Sampling device for three-dimensional motion data of IP image

Abstract

The invention discloses an IP image three-dimensional motion data sampling device, and relates to the technical field of sampling devices.The IP image three-dimensional motion data sampling device comprises a sensor screening module, a sensor deployment module and an animation generation module, and an adaptive actor and a body part deployment set thereof are determined based on a character model of a specified IP image and a three-dimensional data acquisition environment; according to the specified IP image, correspondingly adapting to the body part deployment set of the actor to implement sensor deployment; and converting the deployment set of the body parts of the actors correspondingly adapted to the specified IP images into a three-dimensional motion animation.

Description

Technical Field

[0001] The present invention relates to the technical field of sampling devices, and in particular to a sampling device for three-dimensional motion data of an IP image. Background Art

[0002] With the rapid development of technology, 3D motion data sampling has found widespread application in a wide range of fields, including film and television entertainment, game development, virtual reality, medical rehabilitation, and sports training. In film and television production, this technology can accurately capture actors' movements, imbuing virtual characters with vivid and realistic dynamic performances. The gaming industry uses it to create more immersive virtual worlds, allowing players to operate more smoothly and naturally. In the medical field, it can be used to monitor patient rehabilitation training, assisting doctors in developing more effective treatment plans. In sports training, it can help coaches analyze athletes' movements and improve training effectiveness.

[0003] Traditional devices face numerous technical bottlenecks in capturing 3D motion data for IP characters. Existing solutions often lack a scientific actor adaptation mechanism, resulting in a low degree of overlap between the actor's body model and the IP character model, making it difficult to accurately convey the IP character's motion characteristics. During sensor deployment, due to a lack of targeted screening based on the IP character type and motion requirements, problems such as poor sensor-body compatibility and insufficient wear comfort affect data collection stability. Furthermore, historical performance data is not effectively quantified and utilized, and the assessment of motion interpretation, fluidity, and facial emotion matching lacks a systematic approach. This makes it difficult to accurately reproduce the dynamic characteristics of the IP character when generating 3D motion animations. Overall sampling efficiency and data quality urgently need to be improved. Summary of the Invention

[0004] The purpose of the present invention is to provide a sampling device for three-dimensional motion data of an IP image, which solves the problems existing in the background technology.

[0005] To solve the above technical problems, the present invention adopts the following technical solution: The present invention provides a device for sampling three-dimensional motion data of an IP image, characterized by comprising:

[0006] a sensor screening module, which determines an actor suitable for the specified IP image based on the character model of the specified IP image and the three-dimensional data acquisition environment, and determines a body part deployment set corresponding to the actor suitable for the specified IP image, wherein the body part deployment set includes sensor types corresponding to several body parts;

[0007] Preferably, the specific method for determining the actor suitable for the designated IP image is as follows:

[0008] Extract body models of several actors and models of designated IP images from the database;

[0009] Adjusting the body models of several actors at different proportions to obtain the adjusted body models of the several actors, evaluating the degree of overlap between the adjusted body models of the several actors and the designated IP image, and selecting the maximum overlap as the first fitness degree RY between the several actors and the designated IP image, where Y represents the number of the several actors;

[0010] Extracting historical performance data of several actors from the database to analyze the second compatibility NY of the several actors with the designated IP image, where Y represents the number of the several actors;

[0011] Comprehensively analyze the compatibility of several actors with the designated IP image OY=λ1*(RY)+λ2*(NY).

[0012] Preferably, the specific method for evaluating the degree of coincidence between the adjusted body models of the actors and the designated IP image is as follows:

[0013] Step 1: Voxelize the model;

[0014] Convert the specified IP image model into a voxel grid with a fixed grid resolution of 1mm*1mm*1mm to obtain a binary voxel grid G ip , where G ip (x) = 1 means that the voxel corresponding to the identifier x is inside the specified IP image model, G ip (x)=0 means that the voxel corresponding to the identifier x is not inside the specified IP image model, where x is the globally unique identifier of the voxel unit in the voxel grid;

[0015] The actor's body model is converted into a voxel grid with a fixed grid resolution of 1mm*1mm*1mm to obtain a binary voxel grid G body , where G body (x) = 1 means that the voxel corresponding to the identifier x is inside the actor's body model, G body (x)=0 means that the voxel corresponding to the identifier x is not inside the actor's body model, where x is the globally unique identifier of the voxel unit in the voxel grid;

[0016] Step 2: Calculate the basic volume;

[0017] The physical volume of a single voxel is the voxel volume v l =1mm 3 ;

[0018] Specify the volume V of the IP image ip =N ip *v l , where N ip Represents G ip the number of internal voxels in ;

[0019] The volume V of the actor's body model body =N body *v l , where N body Represents G body the number of internal voxels in ;

[0020] Step 3, calculate the overlap volume;

[0021] In the same voxel grid, check whether the voxel belongs to both models and calculate the coincidence judgment value for each voxel: G ol When (x) = 1, it means that voxel x is inside the specified IP image model and inside the actor's body model, ∧ is a logical symbol and;

[0022] The total number N of voxels with a statistical coincidence judgment value of 1 ol , calculate the overlap volume V of the specified IP image model and the actor's body model ol =N ol *v l ;

[0023] Step 4: Calculate the degree of overlap;

[0024] Calculate the overlap between the specified IP image model and the actor's body model

[0025] Perform the above steps 1-4 on the adjusted body models of several actors and the designated IP image model, calculate the degree of overlap between the adjusted body models of several actors and the designated IP image model, and select the maximum overlap as the first degree of fit between the several actors and the designated IP image, thereby obtaining the first degree of fit RY between the several actors and the designated IP image, where Y represents the number of the actors.

[0026] Preferably, the specific method for analyzing the second compatibility between a plurality of actors and a designated IP image is as follows:

[0027] Step 1: Historical data extraction: extracting historical performance data of several actors from the database, wherein the historical performance data includes the three-dimensional motion animation of the IP image corresponding to each historical performance;

[0028] Step 2: Action interpretation and quantification:

[0029] a. Get the required action set A of the specified IP image stored in the database req ={a req1 ,a req2 ,...,a reqn}, where n is the total number of required actions;

[0030] b. Extracting and interpreting action model A from 3D motion animation act ={a act1 ,a act2 ,...,a actm}, where m is the total number of actual deductive actions;

[0031] c. Calculate the standardized status of each required action:

[0032] for i is the required action index, if there is a reqi ∈A act , j is the index of the deductive action, satisfying sim(a actj ,a reqi )≥θ, where θ is the preset overlap threshold, sim is the action similarity function, then mark a reqi To standardize actions;

[0033] d. Quantified action interpretation is:

[0034] where N standard is the number of standardized actions;

[0035] Step 3: Quantify movement fluency:

[0036] e. Convert the three-dimensional motion animation into a time sequence action sequence S = {(t k ,a k ,d k )|k=1,2,...,K}, where k is the frame index, K is the total number of frames, t k is the action start time, a k is the action identifier, d k The duration of the action;

[0037] f. Quantify the movement fluency as:

[0038] Q fluency =f(S), where function f is generated based on the continuity of the action transition sequence and the stability of the duration;

[0039] Step 4: Calculate facial emotion matching coefficient:

[0040] g. Get the facial emotion requirement set E of the specified IP image req ={e req1 ,e req2 ,...,e reqp}, where p is the total number of demand emotion types;

[0041] h. Extracting and generating facial emotion sequence E from the three-dimensional motion animation act ={e act1,e act2 ,...,e actq}, where q is the number of emotion segments actually captured;

[0042] i. Process the data and calculate the sentiment matching coefficient:

[0043] Among them, sim e It is the emotion similarity function, which returns the similarity in the interval [0,1]; sim e =1 means the emotion matches perfectly, sim e =0 indicates a complete mismatch in sentiment, and intermediate values ​​indicate a partial match;

[0044] Step 5: Second fitness comprehensive quantification

[0045] Generate the second fitness:

[0046] Among them, w1 is the action standard weight, w2 is the action smoothness weight, and w3 is the facial emotion weight;

[0047] Based on the three-dimensional motion animation of the IP image of each historical interpretation of several actors, execute steps 1-5 to calculate the second fitness of each historical interpretation of several actors and the specified IP image, and perform average processing on them to obtain the second fitness of several actors and the specified IP image NY, where Y represents the number of several actors.

[0048] Preferably, the specific method for determining the body part deployment set of the actor corresponding to the designated IP image is as follows:

[0049] Based on the model of the specified IP image, combined with the comparison table of various IP images and basic parts stored in the database, several basic torso parts and several basic facial parts of the specified IP image are screened out;

[0050] According to the type of the specified IP image, all IP images of the same type are filtered from the IP image library, and the common parts of all IP images of the same type are counted;

[0051] Based on the required action set, the required attachment parts for each action are determined according to a required attachment part mapping table pre-stored in the database, and a number of required torso parts and a number of required facial parts of the designated IP image are generated accordingly; the required attachment part mapping table includes data on the correspondence between action types and required attachment parts;

[0052] According to the specified IP image corresponding to the several special required torso parts and several special required facial parts of the actor, and obtaining the collection environment of the specified IP image corresponding to the actor, the sensor types corresponding to the several torso parts and the sensor types corresponding to the several facial parts are screened.

[0053] Preferably, the specific method of screening the sensor types corresponding to the plurality of torso parts and the sensor types corresponding to the plurality of facial parts is as follows:

[0054] Rule 1: Ensure the compatibility of the sensor with the actor's body: The selected sensor must be worn on the actor's designated body part, be consistent with the model's mounting position on that part, and meet the preset wearing comfort requirements;

[0055] Rule 2: Evaluate the historical performance of sensors: Analyze the stability and accuracy of the sensor's historical data collection in various environmental scenarios;

[0056] Based on the first and second screening rules, several sensor types in the database are traversed to obtain several torso-specific and facial-specific sensors.

[0057] Preferably, the method of ensuring the compatibility of the sensor with the actor's body is as follows:

[0058] l. Deployment location determination: Based on the specified IP image corresponding to the adaptation actor's body parts deployment set, determine the designated IP image to be deployed sensors corresponding to the adaptation actor's several torso parts and several facial parts;

[0059] m. Type matching: Based on the multiple body parts and multiple facial parts, select sensor types that meet the following performance indicators as candidates:

[0060] n1. Wearing comfort: Meets the preset comfort threshold;

[0061] n2.Flexibility of movement: Ensure the flexibility of joint movement;

[0062] n3. Freedom of movement: supports high degree of freedom of body movements.

[0063] Preferably, the historical performance of the sensor is evaluated by:

[0064] Step 1: Extract performance data: Extract performance data of various sensors under different environments from the database. The performance data includes:

[0065] p1. Accuracy deviation rate:

[0066] p2. Failure rate:

[0067] p3. Transmission rate: v

[0068] Step 2: Calculate performance effect value: Generate performance effect value E of each sensor under each environmental label and type of environment based on performance data evaluation;

[0069] Step 3: Current environment adaptation screening:

[0070] q1. Get the actual collection environment of the specified IP image;

[0071] q2. Based on the performance value obtained in step 2, extract the performance value E of each sensor in the actual acquisition environment env ;

[0072] The evaluation dimensions of the performance effect value E include:

[0073] r1, accuracy deviation rate;

[0074] r2, failure rate;

[0075] r3. Transmission rate.

[0076] A sensor deployment module is used to implement sensor deployment according to a deployment set of body parts of an adapted actor corresponding to a specified IP image;

[0077] The animation generation module is used to convert the body part deployment set of the actor corresponding to the specified IP image into a three-dimensional motion animation.

[0078] Preferably, the sensor deployment is implemented according to the deployment set of body parts of the adapted actor corresponding to the designated IP image, and the specific method is:

[0079] s. Location Mapping: Locating the physical locations of the torso and facial areas where sensors are to be deployed based on the determined set of body part deployments;

[0080] t. Type matching deployment: Deploy sensors from the determined set of compatible sensor types to the corresponding body parts according to the following rules:

[0081] u1. The trunk position sensor is installed at the joint axis or muscle movement sensitive area;

[0082] u2. Facial sensors fit the distribution of micro-expression muscles;

[0083] v. Calibration verification: Power on the sensor, collect test motion data, and verify whether the matching degree between the sensor output signal and the preset motion model exceeds the threshold.

[0084] The beneficial effects of the present invention are as follows: through multi-module collaborative design, the present invention improves the sampling quality and efficiency of the three-dimensional motion data of the IP image. The sensor screening module accurately determines the suitable actors based on the calculation of the body model overlap and the quantitative analysis of historical interpretation data. The comprehensive evaluation mechanism of the first and second degrees of fitness ensures that the action characteristics of the actors and the IP image are highly consistent. The hierarchical determination method of the body part deployment set combines the IP image type, action requirements and sensor compatibility to achieve scientific deployment of sensors, ensuring the accuracy of data collection and wearing comfort. The multi-dimensional evaluation of the sensor's historical performance can screen out sensor types that are suitable for different collection environments and improve the stability of data collection. The animation generation module uses motion capture technology and key point trajectory modeling to efficiently convert sensor data into three-dimensional motion animation, making the virtual model's movements natural and smooth, comprehensively solving the problems of poor adaptability, low data accuracy, and poor animation generation effects in the existing technology, and providing efficient and reliable technical support for the dynamic presentation of IP images. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0086] Figure 1 This is a schematic diagram of the system structure connection of the present invention. DETAILED DESCRIPTION

[0087] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0088] Reference Figure 1 As shown, the present invention provides a device for sampling three-dimensional motion data of an IP image, comprising:

