to perform pixel point refinement.

13. The method according to claim 1, characterized in that If the input image contains multiple target objects, a bilinear interpolation mechanism is used to extract points corresponding to some affinity fields on the input image.

14. An information processing device, characterized in that: include: A video frame obtaining unit, used for obtaining a video frame to be analyzed in real time, wherein the video frame to be analyzed contains a target object; A position change information obtaining unit, used for analyzing at least two of the video frames to be analyzed through a position information analysis model to obtain position change information of multiple specified parts of the target object; the position information analysis model is used to obtain the position information of multiple specified parts of the target object in the video frame to be analyzed according to the video frame to be analyzed; wherein the position information analysis model is obtained in the following manner: obtaining a key point position information analysis model running on a mobile terminal; wherein the key point position information analysis model running on the mobile terminal runs on the mobile terminal and can obtain the key point position information of the target object contained in the image according to the input image; using the key point position information analysis model running on the mobile terminal as an initial position information analysis model; replacing the original convolution kernel in the initial position information analysis model framework with a convolution kernel of a preset size and a preset number, and replacing the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with partial affinity field frameworks and heat map prediction frameworks with preset ratios and preset numbers to obtain an optimized position information analysis model, and using the optimized position information analysis model as the position information analysis model; A counting processing unit is used to determine that the target object has performed a preset movement if the position change information of multiple designated parts of the target object meets the position change condition, and to perform counting processing on the preset movement performed by the target object.

15. A method for determining the amount of exercise, characterized in that: include: Receiving a video frame containing a moving object captured by a mobile terminal; The video frame containing the moving object is analyzed in real time to obtain position change information of multiple specified parts of the moving object; the real-time analysis is undertaken by a position information analysis model set in the mobile terminal, and the position information analysis model is used to obtain position change information of multiple specified parts of the target object in the video frame according to the video frame containing the moving object collected by the mobile terminal; wherein the position information analysis model is obtained by: Obtain a key point position information analysis model running on a mobile terminal; wherein the key point position information analysis model running on the mobile terminal runs on the mobile terminal and can obtain the key point position information of the target object contained in the image according to the input image; use the key point position information analysis model running on the mobile terminal as the initial position information analysis model; replace the original convolution kernel in the initial position information analysis model framework with a convolution kernel of a preset size and a preset number, and replace the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with a preset ratio and a preset number of partial affinity field frameworks and heat map prediction frameworks to obtain an optimized position information analysis model, and use the optimized position information analysis model as the position information analysis model; If the position change information of the multiple designated parts of the moving object meets the position change condition, counting processing is performed on the movement of the moving object to obtain a counting result.

16. The method according to claim 15, characterized in that Also includes: Get the sports type information set by the user; The position information analysis model is set using analysis parameters corresponding to the motion type information; in the step of performing real-time analysis on the video frame containing the motion object, the position information analysis model with the analysis parameters set is used for analysis.

17. The method according to claim 16, characterized in that Also includes: Obtaining motion parameters corresponding to the motion type information; After obtaining the counting result, performing data processing on the counting result; and comparing the measurement result obtained after data processing with the motion parameter to obtain a comparison result; According to the comparison result, relevant movement prompts are issued through the mobile terminal.

18. An information processing method, characterized in that: include: receiving a video frame containing a moving object captured by a fitness device; The video frame containing the moving object is analyzed in real time to obtain the position change information of the designated part of the moving object; the real-time analysis is undertaken by a position information analysis model set in the fitness equipment, and the position information analysis model is used to obtain the position change information of the designated part of the target object in the video frame according to the video frame containing the moving object collected by the fitness equipment; wherein the position information analysis model is obtained by: Obtain a key point position information analysis model running on a fitness device; wherein the key point position information analysis model running on the fitness device runs in the fitness device and can obtain key point position information of a target object contained in an image according to an input image; use the key point position information analysis model running on the fitness device as an initial position information analysis model; replace the original convolution kernel in the initial position information analysis model framework with a convolution kernel of a preset size and a preset number, and replace the original partial affinity field framework and the original heat map prediction framework in the initial position information analysis model framework with a partial affinity field framework and a heat map prediction framework of a preset ratio and a preset number, obtain an optimized position information analysis model, and use the optimized position information analysis model as the position information analysis model; If the position change information of the designated part of the moving object meets the position change condition, it is determined that the target object has performed a preset movement, and counting processing is performed on the preset movement performed by the target object.

19. An electronic device, characterized in that: The electronic device comprises: a processor; and a memory for storing a computer program, wherein the computer program is run by the processor to execute the method according to any one of claims 1 to 13 and claims 15 to 18.

20. A computer storage medium, characterized in that The computer storage medium stores a computer program, which is executed by a processor to perform the method according to any one of claims 1 to 13 and claims 15 to 18.

Images