CN115954105A - Whole body anatomical structure personalized modeling and posture deformation algorithm for human body biomechanical simulation - Google Patents
Whole body anatomical structure personalized modeling and posture deformation algorithm for human body biomechanical simulation Download PDFInfo
- Publication number
- CN115954105A CN115954105A CN202211636765.5A CN202211636765A CN115954105A CN 115954105 A CN115954105 A CN 115954105A CN 202211636765 A CN202211636765 A CN 202211636765A CN 115954105 A CN115954105 A CN 115954105A
- Authority
- CN
- China
- Prior art keywords
- human body
- point
- model
- skeleton
- joint
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Abstract
The invention discloses a whole body anatomical structure personalized modeling and posture deformation algorithm for human body biomechanics simulation, and belongs to the technical field of personalized modeling. The invention firstly obtains the relative rotation information among all joints when the human body moves through a neural network model, and realizes the posture deformation of the internal anatomical structure of the human body from bottom to top according to the hierarchical relationship of the joints of the human body and the skeleton topological structure. The mechanism of smooth deformation is introduced by removing the intersection after the posture adjustment through a cross algorithm and introducing an optimized skin algorithm, so that the authenticity of the personalized human body model during movement is ensured. And optimizing a model reconstruction result and mapping muscle force lines by combining the body circumference information, and finally constructing a personalized human body biomechanical simulation model containing a whole body anatomical structure, so that the internal anatomical structure of the human body is truly and reasonably deformed when the posture is deformed, and the method can be suitable for biomechanical simulation of different individuals and posture actions.
Description
Technical Field
The invention belongs to the technical field of personalized modeling, and particularly relates to a personalized modeling method of a whole body anatomical structure for a human body biomechanics simulation model and a posture deformation algorithm research.
Background
In recent years, the application demand of personalized human body modeling is increasing, and three-dimensional human body models combining human body internal anatomical structures are applied in many fields, such as anatomical teaching, dynamics simulation, electromagnetic radiation simulation and the like. Personalized human body modeling is not only widely applied in the fields of industrial simulation and medical treatment, but also gradually applied in daily life. With the continuous development of virtual reality technology, the application of virtual characters is increasing, and people hope to construct personalized virtual characters, simulate their own actions in real time, and link the virtual and the reality. For example, human body biomechanical simulation modeling of the personalized whole-body anatomical structure is carried out in real time according to posture transformation of a human body during motion, when no professional guidance is provided for home fitness or exercise training, motion evaluation is carried out according to different postures of the personalized human body model, and the motion evaluation is compared with the whole-body anatomical structure under standard motion, so that wrong motion and possible damage in motion can be pointed out, intelligent motion guidance is carried out in real time, and the daily motion healthy life of people is standardized.
A personalized modeling method for realizing a human motion biomechanics simulation model generally comprises the steps of wearing a special sensing device through a motion capture system and carrying out data acquisition through a multi-view camera. Although the traditional method can obtain high-quality human body model reconstruction and acquire some human body biomechanical characteristics, the traditional method needs a large amount of manual correction and time-consuming processing on data obtained by sampling a specific person, and is not suitable for daily life. The human body posture estimation and human body three-dimensional reconstruction are excellent, and a deep neural network which is formed in recent years is used for three-dimensional reconstruction of a human body in a video or a picture, and is also commonly used in the field of physical fitness. However, the human body reconstruction method focuses more on the external epidermis model, focuses on the posture action of the human body role, ignores the complex and various internal anatomical structures of the three-dimensional human body, lacks the deformation research on the internal anatomical structures, and cannot perform the biomechanical model modeling of the human body movement.
Disclosure of Invention
In order to solve the problems, the invention provides a whole body anatomical structure personalized modeling method and posture deformation algorithm research for a human body biomechanics simulation model. Relative rotation information among all joints during human body movement is obtained through a neural network model, posture transformation of an internal anatomical structure of a human body is realized from bottom to top according to the hierarchical relationship of the joints of the human body and a skeleton topological structure, and meanwhile, the authenticity during the movement of the personalized human body model is ensured through a mechanism that the crossing algorithm is used for removing the crossing after posture adjustment and the optimized skin algorithm is used for introducing smooth deformation. And mapping the muscle force line of the personalized human body model to construct a personalized human body biomechanics simulation model.
