CN107578821A - A kind of real-time efficiently GPU rendering intents being used in system of virtual operation - Google Patents
A kind of real-time efficiently GPU rendering intents being used in system of virtual operation Download PDFInfo
- Publication number
- CN107578821A CN107578821A CN201710748283.1A CN201710748283A CN107578821A CN 107578821 A CN107578821 A CN 107578821A CN 201710748283 A CN201710748283 A CN 201710748283A CN 107578821 A CN107578821 A CN 107578821A
- Authority
- CN
- China
- Prior art keywords
- subdivision
- scalpel
- tissue
- node
- summit
- 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.)
- Granted
Links
Landscapes
- Image Generation (AREA)
Abstract
The present invention is applied to Effective Pictures By A Computer field, there is provided a kind of real-time efficiently GPU rendering intents, methods described being used in system of virtual operation comprises the following steps：A, the Catmull Clark master data grids of tissue and scalpel are inputted；B, the data structure for showing characteristic surface subdivision level is established；C, the related data that surface subdivision needs is calculated in micro-stepping control tinter；D, reference coordinate position is calculated, carries out surface subdivision；E, final subdivision curved surface is obtained in surface subdivision calculates tinter.F, when scalpel and tissue collide, the tissue deformation of dynamic fine dimension is realized.The present invention realizes the Rendering of the real-time high-efficiency needed in system of virtual operation.
Description
Technical field
The present invention relates to Effective Pictures By A Computer field, it is proposed that a kind of new side for the real-time rendering being used in system of virtual operation
Method, the purpose is to make tissue obtain finer surface effect, when scalpel and tissue collide, accelerate real
When rendering speed.
Background technology
Surface subdivision is an infusive new features in OpenGL4.0, and curved surface is mainly used in solving 3D models
Static attribute, including their fineness and polygon quantity.It is exactly specifically when our close-ups one are complicated
During model (such as face), it is therefore desirable to be able to see all details (such as fold of skin) of this model, this is virtual hand
A characteristic required in art system environments.
Subdivision surface has been used for film making for many years.Due to the flexibility that they are provided in modeling, they have been drilled
Become the standard that surface represents industry.With the increase of the demand of the more rich image to increasing visual details, phase
The method for hoping film quality as real-time rendering, subdivision table can be used in content creation tool and interactive video games
Face.
Ideally, it is intended that primitive is rendered to support using subdivision surface as the one-level on GPU, similar to point,
Line and triangle etc..
Subdivision curved surface hardware only support can direct subdividing parametric surface, but directly carry out surface subdivision cost be probably very
High.
In the real-time rendering of virtual operation, when scalpel contacts tissue, the deformation of tissue, which renders, to keep
Speed must quickly on the premise of certain fineness, and this requires the high efficiency of Rendering algorithms.
Key issue is that physical analogy and collision detection are generally run on CPU, and the morphology storage rendered
On GPU.This needs CPU side to access grid data from GPU, so that gauging surface deforms.It will be changed in addition, physical update is related to
GPU is passed back on surface geometry afterwards.Because the bandwidth of CPU-GPU rambus and delay limit, which results in important operation
When expense and have impact on the performance of Current hardware framework.
In summary, existing Rendering is in actual use, it is clear that inconvenience and defect be present, it is therefore necessary to changed
Enter.
The content of the invention
For it is above-mentioned the defects of, rendered it is an object of the invention to provide a kind of real-time GPU for system of virtual operation
New method, significantly reduce CPU-GPU memory Is/O, it is allowed to more preferable frame rate, and realize that the surface of dynamic fine dimension becomes
Shape.
To achieve these goals, the present invention provides a kind of real-time GPU rendering intents for system of virtual operation, and it is special
Sign is that methods described comprises the following steps：
A. using the Catmull-Clark master datas grid of tissue and scalpel as input, including topological relation,
Features tab, the set on basic summit
B. a kind of data structure for showing characteristic surface subdivision level is created, this includes：
B1. the quad-tree structure of surface subdivision level can be reflected
The corresponding recursion step of its internal node (I), leaf node correspondingly generate final vertex position.Leaf node is also
Regular node and boundary node (R), sharp node (C), singular point (E) and terminal node (T) should be included.
Regular node and boundary node (R), will be each in its 16 control points if recursion generation rule face
Individual template is added to template list, and generates the regular node with reference to these templates.
Sharp node (C) can eliminate sharp feature by recursion
For singular point (E) if we have reached the limitation of subdivision depth, final position can directly be corresponded to by not reaching
Sub- face, then we create a singular point.Such node corresponds to the corner in basic grid face, and this is one special
Summit (EV), so we calculate the template in the extreme position in the corner and two tangent lines, and add them to template
In list.
Terminal node (T) is in each subdivision rank, and quaternary tree will have three regular nodes and an internal node, by most
The terminal node of whole rank terminates.
B2. the ordered list for controlling point template needed for surface subdivision is created
Control point is encoded to the weighting sum on a ring neighborhood summit by template, and with a weight array representation template.
B3. 1 ring neighborhood summit can be directly read in internal memory, and uses 16*16 matrix (weight relationship square
Battle array) each ring summit of storage weight, to record pair between topological relation and weight and control point between summit
It should be related to.
The edge in each basic grid face of micro-stepping control shader computations that C. we use in the subdivision stage, internal table
The factor, and the control point needed for surface subdivision are segmented in face, and it includes following feature：
C1. each control point for being subdivided face is calculated by recursive algorithm, and stored into the respective nodes of quaternary tree, together
Shi Jilu surface subdivision levels
C2. it is stored in what is provided by OpenGL4.x by all surface subdivision factor of micro-stepping control shader computations
The floating-point texture of GLRGB32F_ARB forms, it has four colors components (RGBA) and 32 floating number precision, uses red group
Part (high position) storage surface segments the factor, 24 (low levels) storage surface subdivision grade (LTs) behind use
C3.LTs is calculated by formula (1), and wherein f is the corresponding surface subdivision factor
C4. the corresponding weight in 1 ring summit and its control point is read from weight relationship matrix, and calculates their convolution
Obtain the reference position of subdivided vertices.
D. in the gauging surface summit stage, position and its normal vector and from thin at control point are obtained by traveling through quaternary tree
The summit reference position of sub-control tinter output calculates tinter by surface subdivision and obtains the top that can finally render as input
Point position and its normal vector and its topological relation on all summits, its feature are as follows:
D1. the traversal of quaternary tree be not with recurrence but iteration, it is different for traversing different node processing modes.
D11. all child nodes are found by traveling through internal node.
D12. when the child nodes traversed are R nodes, control point position and its normal vector of its record are exported, by song
Line Integral calculates tinter and is mapped to the vertex position and normal vector that can finally render.
D13. when the child nodes traversed are E nodes, surface subdivision calculates the top that tinter directly exports its record
Point position and its normal vector.
D14. when the child nodes traversed during T nodes, traversal is terminated.
D2. surface subdivision calculate all vertex position rendered of tinter final output and normal vector and summit it
Between topological relation, afterwards can with routine rendering pipeline opposite vertexes rendered.
E. when scalpel and tissue collide, for the high frequency distortions of collision area, we use a base
It is as follows in the fine surface deformation algorithm of the algorithm realization dynamic run on GPU, its feature：
E1. our two colliding objects are that scalpel (does not produce deformation, using common triangular mesh as defeated respectively
Enter) and tissue (deformable objects, input being used as using subdivision curved surface caused by preceding step).
E2. we are that scalpel and tissue calculate orientation bounding box, when scalpel collides with tissue,
Two articles bounding box is overlapped, and using real-time binary system voxelization method, voxelization is carried out to the lap of scalpel, and
And from the voxel for obtaining scalpel lap, shape when tissue is collided with scalpel is determined, and produce phase therewith
The displacement control point of matching
E3. displacement control point is released from scalpel overlapping region, deformation and corresponding control point required for producing
Change in location.
Create a kind of data structure for showing characteristic surface subdivision level, it is characterised in that the step B1 is specifically included：
R nodes are stored in frame buffer by a kind of particular form, and three components of X, Y, Z storage control point position have 12 altogether
Individual byte, N are used for storing normal vector, take 12 bytes altogether.The benefit being stored in frame buffer is that GPU can be carried out at a high speed to it
Read write.
Create a kind of data structure for showing characteristic surface subdivision level, it is characterised in that the step B3 is specifically included：
It is 16*16 matrix T, with the elements in a main diagonal Tij(i=j) i-th of 1 ring neighborhood vertex weights are recorded,
With off diagonal element Tij(i ≠ j) records the topological relation between i-th of summit and j-th of summit, is represented with 0 without topology
Relation, represent topological relation be present with 1.
Establishment one is specific based on the fine surface deformation algorithm of the algorithm realization dynamic run on GPU, the step E1
Including：
Scalar value B-spline coefficient is stored in tile-base texture formats, and precalculate each tile borders and
The adjoining pointer in corner, and only the use of dynamic GPU memory managements is that the patch deformed distributes CPU internal memories.
Establishment one is specific based on the fine surface deformation algorithm of the algorithm realization dynamic run on GPU, the step E2
Including：
In order to determine that fine dimension deforms, by calculating scalpel and tissue the border of scalpel and tissue
Lap (hereinafter lap), scalpel includes in lap by voxelization, specific steps：
A. by calculating OBB bounding boxs for scalpel and tissue to obtain overlapping region, and established in GPU internal memories
One rejecting decision-making buffering area
B. rejecting pretreatment is then carried out.First AABB is calculated for each patch in the space by lap
Bounding box, will using the convex closure attribute (self-adapting subdividing curved surface has this attribute) of curved surface in order to obtain close patch borders
All B-spline patch are converted to Bezier base.Then displacement skew is incorporated to by calculating normal cone, at the same institute is in need
The patch of rejecting performs parallelly compressed on decision-making buffering area is rejected
C. the carry out voxelization to lap.The intersecting OBB regions of non-zero to scalpel, use binary system voxelization
Method, by the patch information selected before in rejecting decision-making buffer compression, exclusion that can be quickly takes lap
Patch, with speed-up computation.
Establishment one is specific based on the fine surface deformation algorithm of the algorithm realization dynamic run on GPU, the step E3
Including：
All patch of the human tissue surface of lap will be replaced so that their bodies no longer with scalpel
Element is intersecting, then by all control points of the B-spline (patch not being removed) of all displacements of tissue along bottom surface
Negative normal direction movement, it is known that leave deformation voxel, and calculate the position in its final world coordinate system.
The present invention is rendered by a kind of new method to the scalpel in system of virtual operation and tissue, and it collects
Into into hardware subdivision pipeline.The recursive surface subdivision algorithm of quaternary tree is carried out by only each input basic grid, we
The simplicity and efficiency rendered is improved, and the present invention is proposed using matrix to record topological relation and the top between summit
Point weight, realize single data structure while there is the innovation hair method of multiple functions, not only increase arithmetic speed, also speed up
The fluency rendered, and the robustness of program.CPU physically simulated calculations and processing low frequency deformation are finally separating out, significantly
Reduce CPU-GPU memory Is/O, it is allowed to higher frame rate, and realize the areal deformation of dynamic fine dimension.
Brief description of the drawings
Fig. 1 is the FB(flow block) according to the efficient GPU rendering intents for real-time surface subdivision of the embodiment of the present invention
Fig. 2 is the schematic diagram for the input element figure being related in the embodiment of the present invention
Fig. 3 is a kind of storage form schematic diagram for the data structure being related in the embodiment of the present invention
Fig. 4 is a kind of storage organization schematic diagram being related in the embodiment of the present invention
Fig. 5 is a kind of data store organisation (weight relationship matrix) schematic diagram being related in the embodiment of the present invention
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, it is right below in conjunction with drawings and Examples
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Step (1) uses Fig. 2 institutes representation model (including a thigh model and leader's art in the first step of the present invention
Cutting die type) it is used as basic model.
The information of step (1-1) reading model includes the topological relation on each summit, features tab, basic summit.
Data above is passed in vertex shader by step (1-2), and summit is processed into surface in vertex shader.
We do not set conversion of the local coordinate to Viewport coordinate in vertex shader, because on this summit
In color device, because surface subdivision is not carried out also, summit and imperfection.
Step (2) creates a kind of data structure for showing characteristic surface subdivision level to every sub- face.
This data structure includes the quad-tree structure that can reflect surface subdivision level, and it is to being created by recurrence
The record of one hierarchical data of surface subdivision step；And for controlling the ordered list of point template needed for surface subdivision.
For step (3) in the subdivision stage, micro-stepping control tinter that we use calculates each facilities network grid face
Edge, the internal subdivision surfaces factor, and the control point needed for surface subdivision, it is comprised the following steps that：Here we will be every
A series of control points required for individual sub- surface are referred to as patch.
Step (3-1) we using OpenGL built-in function tell rendering pipeline each patch that input is by 16 set of vertices
Into its specific function is：
glPatchParameteri(GL_PATCH_VERTICES,16)
Step (3-2) then we by recursive algorithm in order to it is each be subdivided face calculate required for control point, and will
Control point information storage records surface subdivision level into the respective nodes of quaternary tree, and we are by according to Fig. 3 storage side
Formula is stored.
Step (3-3) once, will all produce a subdivision level, we are carried out using internal node (I) for every recurrence
Each recursive hierarchical caused by record.
Step (3-4) generates a regular face when recursion, and we will be every in 16 control points corresponding to it
The template of one is added to template list, and generates the regular node (R) of these templates.
R nodes are stored in frame buffer by Fig. 4 modes, and three components of X, Y, Z storage control point position have 12 words altogether
Section, N are used for storing normal vector, take 12 bytes altogether.It is that GPU can be carried out to it to store them in frame buffer and bring benefit
High Su Du writes.
Step (3-5) we sharp feature can also be eliminated by continuous recursion.
For step (3-6) if we have reached the limitation of subdivision depth, final position can directly be corresponded to by not reaching
Sub- face, then we create a singular point (E).Such node corresponds to the corner in basic grid face, and this is a spy
Different summit (EV), so we calculate the template in the extreme position in the corner and two tangent lines, and add them to mould
In panel table.
For each subdivision rank, the quaternary tree that we are established will all have three regular nodes and an internal section
Point, and terminal node (T) is established on final level is other, represent that subdivision terminates.
We are encoded to control point using template the weighting sum on one ring neighborhood summit to step (3-7), and with one
Weight array representation template.
We can directly read 1 ring neighborhood summit in internal memory, and use 16*16 matrix (weight relationship
Matrix, shown in Fig. 5) each ring summit of storage weight, to record topological relation and weight and the control between summit
Corresponding relation between point.
One 16*16 of weight relationship matrix matrix T, with the elements in a main diagonal Tij(i=j) i-th of 1 ring neighborhood tops are recorded
Point weight, with off diagonal element Tij(i ≠ j) records the topological relation between i-th of summit and j-th of summit, represents do not have with 0
There is topological relation, represent topological relation be present with 1.
Step (3-8) is stored in using all surface subdivision factor of micro-stepping control shader computations and carried by OpenGL4.x
The floating-point texture of the GLRGB32F_ARB forms of confession, it has four colors components (RGBA) and 32 floating number precision, use red
Colour cell part (high position) storage surface segments the factor, 24 (low levels) storage surface subdivision grade (LTs) behind use
Step (3-9) calculates LTs, and we are calculated using formula (1), and wherein f is the corresponding surface subdivision factor
Step (3-10) reads the corresponding weight in 1 ring summit and its control point from weight relationship matrix, and calculates it
Convolution obtain the reference position of subdivided vertices.
Step (4) surface subdivision calculate tinter in, we by travel through quaternary tree obtain control point position and its
Normal vector and the summit reference position conduct input from the output of micro-stepping control tinter, and tinter is calculated by surface subdivision and obtained
It is as follows to the vertex position and its normal vector and its topological relation on all summits that can finally render, specific implementation step:
The traversal of step (4-1) our team quaternary trees is not to use recursive algorithm, but iterative strategy, for traversing difference
Node, its processing mode is different：
Step (4-2) finds all child nodes by traveling through internal node
Step (4-3) exports control point position and its method for its record when the child nodes that we traverse are R nodes
Vector, the vertex position and normal vector that can finally render are mapped to by calculation of curved surface integral tinter
When the child nodes that we traverse are E nodes, surface subdivision calculates tinter and directly exports it step (4-4)
The vertex position and its normal vector of record
Step (4-5) during T nodes, terminates traversal when the child nodes that we traverse.
Step (5) surface subdivision calculate in all subdivision surface of tinter final output vertex position and normal vector and
Topological relation between summit
For step (6) when scalpel and tissue collide, for the high frequency distortions of collision area, we use one
It is individual that the fine surface deformation algorithm of dynamic is realized based on the algorithm run on GPU,
Our two colliding objects of step (6-1) are that scalpel (does not produce deformation, uses common triangular mesh respectively
As input) and tissue (deformable objects, input being used as using subdivision curved surface caused by preceding step), by scalar value B
Spline coefficients are stored in tile-base texture formats, and precalculate the adjoining pointer on each tile borders and corner, and
And only distribute CPU internal memories using dynamic GPU memory managements for the patch deformed.
We are that scalpel and tissue calculate orientation bounding box to step (6-2), when scalpel and tissue occur
During collision, two articles bounding box is overlapped, and using real-time binary system voxelization method, body is carried out to the lap of scalpel
Elementization, and from the voxel for obtaining scalpel lap, shape when tissue is collided with scalpel is determined, and produce
Matched displacement control point.In order to determine that fine dimension deforms, by being calculated the border of scalpel and tissue
Scalpel and the lap of tissue (hereinafter lap), scalpel in lap by voxelization, specifically
Step includes：
Step (6-2-1) for scalpel and tissue by calculating OBB bounding boxs to obtain overlapping region, and in GPU
A rejecting decision-making buffering area is established in internal memory
Step (6-2-2) then carries out rejecting pretreatment.It is first each patch in the space by lap
AABB bounding boxs are calculated, in order to obtain close patch borders, (self-adapting subdividing curved surface has using the convex closure attribute of curved surface
This attribute) all B-spline patch are converted into Bezier base.Then displacement skew, while handle are incorporated to by calculating normal cone
Rejecting in need patch reject decision-making buffering area on perform it is parallelly compressed
Carry out voxelization of the step (6-2-3) to lap.The intersecting OBB regions of non-zero to scalpel, use two
System voxelization method, it is heavy by the patch information selected before in rejecting decision-making buffer compression, exclusion expense that can be quickly
The patch of folded part, with speed-up computation.
Step (6-3) is released control point is replaced from scalpel overlapping region, deformation required for producing and corresponding
The change in location at control point.All patch of the human tissue surface of lap will be replaced so that they no longer with hand
The voxel of art knife intersects, then by all control point edges of the B-spline (patch not being removed) of all displacements of tissue
The negative normal direction movement of bottom surface, it is known that leave deformation voxel, and calculate the position in its final world coordinate system.
Can carries out conventional render to these summits to step (7) now.
What the present invention did not elaborated partly belongs to techniques well known.
Although the illustrative embodiment of the present invention is described above, in order to the technology people of this technology neck
Member understands the present invention, it should be apparent that the invention is not restricted to the scope of embodiment, to the ordinary skill of the art
For personnel, as long as various change, in the scope of the present invention that appended claim limits and determines, these changes are aobvious
And be clear to, all are using the innovation and creation of present inventive concept in the row of protection.
Claims (6)
1. a kind of real-time efficiently GPU rendering intents being used in system of virtual operation, it is characterised in that comprise the following steps：
A. using the Catmull-Clark master datas grid of tissue and scalpel as input, including topological relation, characteristic
Label, the set on basic summit
B. a kind of data structure for showing characteristic surface subdivision level is created, this includes：
B1. the quad-tree structure of surface subdivision level can be reflected
The corresponding recursion step of its internal node (I), leaf node correspondingly generate final vertex position.Leaf node should also
Including regular node and boundary node (R), sharp node (C), singular point (E) and terminal node (T).
Regular node and boundary node (R) are if recursion generation rule face, by each in its 16 control points
Template is added to template list, and generates the regular node with reference to these templates.
Sharp node (C) can eliminate sharp feature by recursion
Singular point (E) does not reach the son that can directly correspond to final position if we have reached the limitation of subdivision depth
Face, then we create a singular point.Such node corresponds to the corner in basic grid face, and this is a special top
Point (EV), so we calculate the template in the extreme position in the corner and two tangent lines, and add them to template list
In.
Terminal node (T) is in each subdivision rank, and quaternary tree will have three regular nodes and an internal node, by final level
Other terminal node terminates.
B2. the ordered list for controlling point template needed for surface subdivision is created, control point is encoded to a ring neighborhood top by template
The weighting sum of point, and with a weight array representation template.
B3. 1 ring neighborhood summit can be directly read in internal memory, and uses 16*16 matrix (weight relationship matrix)
The weight on each ring summit is stored, to record the corresponding pass between topological relation and weight and control point between summit
System.
The edge in each basic grid face of micro-stepping control shader computations that C. we use in the subdivision stage, internal surface are thin
Molecular group, and the control point needed for surface subdivision, it includes following feature：
C1. each control point for being subdivided face is calculated by recursive algorithm, and stored into the respective nodes of quaternary tree, remembered simultaneously
Record surface subdivision level
C2. the GLRGB32F_ provided by OpenGL4.x is stored in by all surface subdivision factor of micro-stepping control shader computations
The floating-point texture of ARB forms, it is had four colors components (RGBA) and 32 floating number precision, deposited using red component (high position)
The subdivision surfaces factor is stored up, 24 (low levels) storage surface subdivision grade (LTs) behind use
C3.LTs is calculated by formula (1), and wherein f is the corresponding surface subdivision factor
C4. the corresponding weight in 1 ring summit and its control point is read from weight relationship matrix, and the convolution for calculating them obtains
The reference position of subdivided vertices.
D. in the gauging surface summit stage, the position at control point is obtained by traveling through quaternary tree and its normal vector and is controlled from subdivision
The summit reference position of tinter output processed calculates tinter by surface subdivision and obtains the summit position that can finally render as input
Put and its normal vector and its topological relation on all summits, its feature are as follows:
D1. the traversal of quaternary tree be not with recurrence but iteration, it is different for traversing different node processing modes
D11. all child nodes are found by traveling through internal node
D12. when the child nodes traversed are R nodes, control point position and its normal vector of its record are exported, it is thin by curved surface
Calculating tinter is divided to be mapped to the vertex position and normal vector that can finally render
D13. when the child nodes traversed are E nodes, surface subdivision calculates the summit position that tinter directly exports its record
Put and its normal vector
D14. when the child nodes traversed during T nodes, traversal is terminated.
D2. surface subdivision calculates the final subdivision curved surface of tinter final output, including all vertex positions rendered and
Topological relation between normal vector and summit, when not colliding, it is possible to carried out with the rendering pipeline opposite vertexes of routine
Render.
E. when scalpel and tissue collide, for the high frequency distortions of collision area, we are based on GPU using one
The algorithm of upper operation realizes the fine surface deformation algorithm of dynamic, and its feature is as follows：
E1. our two colliding objects are scalpel (not producing deformation, input is used as using common triangular mesh) respectively
With tissue (deformable objects, input being used as using subdivision curved surface caused by preceding step).
E2. we are that scalpel and tissue calculate orientation bounding box, when scalpel collides with tissue, two things
Body bounding box overlaps, using real-time binary system voxelization method, the lap progress voxelization to scalpel, and from
In the voxel for obtaining scalpel lap, shape when tissue is collided with scalpel is determined, and produce matched
Displacement control point
E3. displacement control point is released from scalpel overlapping region, the position of deformation and corresponding control point required for producing
Put change.
2. according to created in claim 1 it is a kind of show characteristic surface subdivision level data structure, it is characterised in that the step
Rapid B1 is specifically included：
R nodes are stored in frame buffer by a kind of particular form, and three components of X, Y, Z storage control point position have 12 words altogether
Section, N are used for storing normal vector, take 12 bytes altogether.The benefit being stored in frame buffer is that GPU can carry out high Su Du to it
Write.
3. according to created in claim 1 it is a kind of show characteristic surface subdivision level data structure, it is characterised in that the step
Rapid B3 is specifically included：
It is 16*16 matrix T, with the elements in a main diagonal Tij(i=j) i-th of 1 ring neighborhood vertex weights are recorded, use is non-
Diagonal entry Tij(i ≠ j) records the topological relation between i-th of summit and j-th of summit, and no topological relation is represented with 0,
Represent topological relation be present with 1.
4. the fine surface deformation algorithm of dynamic, institute are realized based on the algorithm run on GPU according to creating one in claim 1
Step E1 is stated to specifically include：
Scalar value B-spline coefficient is stored in tile-base texture formats, and precalculates each tile borders and corner
Adjoining pointer, and only the use of dynamic GPU memory managements is that the patch that has deformed distributes CPU internal memories.
5. the fine surface deformation algorithm of dynamic, institute are realized based on the algorithm run on GPU according to creating one in claim 1
Step E2 is stated to specifically include：
In order to determine that fine dimension deforms, by calculating scalpel and the weight of tissue the border of scalpel and tissue
Folded part (hereinafter lap), scalpel are included in lap by voxelization, specific steps：
A. by calculating OBB bounding boxs for scalpel and tissue to obtain overlapping region, and one is established in GPU internal memories
Reject decision-making buffering area
B. rejecting pretreatment is then carried out.AABB is calculated first for each patch in the space by lap to surround
Box, will be all using the convex closure attribute (self-adapting subdividing curved surface has this attribute) of curved surface in order to obtain close patch borders
B-spline patch is converted to Bezier base.Then displacement skew is incorporated to by calculating normal cone, while institute's rejecting in need
Patch reject decision-making buffering area on perform it is parallelly compressed
C. the carry out voxelization to lap.The intersecting OBB regions of non-zero to scalpel, use binary system voxelization side
Method, by the patch information selected before in rejecting decision-making buffer compression, exclusion that can be quickly takes lap
Patch, with speed-up computation.
6. the fine surface deformation algorithm of dynamic is realized based on the algorithm run on GPU according to creating one in claims 1,
The step E3 is specifically included：
All patch of the human tissue surface of lap will be replaced so that their voxel phases no longer with scalpel
Hand over, then by negative method of all control points of the B-spline (patch not being removed) of all displacements of tissue along bottom surface
Move in line direction, it is known that leave deformation voxel, and calculate the position in its final world coordinate system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710748283.1A CN107578821B (en) | 2017-08-28 | 2017-08-28 | A kind of real-time efficient GPU rendering intents in system of virtual operation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710748283.1A CN107578821B (en) | 2017-08-28 | 2017-08-28 | A kind of real-time efficient GPU rendering intents in system of virtual operation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107578821A true CN107578821A (en) | 2018-01-12 |
| CN107578821B CN107578821B (en) | 2018-08-07 |
Family
ID=61029653
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710748283.1A Expired - Fee Related CN107578821B (en) | 2017-08-28 | 2017-08-28 | A kind of real-time efficient GPU rendering intents in system of virtual operation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107578821B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108389202A (en) * | 2018-03-16 | 2018-08-10 | 青岛海信医疗设备股份有限公司 | Calculation method of physical volume, device, storage medium and the equipment of three-dimensional organ |
| CN109670206A (en) * | 2018-11-21 | 2019-04-23 | 杭州电子科技大学 | Simplify method for the hexahedral mesh structure of machine casting part model |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7884823B2 (en) * | 2007-06-12 | 2011-02-08 | Microsoft Corporation | Three dimensional rendering of display information using viewer eye coordinates |
| CN105336003A (en) * | 2015-09-28 | 2016-02-17 | 中国人民解放军空军航空大学 | Three-dimensional terrain model real-time smooth drawing method with combination of GPU technology |
-
2017
- 2017-08-28 CN CN201710748283.1A patent/CN107578821B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7884823B2 (en) * | 2007-06-12 | 2011-02-08 | Microsoft Corporation | Three dimensional rendering of display information using viewer eye coordinates |
| CN105336003A (en) * | 2015-09-28 | 2016-02-17 | 中国人民解放军空军航空大学 | Three-dimensional terrain model real-time smooth drawing method with combination of GPU technology |
Non-Patent Citations (1)
| Title |
|---|
| 王沫楠: "胞元模型在骨组织建模中的应用及评价", 《机械工程学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108389202A (en) * | 2018-03-16 | 2018-08-10 | 青岛海信医疗设备股份有限公司 | Calculation method of physical volume, device, storage medium and the equipment of three-dimensional organ |
| CN108389202B (en) * | 2018-03-16 | 2020-02-14 | 青岛海信医疗设备股份有限公司 | Volume calculation method and device of three-dimensional virtual organ, storage medium and equipment |
| CN109670206A (en) * | 2018-11-21 | 2019-04-23 | 杭州电子科技大学 | Simplify method for the hexahedral mesh structure of machine casting part model |
| CN109670206B (en) * | 2018-11-21 | 2022-12-06 | 杭州电子科技大学 | Hexahedral mesh structure simplification method for mechanical casting part model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107578821B (en) | 2018-08-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Trivariate solid T-spline construction from boundary triangulations with arbitrary genus topology | |
| Guillard et al. | Meshudf: Fast and differentiable meshing of unsigned distance field networks | |
| Biermann et al. | Sharp features on multiresolution subdivision surfaces | |
| DiPaola | Extending the range of facial types | |
| Hu et al. | Centroidal Voronoi tessellation based polycube construction for adaptive all-hexahedral mesh generation | |
| Amresh et al. | Adaptive subdivision schemes for triangular meshes | |
| Kanai et al. | 3D geometric metamorphosis based on harmonic map | |
| Zhao et al. | IGA-based point cloud fitting using B-spline surfaces for reverse engineering | |
| Hui et al. | A feature-based shape blending technique for industrial design | |
| CN112258655A (en) | Three-dimensional grid simplification method applied to VR virtual bank | |
| Warren et al. | A factored approach to subdivision surfaces | |
| Lin et al. | Adaptive patch-based mesh fitting for reverse engineering | |
| CN107578821B (en) | A kind of real-time efficient GPU rendering intents in system of virtual operation | |
| CN103093488A (en) | Virtual haircut interpolation and tweening animation producing method | |
| Bajaj et al. | Free form surface design with A-patches | |
| Fan et al. | Mesh morphing using polycube‐based cross‐parameterization | |
| Garcia et al. | Interactive applications for sketch-based editable polycube map | |
| Duan et al. | Intelligent balloon: a subdivision-based deformable model for surface reconstruction of arbitrary topology | |
| Bourguignon et al. | Relief: A modeling by drawing tool | |
| Pakdel et al. | Incremental subdivision for triangle meshes | |
| CN108876922A (en) | A kind of mesh amending method based on the regularization of interior dihedral angle supplementary angle | |
| CN106887042A (en) | Self adaptation surface subdivision method based on GPU | |
| Zheng et al. | Periodic T-splines and tubular surface fitting | |
| Pang et al. | An effective quad-dominant meshing method for unorganized point clouds | |
| Massarwi et al. | craft from 3D polygonal models using generalized cylinders |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180807 Termination date: 20190828 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |