CN105513136B - A kind of 3D actor model framework extraction methods based on level set central cluster - Google Patents

Abstract

The present invention provides a method for extracting a 3D character model skeleton based on level-set center clustering, comprising the following steps: cutting the 3D character model using a height function as a Morse function to obtain a set of intersection points after cutting; combining the set of intersection points after cutting Clustering is performed according to the connectivity between intersection points, and clustered into several sub-level sets; the level sets whose area is greater than the threshold of the level set area and whose circularity is greater than the threshold of circularity are screened out; The center of the set is clustered, so that the level set is divided into arms, legs and torso of the 3D character model; the resulting skeleton lines are embedded into the 3D character model. The present invention can automatically extract the skeleton for the three-dimensional animation model. Compared with the traditional distance transformation method, the horizontal set clustering method used in the skeleton extraction uses the cutting model, divides the model parts, and calculates the joints of each part respectively. , the average time consumed is less and the error is smaller.

Description

A 3D Character Model Skeleton Extraction Method Based on Level Set Center Clustering

technical field

The invention belongs to the technical field of three-dimensional animation production, in particular to a method for extracting a skeleton of a 3D character model based on horizontal set center clustering.

Background technique

With the development of virtual reality technology and digital media technology, people have more and more demands for the production of 3D character model animation. As an important part of multimedia technology, the technology of 3D animation production has been greatly developed in recent years. However, in the current mainstream 3D character production process, manual production is still the main method, and commercial software such as 3DMax or Maya is mainly used for design and production. However, the animation creation industry at this stage still faces the lack of suitable computer animation production related aspects. Therefore, in the production process of today's 3D animation, a certain amount of labor and a lot of brainpower are still required. People now use more widely computer animation techniques such as key frame and dynamic method. However, these methods all have a common defect: production is time-consuming and labor-intensive, and requires producers to have a relatively high degree of production skills and professional knowledge.

As more people demand for 3D character animation production, a simple way is needed to produce 3D animation, so that more people who lack experience in 3D character animation production can participate in this work. For example, users choose Models, actions, etc., after which the appropriate animations are automatically generated by the program. Therefore, automatic 3D character animation production technology has become a research hotspot for scientific researchers or engineers. This not only saves time and effort, but also allows more people to participate in animation production. In this context, the simplification and automation of the skeleton extraction of the model in the motion redirection technology has become a very critical part in the field of character animation.

Contents of the invention

The object of the present invention is to provide a method for extracting a skeleton of a 3D character model based on horizontal clustering.

Technical scheme of the present invention is:

A method for extracting a skeleton of a 3D character model based on level-set central clustering, comprising the following steps:

Step 1: Use the height function as the Morse function to cut the 3D character model to obtain the set of intersection points after cutting;

Step 2: Cluster the cut intersection points according to the connectivity between the intersection points, and cluster them into several sub-level sets: if some two intersection points are formed by the intersection of a height plane and a triangular mesh surface, then these The two intersection points are saved as a line segment, and the connectivity between the line segments is used for clustering to obtain several sub-level sets;

Step 3: Use the area and circularity of the level set as the screening conditions, and filter out the level sets whose area is greater than the level set area threshold and whose circularity is greater than the circularity threshold;

Step 4: Clustering the filtered level set by using the center of the level set, so as to divide the level set into arms, legs and torso of the 3D character model;

Step 5: Embed the obtained skeleton lines into the 3D character model.

The concrete steps of described step 1 are as follows:

Step 1-1: Select a height function to process the 3D character model to obtain the height h _model of the 3D character model;

Step 1-2: Set the cutting times n of the 3D character model, start cutting from the top of the 3D character model, and take a plane every h _model /n distance, which is the height plane;

Step 1-3: Read the triangular mesh surface in the FBX file, save the maximum value y _max and the minimum value y _min of each triangular mesh surface in the y-axis direction to an array, and set each triangle mesh surface in y The maximum value y _max in the axis direction is sorted;

Step 1-4: Calculate the intersection point between each height plane of the 3D character model and the triangular mesh surface, and obtain a list of triangular mesh surfaces that can be intersected by each height plane;

Steps 1-5: According to the list of triangular mesh surfaces that can be intersected by each height plane, determine the intersection points between each height plane and the triangular mesh surfaces that can intersect, that is, the set of intersection points after cutting.