The technical scheme of the invention is as follows:
the method is used for personalized modeling of the whole body anatomical structure of a human body biomechanics simulation model and research of a posture deformation algorithm, and comprises the following steps:
step A, individualized posture deformation
Step A1, anatomical joint points
Firstly, defining 21 skeleton joint point positions of a motion skeleton of an individualized deformable digital person in a standard state according to human anatomy definition, wherein the joint point positions are pelvic bones, lumbar vertebrae, thoracic vertebrae, cervical vertebrae, right shoulders, right elbows, right wrists, right fingertips, left shoulders, left elbows, left wrists, left fingertips, right hips, right knees, right toes, left hips, left knees, left toes, left clavicles, right clavicles and tops of heads; on the basis of the positions of the skeleton joint points, simulating the normal rotation angle of the skeleton by combining anatomical prior knowledge of each skeleton joint point, judging the rationality of the result after the skeleton is rotated and transformed, and adjusting the skeleton joint points through modeling software; then, the adjusted bone joint points are subjected to the process, and the positions of the anatomical joint points are finally obtained;
step A2, skeleton topological structure
On the basis of the anatomical joint points determined in the step A1, determining inheritance relationships among the skeletal joint points according to a skeleton topological structure, and expressing joint hierarchical relationships by using parent joints and child joints; when the human body moves, the child joints only rotate relative to the father joints and do not translate, so that the Euclidean distance between the father joints and the child joints is constant, and the objective fact that the length of the human body skeleton is constant is met; confirming parent skeletons and child skeletons of all skeletons and root nodes and tail nodes of all skeletons from the root nodes of the skeletons by considering skeleton rotation from bottom to top; the child joint with the lowest generation index moves first and then moves relative to the father joint, and the child joint drives the child joint relative to the father joint, so that the change of the child joint is equivalent to accumulation; after all the bones rotate around the joints, the description of the corresponding posture and the action of the human body is realized;
step A3, removing cross correction
According to the skeleton topological structure in the step A2, obtaining a human body posture transformation result after all skeletons from a child skeleton to a father skeleton are transformed globally; because the joint point and the rotating shaft of the bone rotation can not strictly simulate the complex rotating and sliding process of the human body joint, the crossing between the adjacent bones can also be caused; or crossing of soft tissue of blood vessel with bone and skin; therefore, a de-intersection operation is required to be adopted, the possible curved surface intersection is eliminated, and a simulation available model is generated;
the method for carrying out de-intersection detection on the curved surfaces of all anatomical structures in the human body posture transformation result comprises the following steps: firstly, two curved surfaces needing cross detection are respectively set as S 1 And S 2 Both must ensure a closed mesh surface; traverse S 1 All the top vertices, and any vertex v ∈ S to be detected 1 Starting from the vertex v, making n rays, detecting each ray and S 2 Detecting whether each vertex is at S 2 An inner portion; if all n rays are associated with S 2 The number of the intersections is odd, and the vertex v is shown to be in S 2 Internal, otherwise external;
find S 1 If the detected object is between soft tissue and bone or between bone and bone, the point to be adjusted is S 1 Is centrally located in S 2 An inner point; if the detected object is between the soft tissue and the skinThe point to be adjusted is S 1 Is centrally located in S 2 An outer point; judging that the point needing to be adjusted is positioned on the curved surface S 2 Whether it is external or internal; after each point needing to be adjusted, S is made 2 If the foot falls into the triangular patch, recording the distance from the point to be adjusted to the foot, and selecting the point with the shortest distance in all the distances as the coordinate of the updated point of the point to be adjusted; the updated coordinates of each point to be adjusted are:
v * ←1.5(k-v * )
wherein v is * K is the corresponding foot when the point is the shortest distance from the target point;
after the point needing to be adjusted can not be detected, the curved surface S is aligned 1 And S 2 Performing triangular patch crossing detection to eliminate possible patch crossing; traverse S 1 All triangular patches are arranged, and t is set 1 Is S 1 Detecting three sides and S of any triangular patch 2 Any triangle patch t 2 The cross relationship of (a); as long as t 1 Any one of the three sides of (1) and t 2 Intersecting, then two triangular patches t are considered 1 And t 2 Intersecting; for S 1 Above all with S 2 The vertex of the intersected triangular patch needs to be readjusted until the intersected triangular patch cannot be detected;
step A4, skin optimization
Firstly, realizing a skinning result by a linear hybrid skinning method; binding the skin surface vertex obtained by statistical learning under the silent posture of the human body on each bone joint, wherein each skin surface vertex can be bound and attached to one or more bones (generally, at most four bones are attached); estimating all skin surface vertexes by selecting a least square method with constraint to generate a linear skin weight; according to the skeleton pose transformation and the obtained linear skinning weight, the position transformation of the vertex of the skin surface is expressed as:
wherein, v' i Is the transformed vertex position, N is the number of attached bones, w n,i Is a mixed weight matrix of vertices to the nth bone ni Is a skeleton transformation matrix, v i Is the initial position of the vertex;