[0089] a sensor screening module, which determines an actor suitable for the specified IP image based on the character model of the specified IP image and the three-dimensional data acquisition environment, and determines a body part deployment set corresponding to the actor suitable for the specified IP image, wherein the body part deployment set includes sensor types corresponding to several body parts;

[0090] In a specific embodiment of the present invention, the method for determining the suitable actor for a specified IP image is as follows:

[0091] Extract body models of several actors and models of designated IP images from the database;

[0092] Adjusting the body models of several actors at different proportions to obtain the adjusted body models of the several actors, evaluating the degree of overlap between the adjusted body models of the several actors and the designated IP image, and selecting the maximum overlap as the first fitness degree RY between the several actors and the designated IP image, where Y represents the number of the several actors;

[0093] Extracting historical performance data of several actors from the database to analyze the second compatibility NY of the several actors with the designated IP image, where Y represents the number of the several actors;

[0094] Comprehensively analyze the compatibility of several actors with the designated IP image OY = λ1*(RY)+λ2*(NY);

[0095] It should be noted that the degree of compatibility between the actors and the designated IP image is such that λ1 is the first degree of compatibility weight and λ2 is the second degree of compatibility weight.

[0096] In a specific embodiment of the present invention, the specific method for evaluating the degree of overlap between the adjusted body models of several actors and the designated IP image is as follows:

[0097] Step 1: Voxelize the model;

[0098] Convert the specified IP image model into a voxel grid with a fixed grid resolution of 1mm*1mm*1mm to obtain a binary voxel grid G ip , where G ip (x) = 1 means that the voxel corresponding to the identifier x is inside the specified IP image model, G ip (x)=0 means that the voxel corresponding to the identifier x is not inside the specified IP image model, where x is the globally unique identifier of the voxel unit in the voxel grid;

[0099] The actor's body model is converted into a voxel grid with a fixed grid resolution of 1mm*1mm*1mm to obtain a binary voxel grid G body , where G body (x) = 1 means that the voxel corresponding to the identifier x is inside the actor's body model, G body (x)=0 means that the voxel corresponding to the identifier x is not inside the actor's body model, where x is the globally unique identifier of the voxel unit in the voxel grid;

[0100] Step 2: Calculate the basic volume;

[0101] The physical volume of a single voxel is the voxel volume v l=1mm 3 ;

[0102] Specify the volume V of the IP image ip =N ip *v l , where N ip Represents G ip the number of internal voxels in ;

[0103] The volume V of the actor's body model body =N body *v l , where N body Represents G body the number of internal voxels in ;

[0104] Step 3, calculate the overlap volume;

[0105] In the same voxel grid, check whether the voxel belongs to both models and calculate the coincidence judgment value for each voxel: G ol When (x) = 1, it means that voxel x is inside the specified IP image model and inside the actor's body model, ∧ is a logical symbol and;

[0106] The total number N of voxels with a statistical coincidence judgment value of 1 ol , calculate the overlap volume V of the specified IP image model and the actor's body model ol =N ol *v l ;

[0107] Step 4: Calculate the degree of overlap;

[0108] Calculate the overlap between the specified IP image model and the actor's body model

[0109] Perform the above steps 1-4 on the adjusted body models of several actors and the designated IP image model, calculate the degree of overlap between the adjusted body models of several actors and the designated IP image model, and select the maximum overlap as the first degree of fit between the several actors and the designated IP image, thereby obtaining the first degree of fit RY between the several actors and the designated IP image, where Y represents the number of the actors.

[0110] In a specific embodiment of the present invention, the specific method for analyzing the second compatibility between a plurality of actors and a designated IP image is as follows:

[0111] Step 1: Historical data extraction: extracting historical performance data of several actors from the database, wherein the historical performance data includes the three-dimensional motion animation of the IP image corresponding to each historical performance;

[0112] Step 2: Action interpretation and quantification:

[0113] a. Get the required action set A of the specified IP image stored in the database req ={a req1 ,a req2 ,...,a reqn}, where n is the total number of required actions;

[0114] b. Extracting and interpreting action model A from 3D motion animation act ={a act1 ,a act2 ,...,a actm}, where m is the total number of actual deductive actions;

[0115] c. Calculate the standardized status of each required action:

[0116] for i is the required action index, if there is a reqi ∈A act , j is the index of the deductive action, satisfying sim(a actj ,a reqi )≥θ, where θ is the preset overlap threshold, sim is the action similarity function, then mark a reqi To standardize actions;

[0117] It should be noted that the action similarity function is relatively mature and simple in the existing technology, and will not be described in detail here;