The concrete steps of described steps 1-4 are as follows:

Step 1-4-1: Traverse each triangular mesh surface in the array, if the maximum value y _max of the current triangular mesh surface in the y-axis direction is greater than the current height plane P _current , then execute step 1-4-2, Otherwise, all triangular mesh faces after the current triangular mesh face will no longer intersect with the current height plane P _current , stop the traversal, and the obtained triangular mesh face list is a list of triangular mesh face faces that can intersect with the current height plane P _current ;

Step 1-4-2: Determine whether the minimum value y _min of the current triangular mesh surface in the y-axis direction is greater than the current height plane P _current : if yes, the current triangular mesh surface is above the current height plane P _current , and traverse A triangular mesh surface, return to step 1-4-1; otherwise, the current triangular mesh surface can intersect with the current height plane P _current , save the current triangular mesh surface, traverse the next triangular mesh surface, and return to step 1-4- 1.

The concrete steps of described step 3 are as follows:

Step 3-1: Set filter conditions, including level set area threshold and circularity threshold;

Step 3-2: Calculate the area and circularity of each level set;

Step 3-3: Filter out the level sets whose area is greater than the level set area threshold and whose circularity is greater than the circularity threshold.

The concrete steps of described step 4 are as follows:

Step 4-1: Calculate the center point of each level set to form a center point set;

Step 4-2: Sort all the center points according to the axial length values of the other two directions in the three-dimensional space except the height, and the direction with the largest axial length value is the direction in which the arms of the 3D character model are extended;

Step 4-3: Calculate the average value of the distance between the adjacent center points of the first 5 center points of the sorted center point set, and the average value of the distance between the adjacent center points of the last 5 center points, respectively. 2 times the average value as the two clustering radii;

Step 4-4: respectively select the first center point and the last center point of the sorted center point set as the starting point for clustering, and obtain the level set of the arms of the 3D character model;

Step 4-5: sort all the center points from small to large in the direction of the vertical axis, find the endpoint values of the legs, perform the same clustering as in step 4-4, and obtain the level set of the legs of the 3D character model;

Steps 4-6: use the center point that is not clustered in the center point set as the center point of the level set of the torso of the 3D character model;

Steps 4-7: Calculate the fitting straight line of the center point of the level set of the arms of the 3D character model, the fitting straight line of the center point of the level set of the legs, and the fitting straight line of the center point of the level set of the torso respectively as double The skeleton line of the arm, the skeleton line of the legs, and the skeleton line of the torso.

The concrete steps of described step 5 are as follows:

Step 5-1: Calculate the position of key joint points;

Key joint points include: head, neck, shoulder joint, elbow joint, wrist joint, hand, hip joint, thigh joint, ankle joint and foot;

Step 5-2: use the position of the key joint points to match the standard skeleton structure, and calculate the positions of all other relevant nodes through the position ratio between the joint points;

Step 5-3: Connect all semantically related joint points to get the final skeleton, which is embedded in the 3D character model.

The concrete steps of described step 5-1 are as follows:

Step 5-1-1: Obtain the positions of the head, hands, hip joints, thigh joints and foot joint points from the ends of the skeleton lines of each part of the cut 3D character model: the two ends of the leg skeleton lines are Thigh joints and feet; the lower end of the arm skeleton line is the hand; the two ends of the trunk skeleton line are the head and hip joints;

Step 5-1-2: use the perimeter of the level set obtained by cutting the triangular mesh surface of the arms of the 3D character model with the horizontal plane to determine the position of the wrist joint;

Step 5-1-3: Calculate the position of the neck joint point: use the method of step 5-1-2 to process the level set of the mesh model of the torso position above the arm, and use the midpoint of the level set with the smallest circumference as the neck The location of the ministries;

Step 5-1-4: Calculate the position of the shoulder joint: Use the direction of the skeleton line of the arms to intersect the plane where the shoulder joint is located to obtain an initial shoulder joint position and then correct it;

Step 5-1-5: Calculate the position of the ankle joint: first select the level set corresponding to the leg in the level set screened out in step 3, and then calculate the angle between the connecting lines of the midpoints of the selected adjacent level sets, not less than The angle position where the threshold is set is the ankle joint position.

The concrete steps of described step 5-1-2 are as follows:

Step 5-1-2-1: Calculate a rotation matrix using the angle between the direction of the skeleton line of the arms and the positive direction of the longitudinal axis;

The rotation matrix is represented as follows:

Among them, θ is the angle between the direction of the skeleton line of the arms and the positive direction of the longitudinal axis, (n _x , _ny , _nz ) is the rotation axis, which is obtained by cross-producting the skeleton line of the arms and the positive direction of the longitudinal axis;

Step 5-1-2-2: Multiply the dual-arm grid model by the rotation matrix, and erect the dual-arm grid model;

Step 5-1-2-3: Use the horizontal plane to perform cross-cutting operations on the double-arm grid model, and use the method in step 2 to obtain a series of level sets;

Step 5-1-2-4: If there are multiple level sets under a certain height plane, treat the vertices of all level sets as a point set, use the Graham algorithm to calculate the convex hull of the point set, and use the convex hull as the only level set below that height plane;

Step 5-1-2-5: Calculate the perimeter of each level set, and use the midpoint of the level set with the smallest perimeter as the position of the wrist joint.

The concrete steps of described step 5-1-4 are as follows:

Step 5-1-4-1: use the plane determined by the point with the largest vertical axis value of the skeleton line of both arms and the middle plane of the plane determined by the vertical axis value of the neck joint point to obtain the plane where the shoulder joint is located;

Step 5-1-4-2: Use the direction of the skeleton line of both arms to intersect the plane where the shoulder joint is located to obtain an intersection point, which is the position of the shoulder joint;

Step 5-1-4-3: Use the position of the shoulder joint and the position of the wrist joint point to obtain the initial position of the elbow joint through the position ratio of the elbow joint, shoulder joint and wrist joint point in the standard skeleton structure;

Step 5-1-4-4: Take the point above the plane determined by the value of the longitudinal axis of the elbow joint as the upper arm part, re-fit the center point of the upper arm part with a straight line, and make the intersection of the straight line with the plane where the shoulder joint is located to get correction For the position of the shoulder joint, repeat 5-1-4-3 until the position of the shoulder joint does not change, and the obtained shoulder joint point is the final position of the shoulder joint.

Beneficial effect:

The present invention can automatically extract the skeleton for the three-dimensional animation model. Compared with the traditional distance transformation method, the horizontal set clustering method used in the skeleton extraction uses the cutting model, divides the model parts, and calculates the joints of each part respectively. , the average time consumed is less and the error is smaller. Using this method to replace the traditional manual calculation can reduce the user threshold, facilitate use, and improve the accuracy and efficiency of skeleton extraction.

Description of drawings

Fig. 1 is a schematic diagram of a method for clustering level set central points according to a specific embodiment of the present invention;

Fig. 2 is a flow chart of clustering the set of cut intersection points according to the connectivity between the intersection points according to the specific embodiment of the present invention;

Fig. 3 is a flow chart of clustering the center points of the level set according to a specific embodiment of the present invention;

Fig. 4 (a)～(d) are four models that this method and the distance transformation method of prior art contrast adopt respectively in the embodiment of the present invention;

Fig. 5 (a)～(h) is the comparison experiment result diagram of this method and the distance transformation method of prior art in the specific embodiment of the present invention;

6 is a flow chart of a method for extracting a skeleton of a 3D character model based on horizontal center clustering according to a specific embodiment of the present invention;

Fig. 7 is the specific flowchart of step 1 of the specific embodiment of the present invention;

Fig. 8 is the specific flowchart of steps 1-4 of the specific embodiment of the present invention;

Fig. 9 is a specific flowchart of step 4 of the specific embodiment of the present invention;

FIG. 10 is a specific flow chart of step 5 of a specific embodiment of the present invention;

Fig. 11 is a specific flow chart of step 5-1-2 of the specific embodiment of the present invention.

Detailed ways

The specific implementation manners of the present invention will be described in detail below in conjunction with the accompanying drawings.

A kind of 3D character model skeleton extraction method based on level set central clustering, as shown in Figure 6, comprises the following steps:

Step 1: Use the height function as the Morse function to cut the 3D character model to obtain the set of intersection points after cutting;

As shown in Figure 7, the specific steps are as follows:

Step 1-1: Select a height function to process the 3D character model to obtain the height h _model of the 3D character model;

Step 1-2: Set the number of times of cutting the 3D character model to n=80, start cutting from the top (top of the head) of the 3D character model, and take a plane every h _model /n distance, that is, a height plane;

Step 1-3: In order to improve efficiency, when reading the triangular mesh surface in the FBX file, save the maximum value y _max and minimum value y _min of each triangular mesh surface in the y-axis direction to an array, and for each triangle The grid surface is sorted by the maximum value y _max in the y-axis direction;

Step 1-4: Calculate the intersection point between each height plane of the 3D character model and the triangular mesh surface, and obtain a list of triangular mesh surfaces that can be intersected by each height plane;

As shown in Figure 8, the specific steps are as follows:

Step 1-4-1: Traverse each triangular mesh surface in the array, if the maximum value y _max of the current triangular mesh surface in the y-axis direction is greater than the current height plane P _current , execute step 1-4-2, Otherwise, all triangular mesh faces after the current triangular mesh face will no longer intersect with the current height plane P _current , stop traversal, and the obtained triangular mesh face list is a list of triangular mesh faces that can intersect with the current height plane P _current , execute steps 1-5;

Step 1-4-2: Determine whether the minimum value y _min of the current triangular mesh surface in the y-axis direction is greater than the current height plane P _current : if yes, the current triangular mesh surface is above the current height plane P _current , and traverse A triangular mesh surface, return to step 1-4-1; otherwise, the current triangular mesh surface can intersect with the current height plane P _current , save the current triangular mesh surface, traverse the next triangular mesh surface, and return to step 1-4- 1.

There are two intersection points between the height plane and the triangular mesh surface. The formula for calculating the intersection point is as follows:

x＝x ₁ +(P _current -y ₁ )*(x ₂ -x ₁ )÷(y ₂ -y ₁ )

z＝z ₁ +(P _current -y ₁ )*(z ₂ -z ₁ )÷(y ₂ -y ₁ )

Among them, x and z represent the x-axis and z-axis values of the intersection point of a side of the triangular mesh surface and the current height plane P _current respectively, x ₁ and x ₂ represent the x-axis values of the two endpoints of the side, z ₁ and z ₂ represent the z-axis values of the two endpoints of this side, and the y value of the intersection point is the value of the current height plane P _current .

Steps 1-5: According to the list of triangular mesh surfaces that can be intersected by each height plane, determine the intersection points between each height plane and the triangular mesh surfaces that can intersect, that is, the set of intersection points after cutting.

Step 2: Cluster the cut intersection points according to the connectivity between the intersection points, and cluster them into several sub-level sets: if some two intersection points are formed by the intersection of a height plane and a triangular mesh surface, then these The two intersection points are saved as a line segment, and the connectivity between the line segments is used for clustering to obtain several sub-level sets;

Figure 2 describes the flow chart of clustering the cut intersection points according to the connectivity between the intersection points. The specific method flow is as follows:

Step 2-1: Record the line segment set L{l ₁ ,l ₂ ,…,l _n } formed by the intersection of the current height plane P _current and the triangular mesh surface;

Step 2-2: Take l _k , k∈[1,n] arbitrarily, and record that one endpoint of l _k is v _start , which is the starting point of the entire traversal process, and the other endpoint is v _current , which is the current search point , record the current set of clustering points as E _i , i=1, add v _start to the current set;

Step 2-3: Traverse all other line segments in the line segment set L, if One of the vertex values of l _i is equal to v _current , then take the other endpoint of l _i as v _current , delete l _i from the line segment set L, and add v _current to E _i at the same time, if L is empty at this time then Go to step 2-4, if v _current ＝v _start , jump to step 2-2, and add 1 to i, if a closed curve is not formed after the traversal, then set v _current as the new v _start , and repeat the steps 2-3, traverse to the end of the last connecting curve to form a closed curve, and save the curve, if L is empty, jump to step 2-4, otherwise go to step 2-2;

Step 2-4: Obtain the set of intersection points corresponding to each closed curve as each sub-level set.

Step 3: Use the area and circularity of the level set as the screening conditions, and filter out the level sets whose area is greater than the level set area threshold and whose circularity is greater than the circularity threshold;

Step 3-1: Set filter conditions, including level set area threshold and circularity threshold;

The calculation formula of the level set area threshold T is as follows:

Wherein, _hmodel is the height of the 3D character model, _hstandard is the preset height of the standard model, which is taken as 160 in this embodiment, the Round() function is a rounding function, and ST is the level set area threshold obtained in the standard model , ST=0.2 is taken in this embodiment.

Step 3-2: Calculate the area and circularity of each level set;

Using the theorem "obtained by the sum of the triangular vector area formed by connecting any point and two consecutive points on the polygon", calculate the area of each level set, and use the following formula to calculate:

Among them, S(P) represents the area of the polygon composed of the sub-level set P, M is any point, is the vector formed by connecting the point M with the i-th point in the sublevel set P, is the vector formed by connecting point M with the i+1th intersection point in sub-level set P.

The formula for calculating the circularity of a level set is:

Among them, C(P) represents the circularity of the polygon composed of the sub-level set P, S(P) is the area of the polygon composed of the sub-level set P, and P(P) represents the perimeter of the polygon composed of the sub-level set P.

Step 3-3: Filter out the level sets whose area is greater than the level set area threshold and whose circularity is greater than the circularity threshold. The threshold of circularity in this embodiment is set as 0.4 based on experience.

Step 4: Clustering the filtered level set by using the center of the level set, so as to divide the level set into arms, legs and torso of the 3D character model;

As shown in Figure 9, the specific steps are as follows:

Step 4-1: Calculate the center point of each level set to form a center point set;

Where C is the center point of the level set P.

Step 4-2: Sort all the center points according to the axial length values of the other two directions in the three-dimensional space except the height, and the direction with the largest axial length value is the direction in which the arms of the 3D character model are extended;

Step 4-3: Calculate the average value of the distances between the adjacent center points of the first 5 center points of the sorted center point set C _all , and the average value of the distances between the adjacent center points of the last 5 center points, respectively. 2 times of the two mean values are used as the two clustering radii;

Step 4-4: respectively select the first center point and the last center point of the sorted center point set C _all as the starting point for clustering, and obtain the level set of the arms of the 3D character model;

Figure 1 is a schematic diagram describing the method of clustering the center points of the level set, wherein, the starting point T1 is added to the clustering set, and it is judged whether the distance between T2 and the elements in the clustering set is within a given clustering radius, If there is at least one condition that satisfies the condition, T2 will be added to the clustering set, and so on for the next elements.

Figure 3 is a specific flow chart for clustering the center points of the level set, and the specific execution steps are as follows:

Step 4-4-1: Create a queue Q, add the starting point Cstart to the queue Q, and mark Cstart as processed;

Step 4-4-2: Take the first element of the queue Q as the current judgment point Ccurrent, and add Ccurrent to the clustering result set Cresult;

Step 4-4-3: traverse the central point set C _all , if the distance between the currently traversed central point and Ccurrent is less than the clustering radius R and has not been processed, add the central point to the queue Q, and The center point is marked as processed;

Step 4-4-4: Determine whether the queue Q is empty, if not, execute step 4-4-2; otherwise, execute step 4-4-5;

Step 4-4-5: Cresult is the final clustering result, that is, the level set of the arms of the 3D character model;

Step 4-5: sort all the center points from small to large in the direction of the vertical axis, find the endpoint values of the legs, perform the same clustering as in step 4-4, and obtain the level set of the legs of the 3D character model;

Steps 4-6: use the center point that is not clustered in the center point set as the center point of the level set of the torso of the 3D character model;

Steps 4-7: Calculate the fitting straight line of the center point of the level set of the arms of the 3D character model, the fitting straight line of the center point of the level set of the legs, and the fitting straight line of the center point of the level set of the torso respectively as double The skeleton line of the arm, the skeleton line of the legs, and the skeleton line of the torso.

Step 5: Embed the obtained skeleton lines into the 3D character model.

As shown in Figure 10, the specific steps are as follows:

Step 5-1: Calculate the position of key joint points;

Key joint points include: head, neck, shoulder joint, elbow joint, wrist joint, hand, hip joint, thigh joint, ankle joint and foot;

Step 5-1-1: Obtain the positions of the head, hands, hip joints, thigh joints and foot joint points from the ends of the skeleton lines of each part of the cut 3D character model: the two ends of the leg skeleton lines are Thigh joints and feet; the lower end of the arm skeleton line is the hand; the two ends of the trunk skeleton line are the head and hip joints;

Step 5-1-2: use the perimeter of the level set obtained by cutting the triangular mesh surface of the arms of the 3D character model with the horizontal plane to determine the position of the wrist joint;

As shown in Figure 11, the specific steps are as follows:

Step 5-1-2-1: Calculate a rotation matrix using the angle between the direction of the skeleton line of the arms and the positive direction of the longitudinal axis;

The rotation matrix is represented as follows:

Among them, θ is the angle between the direction of the skeleton line of the arms and the positive direction of the longitudinal axis, (n _x , _ny , _nz ) is the rotation axis, which is obtained by cross-producting the skeleton line of the arms and the positive direction of the longitudinal axis;

Step 5-1-2-2: Multiply the dual-arm grid model by the rotation matrix, and erect the dual-arm grid model;

Step 5-1-2-3: Use the horizontal plane to perform cross-cutting operations on the double-arm grid model, and use the method in step 2 to obtain a series of level sets;

Step 5-1-2-4: If there are multiple level sets under a certain height plane, treat the vertices of all level sets as a point set, use the Graham algorithm to calculate the convex hull of the point set, and use the convex hull as the only level set below that height plane;

Step 5-1-2-5: Calculate the perimeter of each level set, and use the midpoint of the level set with the smallest perimeter as the position of the wrist joint.

Step 5-1-3: Calculate the position of the neck joint point: use the method of step 5-1-2 to process the level set of the mesh model of the torso position above the arm, and use the midpoint of the level set with the smallest circumference as the neck The location of the ministries;

Step 5-1-4: Calculate the position of the shoulder joint: Use the direction of the skeleton line of the arms to intersect the plane where the shoulder joint is located to obtain an initial shoulder joint position and then correct it;

Step 5-1-4-1: Use the plane determined by the point with the largest vertical axis value of the skeleton line of both arms and the middle plane of the plane determined by the vertical axis value of the neck joint point to obtain the plane where the shoulder joint is located;

Step 5-1-4-2: Use the direction of the skeleton line of both arms to intersect the plane where the shoulder joint is located to obtain an intersection point, which is the position of the shoulder joint;

Step 5-1-4-3: Use the position of the shoulder joint and the position of the wrist joint point to obtain the initial position of the elbow joint through the position ratio of the elbow joint, shoulder joint and wrist joint point in the standard skeleton structure;

Step 5-1-4-4: Take the point above the plane determined by the value of the longitudinal axis of the elbow joint as the upper arm part, re-fit the center point of the upper arm part with a straight line, and make the intersection of the straight line with the plane where the shoulder joint is located to get correction For the position of the shoulder joint, repeat 5-1-4-3 until the position of the shoulder joint does not change, and the obtained shoulder joint point is the final position of the shoulder joint.

Step 5-1-5: Calculate the position of the ankle joint: first select the level set corresponding to the leg in the level set screened out in step 3, and then calculate the angle between the connecting lines of the midpoints of the selected adjacent level sets, not less than The angle position at which the threshold is set to 70 degrees is the position of the ankle joint.

Step 5-2: use the position of the key joint points to match the standard skeleton structure, and calculate the positions of all other relevant nodes through the position ratio between the joint points;

Step 5-3: Connect all semantically related joint points to get the final skeleton, which is embedded in the 3D character model.

This method is compared with the distance transformation method of the prior art, and the four models in Fig. 4(a)-(d) are compared, and the skeletons obtained by the two methods are shown in Fig. 5 . Among them, Figure 5(a)(c)(e)(g) is the skeleton obtained by the distance transformation method, and Figure 5(b)(d)(f)(h) is the skeleton obtained by this method. Compare the time efficiency and accuracy parameters of these two methods respectively. Table 1 shows the time comparison of the two methods. Table 2 shows the comparison of the error of the two methods for the arm length. Table 3 shows The error comparison of the two methods for leg length is presented.

Table 1 Time comparison of the two methods

Table 2 Comparison of the error of the two methods for the arm length

Table 3. Error comparison of two methods for leg length

It can be seen from the comparison that the method of the present invention has certain advantages in terms of time efficiency and accuracy for the distance transformation method.

Claims (9)

1. A 3D character model skeleton extraction method based on level set central clustering, characterized in that, comprising the following steps: Step 1: Use the height function as the Morse function to cut the 3D character model to obtain the set of intersection points after cutting; Step 2: Cluster the cut intersection points according to the connectivity between the intersection points, and cluster them into several sub-level sets: if some two intersection points are formed by the intersection of a height plane and a triangular mesh surface, then these The two intersection points are saved as a line segment, and the connectivity between the line segments is used for clustering to obtain several sub-level sets; Step 3: Use the area and circularity of the level set as the screening conditions, and filter out the level sets whose area is greater than the level set area threshold and whose circularity is greater than the circularity threshold; Step 4: Cluster the screened level set using the center of the level set to obtain the center point of the level set of the arms, the level set of the legs, and the level set of the torso of the 3D character model, and calculate the arms of the 3D character model The fitting straight line of the center point of the level set, the fitting straight line of the center point of the level set of the legs, and the fitting straight line of the center point of the level set of the torso are respectively used as the skeleton line of the arms, the skeleton line of the legs, and the trunk the skeleton line; Step 5: Embed the obtained skeleton lines into the 3D character model. 2. the 3D character model skeleton extraction method based on level set center clustering according to claim 1, is characterized in that, the concrete steps of described step 1 are as follows: Step 1-1: Select a height function to process the 3D character model to obtain the height h _model of the 3D character model; Step 1-2: Set the cutting times n of the 3D character model, start cutting from the top of the 3D character model, and take a plane every h _model /n distance, which is the height plane; Step 1-3: Read the triangular mesh surface in the FBX file, save the maximum value y _max and the minimum value y _min of each triangular mesh surface in the y-axis direction to an array, and set each triangle mesh surface in y The maximum value y _max in the axis direction is sorted; Step 1-4: Calculate the intersection point between each height plane of the 3D character model and the triangular mesh surface, and obtain a list of triangular mesh surfaces that can be intersected by each height plane; Steps 1-5: According to the list of triangular mesh surfaces that can be intersected by each height plane, determine the intersection points between each height plane and the triangular mesh surfaces that can intersect, that is, the set of intersection points after cutting. 3. the 3D character model skeleton extraction method based on horizontal set center clustering according to claim 2, is characterized in that, the concrete steps of described step 1-4 are as follows: Step 1-4-1: Traverse each triangular mesh surface in the array, if the maximum value y _max of the current triangular mesh surface in the y-axis direction is greater than the current height plane P _current , then execute step 1-4-2, Otherwise, all triangular mesh faces after the current triangular mesh face will no longer intersect with the current height plane P _current , stop the traversal, and the obtained triangular mesh face list is a list of triangular mesh face faces that can intersect with the current height plane P _current ; Step 1-4-2: Determine whether the minimum value y _min of the current triangular mesh surface in the y-axis direction is greater than the current height plane P _current : if yes, the current triangular mesh surface is above the current height plane P _current , and traverse A triangular mesh surface, return to step 1-4-1; otherwise, the current triangular mesh surface can intersect with the current height plane P _current , save the current triangular mesh surface, traverse the next triangular mesh surface, and return to step 1-4- 1. 4. the 3D character model skeleton extraction method based on level set center clustering according to claim 1, is characterized in that, the specific steps of described step 3 are as follows: Step 3-1: Set filter conditions, including level set area threshold and circularity threshold; Step 3-2: Calculate the area and circularity of each level set; Step 3-3: Filter out the level sets whose area is greater than the level set area threshold and whose circularity is greater than the circularity threshold. 5. the 3D character model skeleton extraction method based on level set center clustering according to claim 1, is characterized in that, the concrete steps of described step 4 are as follows: Step 4-1: Calculate the center point of each level set to form a center point set; Step 4-2: Sort all the center points according to the axial length values of the other two directions in the three-dimensional space except the height, and the direction with the largest axial length value is the direction in which the arms of the 3D character model are extended; Step 4-3: Calculate the average value of the distance between the adjacent center points of the first 5 center points of the sorted center point set, and the average value of the distance between the adjacent center points of the last 5 center points, respectively. 2 times the average value as the two clustering radii; Step 4-4: respectively select the first center point and the last center point of the sorted center point set as the starting point for clustering, and obtain the level set of the arms of the 3D character model; Step 4-5: sort all the center points from small to large in the direction of the vertical axis, find the endpoint values of the legs, perform the same clustering as in step 4-4, and obtain the level set of the legs of the 3D character model; Steps 4-6: use the center point that is not clustered in the center point set as the center point of the level set of the torso of the 3D character model; Steps 4-7: Calculate the fitting straight line of the center point of the level set of the arms of the 3D character model, the fitting straight line of the center point of the level set of the legs, and the fitting straight line of the center point of the level set of the torso respectively as double The skeleton line of the arm, the skeleton line of the legs, and the skeleton line of the torso. 6. the 3D character model skeleton extraction method based on level set center clustering according to claim 1, is characterized in that, the specific steps of described step 5 are as follows: Step 5-1: Calculate the position of key joint points; Key joint points include: head, neck, shoulder joint, elbow joint, wrist joint, hand, hip joint, thigh joint, ankle joint and foot; Step 5-2: use the position of the key joint points to match the standard skeleton structure, and calculate the positions of all other relevant nodes through the position ratio between the joint points; Step 5-3: Connect all semantically related joint points to get the final skeleton, which is embedded in the 3D character model. 7. The 3D character model skeleton extraction method based on horizontal set center clustering according to claim 6, wherein the specific steps of the step 5-1 are as follows: Step 5-1-1: Obtain the positions of the head, hands, hip joints, thigh joints and foot joint points from the ends of the skeleton lines of each part of the cut 3D character model: the two ends of the leg skeleton lines are Thigh joints and feet; the lower end of the arm skeleton line is the hand; the two ends of the trunk skeleton line are the head and hip joints; Step 5-1-2: use the perimeter of the level set obtained by cutting the triangular mesh surface of the arms of the 3D character model with the horizontal plane to determine the position of the wrist joint; Step 5-1-3: Calculate the position of the neck joint point: use the method of step 5-1-2 to process the level set of the mesh model of the torso position above the arm, and use the midpoint of the level set with the smallest circumference as the neck The positions of the joints; Step 5-1-4: Calculate the position of the shoulder joint: Use the direction of the skeleton line of the arms to intersect the plane where the shoulder joint is located to obtain an initial shoulder joint position and then correct it; Step 5-1-5: Calculate the position of the ankle joint: first select the level set corresponding to the leg in the level set screened out in step 3, and then calculate the angle between the connecting lines of the midpoints of the selected adjacent level sets, not less than The angle position where the threshold is set is the ankle joint position. 8. The 3D character model skeleton extraction method based on horizontal set center clustering according to claim 7, wherein the specific steps of the step 5-1-2 are as follows: Step 5-1-2-1: Calculate a rotation matrix using the angle between the direction of the skeleton line of the arms and the positive direction of the longitudinal axis; The rotation matrix is represented as follows: <mrow><mi>M</mi><mo>=</mo><mfenced open = "[" close = "]"><mtable><mtr><mtd><mrow><msubsup><mi>n</mi><mi>x</mi><mn>2</mn></msubsup><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>+</mo><mi>cos</mi><mi>&amp;theta;</mi></mrow></mtd><mtd><mrow><msub><mi>n</mi><mi>x</mi></msub><msub><mi>n</mi><mi>y</mi></msub><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>+</mo><msub><mi>n</mi><mi>z</mi></msub><mi>sin</mi><mi>&amp;theta;</mi></mrow></mtd><mtd><mrow><msub><mi>n</mi><mi>x</mi></msub><msub><mi>n</mi><mi>z</mi></msub><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>-</mo><msub><mi>n</mi><mi>y</mi></msub><mi>sin</mi><mi>&amp;theta;</mi></mrow></mtd></mtr><mtr><mtd><mrow><msub><mi>n</mi><mi>x</mi></msub><msub><mi>n</mi><mi>y</mi></msub><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>-</mo><msub><mi>n</mi><mi>z</mi></msub><mi>sin</mi><mi>&amp;theta;</mi></mrow></mtd><mtd><mrow><msubsup><mi>n</mi><mi>y</mi><mn>2</mn></msubsup><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>+</mo><mi>cos</mi><mi>&amp;theta;</mi></mrow></mtd><mtd><mrow><msub><mi>n</mi><mi>y</mi></msub><msub><mi>n</mi><mi>z</mi></msub><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>+</mo><msub><mi>n</mi><mi>x</mi></msub><mi>sin</mi><mi>&amp;theta;</mi></mrow></mtd></mtr><mtr><mtd><mrow><msub><mi>n</mi><mi>x</mi></msub><msub><mi>n</mi><mi>z</mi></msub><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>+</mo><msub><mi>n</mi><mi>y</mi></msub><mi>sin</mi><mi>&amp;theta;</mi></mrow></mtd><mtd><mrow><msub><mi>n</mi><mi>y</mi></msub><msub><mi>n</mi><mi>z</mi></msub><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>-</mo><msub><mi>n</mi><mi>x</mi></msub><mi>sin</mi><mi>&amp;theta;</mi></mrow></mtd><mtd><mrow><msubsup><mi>n</mi><mi>z</mi><mn>2</mn></msubsup><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>cos</mi><mi>&amp;theta;</mi><mo>)</mo></mrow><mo>+</mo><mi>cos</mi><mi>&amp;theta;</mi></mrow></mtd></mtr></mtable></mfenced></mrow> Among them, θ is the angle between the direction of the skeleton line of the arms and the positive direction of the longitudinal axis, (n _x , _ny , _nz ) is the rotation axis, which is obtained by cross multiplying the skeleton line of the arms and the positive direction of the longitudinal axis; Step 5-1-2-2: Multiply the dual-arm grid model by the rotation matrix, and erect the dual-arm grid model; Step 5-1-2-3: Use the horizontal plane to perform cross-cutting operations on the double-arm grid model, and use the method in step 2 to obtain a series of level sets; Step 5-1-2-4: If there are multiple level sets under a certain height plane, treat the vertices of all level sets as a point set, use the Graham algorithm to calculate the convex hull of the point set, and use the convex hull as the only level set below that height plane; Step 5-1-2-5: Calculate the perimeter of each level set, and use the midpoint of the level set with the smallest perimeter as the position of the wrist joint. 9. The 3D character model skeleton extraction method based on horizontal set center clustering according to claim 7, wherein the specific steps of the step 5-1-4 are as follows: Step 5-1-4-1: use the plane determined by the point with the largest vertical axis value of the skeleton line of both arms and the middle plane of the plane determined by the vertical axis value of the neck joint point to obtain the plane where the shoulder joint is located; Step 5-1-4-2: Use the direction of the skeleton line of both arms to intersect the plane where the shoulder joint is located to obtain an intersection point, which is the position of the shoulder joint; Step 5-1-4-3: Use the position of the shoulder joint and the position of the wrist joint point to obtain the initial position of the elbow joint through the position ratio of the elbow joint, shoulder joint and wrist joint point in the standard skeleton structure; Step 5-1-4-4: Take the point above the plane determined by the value of the longitudinal axis of the elbow joint as the upper arm part, re-fit the center point of the upper arm part with a straight line, and make the intersection of the straight line with the plane where the shoulder joint is located to get correction For the position of the shoulder joint, repeat 5-1-4-3 until the position of the shoulder joint does not change, and the obtained shoulder joint point is the final position of the shoulder joint.