at present, the skin methods widely used in experiments are Linear Blend skin (Linear Blend skin) and Dual Quaternion skin (Dual Quaternion skin), and the two skin methods have high calculation efficiency in practical application. On the basis of linear hybrid skin, a method for skin optimization through a rotation center is introduced, and the skin deformation result is optimized; calculating the rotation center of each skin surface vertex, and finding out points similar to the linear skin weight distribution in the skin surface vertices; calculating the rotation transformation in the skin deformation by using a dual quaternion skin transformation method and through a quaternion superposition rotation matrix; calculating translation transformation in skin deformation according to the calculated linear transformation result of the rotation center of the vertex of the skin surface and the result of transformation of the rotation matrix obtained by quadruple superposition; this process is represented as:
wherein, t is the translation amount,
is a linear transformation of the center of rotation, rv * Is a quadruple superimposed rotation matrix transformation;
and B, step B: the personalized modeling of the human body biomechanics simulation model is combined with the real body circumference size measured in the actual biomechanics simulation, so that the personalized modeling result of the human body biomechanics simulation model is more accurate;
step B1, three-dimensional human body model reconstruction
Firstly, respectively extracting the features of images of each frame in a photo or a video through a neural network model, regressing the spatial position coordinates of the anatomical joint points defined in the step A1 under different postures, calculating the rotation angle of each skeletal joint point relative to the father node thereof by combining the father node defined for each skeletal joint point by the skeleton topological structure in the step A2, and obtaining the relative transformation matrix of each skeleton under different human postures;
converting the relative transformation matrix of each skeleton into an absolute transformation matrix by the personalized posture deformation algorithm in the step A; converting the human body from the silent posture to the current posture according to the absolute transformation matrix of the skeleton and a human body mean template obtained by statistical learning, and finally performing cross correction and skinning result optimization on the posture transformation result to reconstruct a three-dimensional human body model of the moving personnel;
step B2, measuring body circumference
Measuring the body circumference based on the three-dimensional human body model of the moving person reconstructed in the step B1; firstly, defining position information passing through feature points when measuring each circumference of a human body on a human body epidermis mean value model obtained by statistical learning according to different measurement standards; interpolating a deformation result of the body circumference characteristic points in a three-dimensional space interpolation mode according to a deformation field constructed by the vertex displacement between the reconstructed three-dimensional human body model body surface of the moving personnel and the human body mean value model body surface;
as the body circumference characteristic points are distributed sparsely in the human body model, in order to ensure the interpolation effect, a plurality of points with the nearest distance of all the characteristic points in each circumference are taken and collected to form a point set, and the interpolation of the circumference characteristic points is realized by using Gaussian interpolation according to the deformation field of the vertex displacement of each circumference near point in the point set; calculating the absolute distance between two points is expressed as:
wherein d is a three-dimensional space point a (x) 1 ,y 1 ,z 1 ) And b (x) 2 ,y 2 ,z 2 ) The Euclidean distance between them; calculating the distance between all vertexes in the model and the current circumference characteristic pointAfter the Euclidean distances are arranged according to the distance increasing order, selecting vertex coordinates of a plurality of points with the nearest distance, combining the vertex coordinates to form a point set, and performing Gaussian interpolation;
then defining the vertex, pelvis and sole of the characteristic points, obtaining the height of the model according to the distance between the characteristic points after interpolation registration, obtaining the real proportion after model reconstruction by acquiring the actual height data of the sportsman, obtaining the posture information according to the actual weight data of the sportsman, and estimating the actual body circumference of the sportsman; on the basis of the estimated body circumference size, the body circumference size obtained by actual measurement in biomechanical simulation is combined, and the personalized modeling result is further optimized, so that the reconstructed human body model is more accurate;
step B3, mapping muscle force lines
In order to enable the constructed human body biomechanical simulation model to carry out biomechanical simulation, mapping of muscle force lines is carried out on the reconstructed three-dimensional human body model of the sportsman; according to a muscle-bone model in the existing biomechanics simulation experiment, the relative position relation of parent-child bones in a human anatomy model, the relative positions of a starting point and a starting point of muscle on each bone are arranged;
when mapping the muscle force lines, firstly, uniform resampling is carried out on a skeletal model obtained by statistical learning, based on a calibration point, the muscle-skeletal model point cloud is registered to a standard skeletal model of a digital human body model obtained by statistical learning through nonlinear space transformation, and the registration of the skeletal model and the mapping of the muscle force lines are completed.