[0118] d. Quantified action interpretation is:

[0119] where N standard is the number of standardized actions;

[0120] Step 3: Quantify movement fluency:

[0121] e. Convert the three-dimensional motion animation into a time sequence action sequence S = {(t k ,a k ,d k )|k=1,2,...,K}, where k is the frame index, K is the total number of frames, t k is the action start time, a k is the action identifier, d k The duration of the action;

[0122] f. Quantify the movement fluency as:

[0123] Q fluency =f(S), where function f is generated based on the continuity of the action transition sequence and the stability of the duration;

[0124] It should be noted that the continuity of the action conversion sequence refers to the stability of the action conversion duration; the stability of the duration includes the stability of the duration of each action and the standard duration;

[0125] Step 4: Calculate facial emotion matching coefficient:

[0126] g. Get the facial emotion requirement set E of the specified IP image req ={e req1 ,e req2 ,...,e reqp}, where p is the total number of demand emotion types;

[0127] h. Extracting and generating facial emotion sequence E from the three-dimensional motion animation act ={e act1 ,e act2 ,...,e actq}, where q is the number of emotion segments actually captured;

[0128] i. Process the data and calculate the sentiment matching coefficient:

[0129] Among them, sim e It is the emotion similarity function, which returns the similarity in the interval [0,1]; sim e =1 means the emotion matches perfectly, sim e =0 indicates a complete mismatch in sentiment, and intermediate values ​​indicate a partial match;

[0130] It should be noted that the emotion similarity function is relatively mature and simple in the existing technology, and will not be described in detail here;

[0131] Step 5: Second fitness comprehensive quantification

[0132] Generate the second fitness:

[0133] Among them, w1 is the action standard weight, w2 is the action smoothness weight, and w3 is the facial emotion weight;

[0134] Based on the three-dimensional motion animation of the IP image of each historical interpretation of several actors, execute steps 1-5 to calculate the second fitness of each historical interpretation of several actors and the specified IP image, and perform average processing on them to obtain the second fitness of several actors and the specified IP image NY, where Y represents the number of several actors.

[0135] In a specific embodiment of the present invention, the specific method for determining the body part deployment set of the actor corresponding to the specified IP image is as follows:

[0136] Based on the model of the specified IP image, combined with the comparison table of various IP images and basic parts stored in the database, several basic torso parts and several basic facial parts of the specified IP image are screened out;

[0137] It should be noted that, taking the IP image types of virtual idols, animal anthropomorphism, and mechanical structures as examples, the comparison table of various IP images and basic parts is shown in Table 1;

[0138] Table 1 - Comparison table of various IP images and basic parts

[0139] IP image type Basic torso Basic facial parts Virtual idols Spine, shoulders Eyebrows, eyes, mouth Animal anthropomorphism Spine, tail ears, whiskers, eyelids Mechanical structure Hydraulic transmission shaft, gear set Optical lens, sound generator

[0140] According to the type of the specified IP image, all IP images of the same type are filtered from the IP image library, and the common parts of all IP images of the same type are counted;

[0141] Based on the required action set, the required attachment parts for each action are determined according to a required attachment part mapping table pre-stored in the database, and a number of required torso parts and a number of required facial parts of the designated IP image are generated accordingly; the required attachment part mapping table includes data on the correspondence between action types and required attachment parts;

[0142] It should be noted that, taking the action type as expression and action as an example, the required carrying part mapping table is shown in Table 2;

[0143] Table 2-Required loading location mapping table

[0144]

[0145] According to the specified IP image corresponding to the several special required torso parts and several special required facial parts of the actor, and obtaining the collection environment of the specified IP image corresponding to the actor, the sensor types corresponding to the several torso parts and the sensor types corresponding to the several facial parts are screened.

[0146] In a specific embodiment of the present invention, the specific method for selecting sensor types corresponding to a plurality of torso parts and sensor types corresponding to a plurality of facial parts is as follows:

[0147] Rule 1: Ensure the compatibility of the sensor with the actor's body: The selected sensor must be worn on the actor's designated body part, be consistent with the model's mounting position on that part, and meet the preset wearing comfort requirements;

[0148] Rule 2: Evaluate the historical performance of sensors: Analyze the stability and accuracy of the sensor's historical data collection in various environmental scenarios;

[0149] Based on the first and second screening rules, several sensor types in the database are traversed to obtain several torso-specific and facial-specific sensors.

[0150] In a specific embodiment of the present invention, the method for ensuring the compatibility of the sensor with the actor's body is as follows:

[0151] l. Deployment location determination: Based on the specified IP image corresponding to the adaptation actor's body parts deployment set, determine the designated IP image to be deployed sensors corresponding to the adaptation actor's several torso parts and several facial parts;

[0152] m. Type matching: Based on the multiple body parts and multiple facial parts, select sensor types that meet the following performance indicators as candidates:

[0153] n1. Wearing comfort: Meets the preset comfort threshold;

[0154] n2.Flexibility of movement: Ensure the flexibility of joint movement;

[0155] n3. Freedom of movement: supports high degree of freedom of body movements;

[0156] It should be noted that, taking the sensor types of accelerometer, gyroscope, inertial measurement unit, strain gauge sensor, myoelectric sensor, and optical facial marker as examples, the sensor type performance index table is shown in Table 3;

[0157] Table 3 - Sensor type performance index table

[0158]

[0159] It should be noted that, taking the body parts of spine, shoulder, knee, eyebrow, mouth, and eye as an example, the body part characteristic requirement table is shown in Table 4;

[0160] Table 4 - Body part characteristic requirements

[0161]

[0162]

[0163] In a specific embodiment of the present invention, the method for evaluating the historical performance of the sensor is as follows:

[0164] Step 1: Extract performance data: Extract performance data of various sensors under different environments from the database. The performance data includes:

[0165] p1. Accuracy deviation rate:

[0166] p2. Failure rate:

[0167] p3. Transmission rate: v

[0168] Step 2: Calculate performance effect value: Generate performance effect value E of each sensor under each environmental label and type of environment based on performance data evaluation;

[0169] Step 3: Current environment adaptation screening:

[0170] q1. Get the actual collection environment of the specified IP image;

[0171] q2. Based on the performance value obtained in step 2, extract the performance value E of each sensor in the actual acquisition environment env ;

[0172] The evaluation dimensions of the performance effect value E include:

[0173] r1, accuracy deviation rate;

[0174] r2, failure rate;

[0175] r3. Transmission rate.

[0176] It should be noted that the performance effect value E is calculated based on the performance data. Among them, α, β, γ, κ are preset weight coefficients, v max is the preset maximum transmission rate benchmark value, e -βδ represents the precision deviation penalty term (the larger the precision deviation, the smaller the value), Indicates the failure rate suppression item (the higher the failure rate, the smaller the value).

[0177] A sensor deployment module is used to implement sensor deployment according to a deployment set of body parts of an adapted actor corresponding to a specified IP image;

[0178] The animation generation module is used to convert the body part deployment set of the actor corresponding to the specified IP image into a three-dimensional motion animation.

[0179] In a specific embodiment of the present invention, the sensor deployment is implemented according to the deployment set of body parts of the actor corresponding to the designated IP image, and the specific method is:

[0180] s. Location Mapping: Locating the physical locations of the torso and facial areas where sensors are to be deployed based on the determined set of body part deployments;

[0181] t. Type matching deployment: Deploy sensors from the determined set of compatible sensor types to the corresponding body parts according to the following rules:

[0182] u1. The trunk position sensor is installed at the joint axis or muscle movement sensitive area;

[0183] u2. Facial sensors fit the distribution of micro-expression muscles;

[0184] v. Calibration verification: Power on the sensor, collect test motion data, and verify whether the matching degree between the sensor output signal and the preset motion model exceeds the threshold.

[0185] It should be noted that the specific method for converting the designated IP image into a three-dimensional motion animation is as follows:

[0186] The motion capture technology is used to record and replay the motion trajectory information of the actor corresponding to the designated IP image. Sensors are used to detect and record the motion trajectory of the actor's body, face, and even the camera and light source in three-dimensional space in real time. These records are converted into data and then assigned to the virtual character model generated in the animation software. This model and the actor corresponding to the designated IP image have the same movements and generate a motion sequence.

[0187] The motion capture technology records the motion trajectory of the actor corresponding to the specified IP image, then processes the recorded motion data and converts it into points in three-dimensional space, uses these action data to drive the generated model, and then forms a series of computer animations.

[0188] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should all fall within the scope of protection of the present invention.

Claims

1. A device for sampling three-dimensional motion data of an IP image, characterized in that: include: a sensor screening module, which determines an actor suitable for the specified IP image based on the character model of the specified IP image and the three-dimensional data acquisition environment, and determines a body part deployment set corresponding to the actor suitable for the specified IP image, wherein the body part deployment set includes sensor types corresponding to several body parts; A sensor deployment module is used to implement sensor deployment according to a deployment set of body parts of an adapted actor corresponding to a specified IP image; The animation generation module is used to convert the body part deployment set of the actor corresponding to the specified IP image into a three-dimensional motion animation.

2. The device for sampling three-dimensional motion data of an IP image according to claim 1, characterized in that: The specific method for determining the actor suitable for the specified IP image is as follows: Extract body models of several actors and models of designated IP images from the database; Adjusting the body models of several actors at different proportions to obtain the adjusted body models of the several actors, evaluating the degree of overlap between the adjusted body models of the several actors and the designated IP image, and selecting the maximum overlap as the first fitness degree RY between the several actors and the designated IP image, where Y represents the number of the several actors; Extracting historical performance data of several actors from the database to analyze the second compatibility NY of the several actors with the designated IP image, where Y represents the number of the several actors; Comprehensively analyze the compatibility of several actors with the designated IP image OY=λ1*(RY)+λ2*(NY).

3. The device for sampling three-dimensional motion data of an IP image according to claim 2, characterized in that: The specific method for evaluating the degree of overlap between the adjusted body models of several actors and the designated IP image is as follows: Step 1: Voxelize the model; Convert the specified IP image model into a voxel grid with a fixed grid resolution of 1mm*1mm*1mm to obtain a binary voxel grid G ip , where G ip (x) = 1 means that the voxel corresponding to the identifier x is inside the specified IP image model, G ip (x)=0 means that the voxel corresponding to the identifier x is not inside the specified IP image model, where x is the globally unique identifier of the voxel unit in the voxel grid; The actor's body model is converted into a voxel grid with a fixed grid resolution of 1mm*1mm*1mm to obtain a binary voxel grid G body , where G body (x) = 1 means that the voxel corresponding to the identifier x is inside the actor's body model, G body (x)=0 means that the voxel corresponding to the identifier x is not inside the actor's body model, where x is the globally unique identifier of the voxel unit in the voxel grid; Step 2: Calculate the basic volume; The physical volume of a single voxel is the voxel volume v l =1mm 3 ; Specify the volume V of the IP image ip =N ip *v l , where N ip Represents G ip the number of internal voxels in ; The volume V of the actor's body model body =N body *v l , where N body Represents G body the number of internal voxels in ; Step 3, calculate the overlap volume; In the same voxel grid, check whether the voxel belongs to both models and calculate the coincidence judgment value for each voxel: G ol When (x) = 1, it means that voxel x is inside the specified IP image model and inside the actor's body model, ∧ is a logical symbol and; The total number N of voxels with a statistical coincidence judgment value of 1 ol , calculate the overlap volume V of the specified IP image model and the actor's body model ol =N ol *v l ; Step 4: Calculate the degree of overlap; Calculate the overlap between the specified IP image model and the actor's body model Perform the above steps 1-4 on the adjusted body models of several actors and the designated IP image model, calculate the degree of overlap between the adjusted body models of several actors and the designated IP image model, and select the maximum overlap as the first degree of fit between the several actors and the designated IP image, thereby obtaining the first degree of fit RY between the several actors and the designated IP image, where Y represents the number of the actors.

4. The device for sampling three-dimensional motion data of an IP image according to claim 2, characterized in that: The specific method for analyzing the second compatibility between a number of actors and a designated IP image is as follows: Step 1: Historical data extraction: extracting historical performance data of several actors from the database, wherein the historical performance data includes the three-dimensional motion animation of the IP image corresponding to each historical performance; Step 2: Action interpretation and quantification: a. Get the required action set A of the specified IP image stored in the database req ={a req1 ,a req2 ,...,a reqn }, where n is the total number of required actions; b. Extracting and interpreting action model A from 3D motion animation act ={a act1 ,a act2 ,...,a actm }, where m is the total number of actual deductive actions; c. Calculate the standardized status of each required action: for i is the required action index, if there is a reqi ∈A act , j is the index of the deductive action, satisfying sim(a actj ,a reqi )≥θ, where θ is the preset overlap threshold, sim is the action similarity function, then mark a reqi To standardize actions; d. Quantified action interpretation is: where N standard is the number of standardized actions; Step 3: Quantify movement fluency: e. Convert the three-dimensional motion animation into a time sequence action sequence S = {(t k ,a k ,d k )|k=1,2,...,K}, where k is the frame index, K is the total number of frames, t k is the action start time, a k is the action identifier, d k The duration of the action; f. Quantify the movement fluency as: Q fluency =f(S), where function f is generated based on the continuity of the action transition sequence and the stability of the duration; Step 4: Calculate facial emotion matching coefficient: g. Get the facial emotion requirement set E of the specified IP image req ={e req1 ,e req2 ,...,e reqp }, where p is the total number of demand emotion types; h. Extracting and generating facial emotion sequence E from the three-dimensional motion animation act ={e act1 ,e act2 ,...,e actq }, where q is the number of emotion segments actually captured; i. Process the data and calculate the sentiment matching coefficient: Among them, sim e It is the emotion similarity function, which returns the similarity in the interval [0,1]; sim e =1 means the emotion matches perfectly, sim e =0 indicates a complete mismatch in sentiment, and intermediate values ​​indicate a partial match; Step 5: Second fitness comprehensive quantification Generate the second fitness: Among them, w1 is the action standard weight, w2 is the action smoothness weight, and w3 is the facial emotion weight; Based on the three-dimensional motion animation of the IP image of each historical interpretation of several actors, execute steps 1-5 to calculate the second fitness of each historical interpretation of several actors and the specified IP image, and perform average processing on them to obtain the second fitness of several actors and the specified IP image NY, where Y represents the number of several actors.

5. The device for sampling three-dimensional motion data of an IP image according to claim 1, characterized in that: The specific method for determining the body part deployment set of the actor corresponding to the specified IP image is as follows: Based on the model of the specified IP image, combined with the comparison table of various IP images and basic parts stored in the database, several basic torso parts and several basic facial parts of the specified IP image are screened out; According to the type of the specified IP image, all IP images of the same type are filtered from the IP image library, and the common parts of all IP images of the same type are counted; Based on the required action set, the required attachment parts for each action are determined according to a required attachment part mapping table pre-stored in the database, and a number of required torso parts and a number of required facial parts of the designated IP image are generated accordingly; the required attachment part mapping table includes data on the correspondence between action types and required attachment parts; According to the specified IP image corresponding to the several special required torso parts and several special required facial parts of the actor, and obtaining the collection environment of the specified IP image corresponding to the actor, the sensor types corresponding to the several torso parts and the sensor types corresponding to the several facial parts are screened.

6. The device for sampling three-dimensional motion data of an IP image according to claim 5, characterized in that: The specific method of screening the sensor types corresponding to the body parts and the sensor types corresponding to the facial parts is as follows: Rule 1: Ensure the compatibility of the sensor with the actor's body: The selected sensor must be worn on the actor's designated body part, be consistent with the model's mounting position on that part, and meet the preset wearing comfort requirements; Rule 2: Evaluate the historical performance of sensors: Analyze the stability and accuracy of the sensor's historical data collection in various environmental scenarios; Based on the first and second screening rules, several sensor types in the database are traversed to obtain several torso-specific and facial-specific sensors.

7. The device for sampling IP image 3D motion data according to claim 6, characterized in that: The specific method for ensuring the compatibility of the sensor with the actor's body is as follows: l. Deployment location determination: Based on the specified IP image corresponding to the adaptation actor's body parts deployment set, determine the designated IP image to be deployed sensors corresponding to the adaptation actor's several torso parts and several facial parts; m. Type matching: Based on the multiple body parts and multiple facial parts, select sensor types that meet the following performance indicators as candidates: n1. Wearing comfort: Meets the preset comfort threshold; n2.Flexibility of movement: Ensure the flexibility of joint movement; n3. Freedom of movement: supports high degree of freedom of body movements.

8. The device for sampling three-dimensional motion data of an IP image according to claim 6, characterized in that: The specific method for evaluating the historical performance of the sensor is as follows: Step 1: Extract performance data: Extract performance data of various sensors under different environments from the database. The performance data includes: p1. Accuracy deviation rate: p2. Failure rate: p3. Transmission rate: v; Step 2: Calculate performance effect value: Generate performance effect value E of each sensor under each environmental label and type of environment based on performance data evaluation; Step 3: Current environment adaptation screening: q1. Get the actual collection environment of the specified IP image; q2. Based on the performance value obtained in step 2, extract the performance value E of each sensor in the actual acquisition environment env ; The evaluation dimensions of the performance effect value E include: r1, accuracy deviation rate; r2, failure rate; r3. Transmission rate.

9. The device for sampling three-dimensional motion data of an IP image according to claim 8, characterized in that: The specific method of implementing sensor deployment according to the deployment set of body parts of the actor corresponding to the designated IP image is as follows: s. Location Mapping: Locating the physical locations of the torso and facial areas where sensors are to be deployed based on the determined set of body part deployments; t. Type matching deployment: Deploy sensors from the determined set of compatible sensor types to the corresponding body parts according to the following rules: u1. The trunk position sensor is installed at the joint axis or muscle movement sensitive area; u2. Facial sensors fit the distribution of micro-expression muscles; v. Calibration verification: Power on the sensor, collect test motion data, and verify whether the matching degree between the sensor output signal and the preset motion model exceeds the threshold.

Images