The invention has the beneficial effects that: the invention carries out the construction of the individual human body biomechanics simulation model and the optimization algorithm research of the model posture deformation, can construct the individual biomechanics model for different individuals and is used for biomechanics simulation. Under the optimization algorithm of the posture deformation, no matter how large the angle of the posture joint is changed, the real motion deformation result of the personalized model from the internal anatomical structure to the external epidermis can be obtained. In addition, the body circumference of the sportsman is measured and a personalized human body biomechanics simulation model is constructed through the pictures and the videos, the physical condition and the actions in different postures are evaluated, wrong actions and possible damages in the sports are pointed out, and a reasonable exercise training proposal and intelligent exercise guidance are provided.
Drawings
Fig. 1 is a flow chart of the positioning method of the present invention.
FIG. 2 is a flow chart of the construction of the digital human body personalized posture deformation algorithm.
FIG. 3 is a flow chart of personalized modeling of a biomechanical simulation model of a human body.
Fig. 4 is a flow chart illustrating the posture deformation algorithm and the method for constructing the human body biomechanics simulation model according to the present invention by taking the video of the moving person as an example.
Detailed Description
The whole body anatomy structure personalized modeling method and the posture deformation algorithm research for the human body biomechanics simulation model are shown in figure 1. The method mainly comprises two parts: a digital human body personalized posture deformation algorithm; a personalized modeling method for a human body biomechanics simulation model. The present invention will be further described with reference to specific embodiments.
Step A, researching an individualized attitude deformation algorithm, as shown in figure 2.
Step A1, anatomical joint point definition
According to human anatomy definition, 21 skeleton joint point positions are defined for a motion skeleton of an individual deformable digital human in a standard state, wherein the 21 skeleton joint point positions are pelvic bones, lumbar vertebrae, thoracic vertebrae, cervical vertebrae, right shoulders, right elbows, right wrists, right fingertips, left shoulders, left elbows, left wrists, left fingertips, right hips, right knees, right toes, left hips, left knees, left toes, left clavicles, right clavicles and tops of heads.
On the basis of the joint point position defined in the prior art, the prior knowledge of each skeleton anatomical structure is combined to simulate the relative rotation between the skeletons, and the joint point position is adjusted by judging the relative rotation condition between the skeletons and using modeling software. And determining the position of the anatomical joint after testing whether the bone rotation transformation of the new joint is reasonable.
The number of the selected joint points is not fixed to 21, and the joint points can be increased or deleted automatically according to requirements and algorithm expression. The anatomical feature points determined in this step serve as the relative rotation centers of the respective joints at the time of the subsequent posture change.
Step A2, skeleton topological structure
On the basis of the definition of the joint points, the inheritance relationship between the relation points is determined according to the skeleton topological structure, and the joint hierarchical relationship is expressed by the father joint and the son joint. From the root node of the skeleton, the parent skeleton and the child skeleton of each skeleton and the root node and the tail node of each skeleton are confirmed from top to bottom.
When the posture is deformed, the child joint with the lowest generation hierarchy defined in the skeleton topological structure and the joint hierarchical relationship moves first, moves relative to the father joint and drives the child joint relative to the father joint, and the change of the child joint is accumulated; after all the bones rotate around the joints, the description of the corresponding posture and the action of the human body is realized;
step A3, removing cross correction
The method for detecting whether the patches between the curved surfaces are intersected and removing the intersections between the patches comprises the following steps:
firstly, two curved surfaces needing cross detection are respectively set as S 1 And S 2 Both are curved surfaces of organs in the whole body anatomy, which must be guaranteed to be a closed mesh surface in three-dimensional space. Traverse S 1 All the vertexes are checked, and any vertex v epsilon to be detected is set 1 Starting from v, making n rays, detecting each ray and S 2 Detecting whether each vertex is at S 2 Inside. If all n rays are associated with S 2 The number of the intersections is odd, and v is stated as S 2 Internal, otherwise external. All test cases can be satisfied by taking n =10 in the project.
Find S 1 The point in the middle of which adjustment is required. Taking blood vessels, bones, and skin as an example, if the object to be detected is between blood vessels and bones or between bones, the point to be adjusted is S 1 Is centrally located in S 2 An inner point; if the detected object is between a blood vessel and the skin, the point to be adjusted is S 1 Is centrally located in S 2 An outer point;judging that the point needing to be adjusted is positioned on the curved surface S 2 Whether it is external or internal; judging whether the point needing to be adjusted is located on the curved surface S 2 Is the interior or exterior of. For each point v to be adjusted * Through v * To S 2 If the foot falls inside the triangle, v is recorded * Distance to the drop foot. The shortest of all distances is taken as v * To S 2 Then the foot k corresponding to the shortest distance is taken as v * Target point of (1), then v * Should be updated as:
v * ←1.5(k-v * )
and when the point needing to be adjusted cannot be detected, performing patch crossing detection. Traverse S 1 All triangular patches are arranged, and t is set 1 Is S 1 Applying any one of the sheets, and detecting its three edges and S respectively 2 On any dough sheet t 2 The cross-relation of (c). As long as any one of three sides is connected with t 2 Intersect, then consider two t 1 And t 2 And (4) intersecting. For S 1 Above all and S 2 And the vertex of the intersected patch is readjusted until the intersected patch cannot be detected, and the de-intersection correction is finished.
Step A4, skin optimization
Firstly, realizing a skinning result by a linear hybrid skinning method; binding skin surface vertexes under the silent posture of the human body, which are obtained by statistical learning, to a bone posture, wherein each skin surface vertex can be bound and attached to one or more bones; the constrained least squares method is chosen to estimate all vertices, generating linear skinning weights, with the sum of the weights of all bones for one vertex being 1. From the skeletal pose transformation and the derived skinning weights, the position transformation of the model vertices can be expressed as:
wherein, v' i Is the transformed vertex position, N is the number of attached bones, w n,i Is vertex to vertexMixed weight matrix of n attached bones, G ni Is a skeleton transformation matrix, v i Is the initial position of the vertex.
On the basis of linear hybrid skin, a method for skin optimization through a rotation center is introduced, and the skin deformation result is optimized; calculating the rotation center of each skin surface vertex, and finding out points similar to the linear skin weight distribution in the skin surface vertices; calculating the rotation transformation in the skin deformation by using a dual quaternion skin transformation method and through a quaternion superposition rotation matrix; calculating translation transformation in skin deformation according to the calculated linear transformation result of the rotation center of the vertex of the skin surface and the result of transformation of the rotation matrix obtained by quadruple superposition; this process is represented as:
where t is the amount of translation and,
is a linear transformation of the center of rotation, rv * Is a rotation matrix transformation of a quadruple superposition.
And step B, personalized modeling of the human body biomechanics simulation model, as shown in figure 3.
The individual modeling of a biomechanical simulation model is carried out on the sports personnel, and comprises the steps of optimizing a model reconstruction result by body circumference measurement, mapping muscle force lines and the like. Firstly, detecting anatomical joint point information and joint rotation angles of the real-time movement postures of the moving personnel through a neural network model, and carrying out personalized three-dimensional human body model reconstruction according to the personalized human body posture deformation algorithm in the step A. Optimizing the result of the human body model reconstruction by combining the information of the degree of enclosure actually measured for the moving personnel and the result of the body degree of enclosure measurement estimated for the reconstructed human body model; and then mapping the muscle force line to realize the personalized modeling of the human body biomechanics simulation model.
Step B1, rebuilding three-dimensional human body model
The Frankmocap network and the vibe network are used as basic models for three-dimensional human body model reconstruction and are respectively used for extracting feature points of the pictures and the videos to obtain relative transformation information among different human body posture joints in the pictures or the videos. And then, performing de-cross correction and skinning result optimization by adopting a researched attitude deformation algorithm to obtain the three-dimensional human body model reconstruction of the moving personnel.
Step B2, body circumference measurement
Firstly, defining position information passing through feature points when measuring each circumference of a human body on a human body epidermis mean value model obtained by statistical learning according to different measurement standards, and taking the position information as the definition of each circumference size;
b, interpolating a deformation result of the body circumference characteristic points in a three-dimensional space interpolation mode according to a deformation field constructed by vertex displacement between the three-dimensional human body model body surface of the moving person reconstructed in the step B1 and the human body mean value model body surface, and measuring the body circumference;
as the body circumference characteristic points are sparsely distributed in the human body model, in order to ensure the interpolation effect, a point set with the shortest Euclidean distance with the current circumference characteristic points is selected in each circumference area, the deformation vector of the vertex in the point set is used for interpolating the circumference,
then defining the vertex, pelvis and sole of the characteristic points, obtaining the height of the model according to the distance between the characteristic points after interpolation registration, obtaining the real proportion after model reconstruction by acquiring the actual height data of the sportsman, obtaining the posture information according to the actual weight data of the sportsman, and estimating the actual body circumference of the sportsman; on the basis of the estimated body circumference size, the body circumference size obtained by actual measurement in biomechanical simulation is combined, and the personalized modeling result is further optimized, so that the reconstructed human body model is more accurate;
step B3, mapping muscle force lines
Firstly, according to a musculoskeletal model in the existing biomechanical simulation experiment, the relative positions of the paternal-child relationship and position information among bones, a muscle starting point and a generation starting point on the bones in the musculoskeletal model are extracted. Because the number of the vertexes and patches of the bone model obtained by statistical learning is more than that of the muscle-bone model in biomechanical simulation, firstly, the bone model obtained by statistical learning is uniformly resampled; the bones of the human anatomical model are then registered to a standard bone model of the digital human model by a non-linear spatial transformation. Finding out a plurality of groups of calibration points of the same part in the anatomical model and the standard template, and mapping the cloud of the point to be registered to the corresponding spatial position in the target point cloud according to the spatial transformation between the calibration points to complete the registration of the skeleton model and the mapping of the muscle force line.
Claims (1)
1. A whole body anatomy structure personalized modeling and posture deformation algorithm for human body biomechanical simulation is characterized by comprising the following steps:
step A, personalized posture deformation
Step A1, anatomical joint points
Firstly, defining 21 skeleton joint point positions of a motion skeleton of an individualized deformable digital person in a standard state according to human anatomy definition, wherein the joint point positions are pelvic bones, lumbar vertebrae, thoracic vertebrae, cervical vertebrae, right shoulders, right elbows, right wrists, right fingertips, left shoulders, left elbows, left wrists, left fingertips, right hips, right knees, right toes, left hips, left knees, left toes, left clavicles, right clavicles and tops of heads; on the basis of the positions of the skeleton joint points, simulating the normal rotation angle of the skeleton by combining anatomical prior knowledge of each skeleton joint point, judging the rationality of the result after the skeleton is rotated and transformed, and adjusting the skeleton joint points through modeling software; then, the adjusted bone joint points are subjected to the process, and the positions of the anatomical joint points are finally obtained;
step A2, skeleton topological structure
On the basis of the anatomical joint points determined in the step A1, determining inheritance relationships among the skeletal joint points according to a skeleton topological structure, and expressing joint hierarchical relationships by using parent joints and child joints; when the human body moves, the child joints only rotate relative to the father joints and do not translate, so that the Euclidean distance between the father joints and the child joints is constant, and the objective fact that the length of the human body skeleton is constant is met; confirming parent skeletons and child skeletons of all skeletons and root nodes and tail nodes of all skeletons from the root nodes of the skeletons by considering skeleton rotation from bottom to top; the child joint with the lowest generation score moves first and then moves relative to the father joint, and the child joint is driven relative to the father joint, and the change of the child joint is equivalent to accumulation; after all the bones rotate around the joints, the description of the corresponding posture and the action of the human body is realized;
step A3, removing cross correction
According to the skeleton topological structure in the step A2, obtaining a human body posture transformation result after all skeletons from a child skeleton to a father skeleton are transformed globally; because the joint point and the rotating shaft of the bone rotation can not strictly simulate the complex rotating and sliding process of the human body joint, the crossing between the adjacent bones can also be caused; or crossing of soft tissue of blood vessel with bone and skin; therefore, a de-intersection operation is required to be adopted, the possible generated curved surface intersection is eliminated, and a simulation available model is generated;
the method for carrying out de-intersection detection on the curved surfaces of all anatomical structures in the human body posture transformation result comprises the following steps: firstly, two curved surfaces needing cross detection are respectively set as S 1 And S 2 Both must ensure a closed mesh surface; traverse S 1 All the top vertices, and any vertex v ∈ S to be detected 1 Starting from the vertex v, making n rays, detecting each ray and S 2 Detecting whether each vertex is at S 2 An inner portion; if all n rays are associated with S 2 The number of the intersections is odd, and the vertex v is shown to be in S 2 Internal, otherwise external;
find S 1 If the detected object is between soft tissue and bone or between bone and bone, the point to be adjusted is S 1 Is centrally located in S 2 An inner point; if the detected object is between the soft tissue and the skin, the point needing to be adjusted is S 1 Is centrally located in S 2 An outer point; judging that the point needing to be adjusted is positioned on the curved surface S 2 Whether it is external or internal; after each point needing to be adjusted, S is made 2 All of (2)If the foot falls into the triangular patch, recording the distance from the point to be adjusted to the foot, and selecting the point with the shortest distance in all the distances as the coordinate of the updated point of the point to be adjusted; the updated coordinates of each point to be adjusted are:
v * ←1.5(k-v * )
wherein v is * K is the corresponding foot when the shortest distance of the target point is the point needing to be adjusted;
after the point needing to be adjusted can not be detected, the curved surface S is aligned 1 And S 2 Performing triangular patch crossing detection to eliminate possible patch crossing; traverse S 1 All triangular patches are arranged, and t is set 1 Is S 1 Detecting three sides and S of any triangular patch 2 Any triangle patch t 2 The cross relationship of (a); as long as t 1 Any one of three sides of the T-bar is connected with t 2 Intersecting, then two triangular patches t are considered 1 And t 2 Intersecting; for S 1 Above all with S 2 The vertex of the intersected triangular patch needs to be readjusted until the intersected triangular patch cannot be detected;
step A4, skin optimization
Firstly, realizing a skinning result by a linear hybrid skinning method; binding a skin surface vertex obtained by statistical learning under a human body silent posture on each bone joint, wherein each skin surface vertex can be bound and attached to one or more bones; estimating all skin surface vertexes by selecting a least square method with constraint to generate a linear skin weight; according to the bone pose transformation and the obtained linear skinning weight, the position transformation of the vertex of the skin surface is expressed as:
wherein, v' i Is the transformed vertex position, N is the number of attached bones, w n,i Is a mixed weight matrix of vertices to the nth attached bone, G ni Is a skeleton transformation matrix, v i Is the initial position of the vertex;
on the basis of linear hybrid skin, a method for skin optimization through a rotation center is introduced, and the skin deformation result is optimized; calculating the rotation center of each skin surface vertex, and finding out points similar to the linear skin weight distribution in the skin surface vertices; calculating the rotation transformation in the skin deformation by using a dual quaternion skin transformation method and through a quaternion superposition rotation matrix; calculating translation transformation in skin deformation according to the calculated linear transformation result of the rotation center of the vertex of the skin surface and the result of transformation of a rotation matrix obtained by quadruple superposition; this process is represented as:
wherein, t is the translation amount,
is a linear transformation of the center of rotation, rv * Is a quadruple superimposed rotation matrix transformation;
and B: the personalized modeling of the human body biomechanics simulation model is combined with the real body circumference size measured in the actual biomechanics simulation, so that the personalized modeling result of the human body biomechanics simulation model is more accurate;
step B1, three-dimensional human body model reconstruction
Firstly, respectively extracting the features of images of each frame in a photo or a video through a neural network model, regressing the spatial position coordinates of the anatomical joint points defined in the step A1 under different postures, calculating the rotation angle of each skeletal joint point relative to the father node thereof by combining the father node defined for each skeletal joint point by the skeleton topological structure in the step A2, and obtaining the relative transformation matrix of each skeleton under different human postures;
converting the relative transformation matrix of each skeleton into an absolute transformation matrix by the personalized posture deformation algorithm in the step A; converting the human body from the silent posture to the current posture according to the absolute transformation matrix of the skeleton and a human body mean template obtained by statistical learning, and finally performing cross correction and skinning result optimization on the posture transformation result to reconstruct a three-dimensional human body model of the moving personnel;
step B2, measuring body circumference
Measuring the body circumference based on the three-dimensional human body model of the moving person reconstructed in the step B1; firstly, defining position information passing through feature points when measuring each circumference of a human body on a human body epidermis mean value model obtained by statistical learning according to different measurement standards; interpolating a deformation result of the body circumference characteristic points in a three-dimensional space interpolation mode according to a deformation field constructed by the vertex displacement between the reconstructed three-dimensional human body model body surface of the moving personnel and the human body mean value model body surface;
as the body circumference characteristic points are sparsely distributed in the human body model, in order to ensure the interpolation effect, a plurality of points with the nearest distance of all the characteristic points in each circumference are taken and are integrated to form a point set, and the interpolation of the circumference characteristic points is realized by using Gaussian interpolation according to the deformation field of the vertex displacement of each circumference near point in the point set; calculating the absolute distance between two points is expressed as:
wherein d is a three-dimensional space point a (x) 1 ,y 1 ,z 1 ) And b (x) 2 ,y 2 ,z 2 ) The Euclidean distance between them; calculating Euclidean distances between all vertexes in the model and the current girth feature points, arranging the Euclidean distances according to the distance increasing order, selecting vertex coordinates of a plurality of points with the nearest distances, combining the vertex coordinates to form a point set, and performing Gaussian interpolation;
then defining the vertex, pelvis and sole of the characteristic points, obtaining the height of the model according to the distance between the characteristic points after interpolation registration, obtaining the real proportion after model reconstruction by acquiring the actual height data of the sportsman, obtaining the posture information according to the actual weight data of the sportsman, and estimating the actual body circumference of the sportsman; on the basis of the estimated body circumference size, the body circumference size obtained by actual measurement in biomechanical simulation is combined, and the personalized modeling result is further optimized, so that the reconstructed human body model is more accurate;
step B3, mapping muscle force lines
In order to enable the constructed human body biomechanics simulation model to carry out biomechanics simulation,
mapping muscle force lines on the reconstructed three-dimensional human body model of the sportsman; according to a muscle-bone model in the existing biomechanics simulation experiment, the relative position relation of parent-child bones in a human anatomy model, the relative positions of a starting point and a starting point of muscle on each bone are arranged;
when mapping muscle force lines, firstly, the bone model obtained by statistical learning is uniformly resampled,
and registering the muscle-skeleton model point cloud to a standard skeleton model of the digital human body model obtained by statistical learning through nonlinear space transformation based on the calibration point, and finishing the registration of the skeleton model and the mapping of the muscle force line.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211636765.5A CN115954105A (en) | 2022-12-16 | 2022-12-16 | Whole body anatomical structure personalized modeling and posture deformation algorithm for human body biomechanical simulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211636765.5A CN115954105A (en) | 2022-12-16 | 2022-12-16 | Whole body anatomical structure personalized modeling and posture deformation algorithm for human body biomechanical simulation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115954105A true CN115954105A (en) | 2023-04-11 |
Family
ID=87282089
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211636765.5A Pending CN115954105A (en) | 2022-12-16 | 2022-12-16 | Whole body anatomical structure personalized modeling and posture deformation algorithm for human body biomechanical simulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115954105A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117808855A (en) * | 2024-02-29 | 2024-04-02 | 新西旺智能科技(深圳)有限公司 | Target alignment method and system based on visual image |
-
2022
- 2022-12-16 CN CN202211636765.5A patent/CN115954105A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117808855A (en) * | 2024-02-29 | 2024-04-02 | 新西旺智能科技(深圳)有限公司 | Target alignment method and system based on visual image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7061694B2 (en) | Image processing methods and equipment, imaging equipment, and storage media | |
| Kadleček et al. | Reconstructing personalized anatomical models for physics-based body animation | |
| Delingette et al. | Craniofacial surgery simulation testbed | |
| Zhu et al. | An efficient human model customization method based on orthogonal-view monocular photos | |
| CN1864074B (en) | Method for determining patient-related information, control apparatus and magnetic resonance tomography instrument | |
| CN103824326B (en) | Dynamic human body three-dimensional modeling method | |
| CN110335358A (en) | The personalized deformation method of deformable Digital Human anatomical model | |
| Corazza et al. | Automatic generation of a subject-specific model for accurate markerless motion capture and biomechanical applications | |
| CN101882326A (en) | Three-dimensional craniofacial reconstruction method based on overall facial structure shape data of Chinese people | |
| CN107016666A (en) | The post processing of image method of pelvis CT three-dimensional reconstructions based on coordinate-system | |
| JP6741305B2 (en) | Vertebral posture estimation device | |
| Dicko et al. | Anatomy transfer | |
| Bauer et al. | Anatomical augmented reality with 3D commodity tracking and image-space alignment | |
| CN109118455B (en) | Ancient human skull craniofacial interactive restoration method based on modern soft tissue distribution | |
| CN107993293A (en) | A kind of three-dimensional interactive anatomic teaching method for including Chinese population difference | |
| JP2022501732A (en) | Image processing methods and devices, image devices and storage media | |
| CN115954105A (en) | Whole body anatomical structure personalized modeling and posture deformation algorithm for human body biomechanical simulation | |
| Nedel et al. | Anatomic modeling of deformable human bodies | |
| Beylot et al. | 3D interactive topological modeling using visible human dataset | |
| Schleicher et al. | BASH: Biomechanical Animated Skinned Human for Visualization of Kinematics and Muscle Activity. | |
| Assassi et al. | From MRI to anatomical simulation of the hip joint | |
| Chakravarthi et al. | Real-time human motion tracking and reconstruction using imu sensors | |
| CN107085866A (en) | The analog simulation method and system of a kind of straight bone CT images of human body | |
| Bond et al. | Series of 4D adult XCAT phantoms for imaging research and dosimetry | |
| Robertini et al. | Capture of arm-muscle deformations using a depth-camera |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |