CN103886635A - Self-adaption LOD model establishing method based on face clustering - Google Patents

Abstract

The invention discloses a self-adaption LOD model establishing method based on face clustering and belongs to the field of three-dimensional digital urban and geographic information system. The data volume of an LOD model is decreased. Due to the fact that the simplified LOD model obtained by utilizing the self-adaption LOD model establishing method is a progressive load, each LOD is formed by the partial face of an original model, and vertex attribute disorder is not caused. The self-adaption LOD model establishing method is high in algorithm efficiency, can be used for real-time self-adaption to generate the LOD model and effectively prevents the LOD switching from being excessive and unnatural. Texture optimization is standardized according to texture resolution ratio.

Description

Self-adaptation LOD model building method based on face cluster

Technical field

The invention belongs to three-dimensional digital city and Geographic Information System field, particularly relate to a kind of LOD model building method.

Background technology

LOD(Levels of Detail, level of detail) be that conventional three-dimensional model loads thought, its core is to generate good LOD model and the loading strategy of LOD model.Good model optimization result can improve rendering efficiency under identical stress model; And for identical model, good loading mode can improve rendering speed equally.

Existing BUILDINGS MODELS short-cut method, adopts the short-cut method based on free grid mostly, deleting face or merging in the process of face, can make buildings produce distortion, poor effect; Secondly, because BUILDINGS MODELS is all posted data texturing conventionally, and in most Model Simplification Method, vertex attribute is considered not, short-cut method can bring the confusion of Texture Coordinates, thereby causes texture mapping entanglement, has greatly reduced the quality of model simplification.

The loading efficiency of LOD model is not high.Many three-dimensional platforms are in loading magnanimity model data, exist model loading efficiency low, between LOD model, switch smooth not, load and the slow problem of rendering speed, affect visual experience effect, affect the further sector application of three-dimensional platform, how to have improved the difficult problem that the dynamic throughput speed of model and the 3 D rendering speed of quickening large scene need to solve.

Summary of the invention

Because the above-mentioned defect of prior art, technical matters to be solved by this invention is to provide in a kind of digital city, improves the LOD model building method of large scene three-dimensional model loading efficiency.

For achieving the above object, the invention provides a kind of self-adaptation LOD model building method based on face cluster, carry out according to the following steps:

Step 1, generation shape LOD; Generate and optimize texture LOD;

Described generation shape LOD carries out according to the following steps:

A1, obtain the important factor of each;

A2, the segmentation threshold of determining important factor partitioning model;

A3, generation simplified model, build shape LOD;

Step 2, employing accumulate mode load shape LOD, utilize traditional switching mode to load texture LOD.

Preferably, the important factor that obtains each described in step 1 carries out according to the following steps:

For each face of BUILDINGS MODELS defines its important factor for whole BUILDINGS MODELS outward appearance percentage contribution, the summation of setting the important factor of all is 1, and the important factor summation of setting simplified model is E, 0<E≤1; For onesize important factor summation E, obtain the combination that uses the minimum face of number.

Preferably, segmentation threshold the partitioning model of described in step 1, determining important factor carry out according to the following steps:

All faces, according to the large minispread of important factor, by setting the segmentation threshold of two or more important factors, are cut apart building sides, obtained face cluster result, thereby BUILDINGS MODELS is divided into multiple parts; Again the part of these divisions is combined to the shape LOD that obtains different stage.

Preferably, described in step 1, generate simplified model, build shape LOD and carry out according to the following steps:

The simplified model of all these ranks of formation in threshold value; By different segmentation thresholds, divide the simplified model that obtains different stage, construct shape LOD by the simplified model combination of different stage.

Preferably, described in step 1, carrying out the optimization of LOD model texture carries out according to the following steps:

The fine textures of setting LOD model texture is T1, and the simplification texture of setting LOD model texture is T2;

For fine textures T1, optimize refined model dimension of picture:

Setting for the texture standard resolution of pinup picture is R ₁(r ₁x,r ₁y), r ₁x and r ₁y is respectively texture standard for the pinup picture resolution in coordinate axis x direction and y direction; Setting refined model photo resolution is R ₂(r ₂x,r ₂y), r ₂x and r ₂y is respectively the resolution of refined model picture in coordinate axis x direction and y direction; The reduction multiple of setting refined model picture is N (nx, ny), and nx and ny are respectively the reduction multiple of fine textures in coordinate axis x direction and y direction; Calculate

obtain the reduction multiple of refined model picture; Be original 1/N (nx, ny) by the length of refined model picture and width reduction;

The render to texture technology that uses 3Dmax, is merged into a figure by all simplification texture T2, and utilizes method for resampling to reduce and simplify the resolution after texture T2 merges.

The invention has the beneficial effects as follows:

1, the present invention has reduced the data volume of LOD model, has improved the loading efficiency of model.General LOD model at least comprises original refined model, and the simplified model generating, if multistage LOD, model value is more.And the simplified model that uses this method to obtain is to utilize the part of original refined model to form, all total model values are still original refined model.Texture aspect, it is all same texture that this paper method generates simplification texture for multistage LOD model, and the LOD model conventionally generating, owing to there is distortion, may need more cover textures and overlap texture coordinate more just can meet the demands.

2, the present invention can not cause the entanglement of vertex attribute; Utilize simplification LOD model that this method obtains owing to being a kind of laddering loading, the part face that the formation of each LOD derives from master pattern forms, and therefore can not cause the entanglement of vertex attribute.The texture coordinate that for example summit comprises can there is not entanglement in normal information after LOD generates.

3, efficiency of algorithm of the present invention is high, can be used for real-time adaptive and generates LOD model; LOD model texture can be fixedly installed as simplifying texture and fine textures two-layer, and in shape LOD generation, simplified model is to be made up of the part face of refined model, only need to be according to the fine degree of setting, then according to the size of the important factor of face, increase and decrease in real time part face, obtain new clustering, thereby obtain new shape LOD, form LOD model with LOD model texture.

4, it is natural excessively not when the present invention has effectively avoided LOD to switch; Switching for common LOD, because simplified model and refined model are different, is therefore the inconsistent situation that inevitably has of switching between different models, simplifies effect better, crosses and gets over nature.While utilizing this method to switch, on the LOD of front portion model basis, constantly increase new face, form new more meticulous LOD model, finally reach the demonstration of refined model.

5, according to texture resolution, texture is optimized to specification more; For three-dimensional digital city model, not only should have model accuracy, the precision that also should paste texture to model is carried out unified consideration, the present invention utilizes resolution to carry out Automatic Optimal to texture, more specification the pinup picture standard of texture.

Brief description of the drawings

Fig. 1 is the schematic flow sheet of the embodiment of the invention.

Embodiment

Below in conjunction with drawings and Examples, the invention will be further described:

As shown in Figure 1, a kind of self-adaptation LOD model building method based on face cluster, carries out according to the following steps:

Step 1, generate shape LOD, generate and optimize texture LOD;

Described generation shape LOD carries out according to the following steps:

A1, obtain the important factor of each;

A2, determine division threshold value the partitioning model of important factor;

A3, generation simplified model, build shape LOD;

Step 2, employing accumulate mode load shape LOD, utilize traditional switching mode to load texture LOD.

The important factor that obtains each described in step 1 carries out according to the following steps:

For each face of BUILDINGS MODELS defines its important factor for whole BUILDINGS MODELS outward appearance percentage contribution, according to the important factor of the practical significance definition face of mould shapes itself and model face.In the present embodiment, according to the important factor of the practical significance definition face of mould shapes itself be: the large important factor of area of face is greater than the important factor of the face that area is little; According to the important factor of the practical significance definition face of model face be: the important factor of the agent structure face of building is greater than the important factor of building detail moulding face.The summation that defines the important factor of all is 1, and the important factor summation of setting simplified model is E, 0<E≤1; Simplified model is for onesize important factor summation E, obtains the combination that uses the minimum face of number.E is larger, and simplified model is meticulousr.

Segmentation threshold the partitioning model of described in step 1, determining important factor carry out according to the following steps:

By all faces according to the large minispread of important factor, by setting the segmentation threshold of two or more important factors, by BUILDINGS MODELS not coplanar carry out cluster, thereby be divided into multiple parts, then the part of these divisions is combined to the shape LOD that obtains different stage.In the present embodiment, segmentation threshold adopts adaptive approach to obtain, and according to when the loaded and displayed, the requirement of the fine degree to this rank simplified model is set.

Described in step 1, generate simplified model, build shape LOD and carry out according to the following steps:

Each segmentation threshold can be divided into all faces two parts, the simplified model of all these ranks of formation in threshold value; By different segmentation thresholds, divide the simplified model that obtains different stage, construct shape LOD by the simplified model combination of different stage; Adjacent two-stage LOD, meticulousr simplified model be compared with all of simple model and increase progressively all of part (be included in both and divide the face between threshold value) and.

In the present embodiment, taking three grades of LOD models as example, more by that analogy multi-level.Shown in (1-4):

f ₁＞＝f ₂……＞＝f _k1……＞＝f _k2……＞＝f _n； （1）

F _all={F ₁,F ₂,……,F _k1,……,F _k2,……,F _n}； （2）

$\left\{\begin{array}{c}M1=\{F_1,F_2,......,F_{k1}\}\\ M2=\{F_{k1+1},......,F_{k2}\}\\ M3=\{F_{k2+1},......,F_n\}\end{array}\right.;---\left(3\right)$

Formula (1) is the important factor by the face of arranging from big to small, this embodiment carrys out the value of given important factor with the size of each face, formula (2) is the arrangement of the corresponding face of each important factor, formula (3) is face clustering result, obtain formula (4) according to division result, be LOD model, comprise LOD0, LOD1 and LOD2.

F in formula (1) _ifor coming the important factor of i position in sequence, i is positive integer, f _k1and f _k2respectively the minimum value of the important factor value f of the face of composition M1 and M2, i.e. segmentation threshold, subscript k1 and k2 represent its position in important factor sequence; In formula (2), F _irepresent f _icorresponding face, F _allrefer to the set of all, and refined model all are comprised, so F _allit is refined model.M1, M2, M3 represents respectively the face set of three parts that are divided into, for example M1 is by face F ₁, F ₂..., F _k1composition.In formula (4), refined model LOD0 is by { M3} constitutes for M1, M2, and what LOD1 and LOD2 represented is the simplified model of the different stage of generation, respectively by { M1, M2} is with { M1} composition, the fine degree of LOD1 is greater than LOD2.

Divide minimum value f _k1and f _k2definite constraint that comprises two aspects:

1) constraint condition 1:

Can keep the aspect ratio of original refined model to set from simplified model, and only need to meet at LOD model the sharpness between its viewing area, for example LOD1 shows between modal distance 500-1000 rice, as long as ensure that LOD1 meets the demands 500 meters of outward appearances far away, can adaptively obtain the Preliminary division of LOD model face by given sharpness size.

Taking by all faces of model be all triangular facet as example, suppose the important factor taking the area of triangular facet as face, and to generate two-layer LOD model as example, comprise LOD0, and LOD1.First threshold value can be determined as follows so:

V _screen=V _{the world}* MVPW; (5)

Above formula shows model coordinate in the world coordinate system transformational relation to screen coordinate system, V _screenfor screen coordinate, V _{the world}for the coordinate of point in world coordinate system, MVPW is transition matrix.Setting screen allows the smallest triangle size of ignoring, inverse its in actual coordinates, the switching that is positioned at simplified model and refined model apart from time triangle area size, be the area of minimal face in simplified model set, it is the important factor of corresponding minimal face, in the sequence of the face representing in formula (2), obtain correspondence and try to achieve P1 triangular facet as dividing interval, P1 is positive integer.

2) constraint condition 2:

The restriction that only has P1 is inadequate, if a model is all made up of facet, can be by Ignore All, and the face set of gained simplified model is empty, therefore also needs to increase P2, P2 is set as follows:

By the important factor of face by progressively cumulative from big to small, in its calculating formula (1) before P2 item with all summation sizes of E(be 1), E meets:

$E=\underset{i=0}{\overset{i=P2}{\mathrm{\&Sigma;}}}{f}_i\&GreaterEqual;E_{\min };---\left(6\right)$

F _ibe i leg-of-mutton important factor, E _minfor the E minimum value allowing, can solve by formula (6) value that obtains P2.

In conjunction with the size of constraint 1 P1 trying to achieve and the value of constraint 2 P2 that try to achieve, have:

$\left\{\begin{array}{c}M1=\{F_1,F_2,......,F_k\}\\ M2=\{F_k,......,F_n\}\end{array}\right.,$ Wherein k=max (P1, P2); (7)

And then obtain being constructed as follows of shape LOD:

$\left\{\begin{array}{c}\mathrm{LOD}1=\left\{M1\right\}\\ \mathrm{LOD}0=\{M1,M2\}\end{array}\right.;---\left(8\right)$

Described in step 1, generate and optimize texture LOD and carry out according to the following steps:

The fine textures of setting LOD model texture is T1, and the simplification texture of setting LOD model texture is T2;

For fine textures T1, optimize refined model dimension of picture:

Setting for the texture standard resolution of pinup picture is R ₁(r ₁x,r ₁y), r ₁x and r ₁y is respectively texture standard for the pinup picture resolution in coordinate axis x direction and y direction; Setting refined model photo resolution is R ₂(r ₂x,r ₂y), r ₂x and r ₂y is respectively the resolution of refined model picture in coordinate axis x direction and y direction; The reduction multiple of setting refined model picture is N (nx, ny), and nx and ny are respectively the reduction multiple of fine textures in coordinate axis x direction and y direction; Calculate

obtain the reduction multiple of refined model picture; Be original 1/N (nx, ny) by the length of refined model picture and width reduction.

Use the render to texture technology of 3Dmax, all simplification texture T2 are merged into a figure, and utilize method for resampling by the decrease resolution 3/4 or 15/16 of simplifying after texture T2 merges, in the present embodiment, by the method for sampling by the decrease resolution 3/4 of the simplification texture T2 after merging.

In step 2, adopt cumulative mode to load described shape LOD and LOD model texture, the present embodiment is taking three layers of LOD model and two-layer LOD texture as example, and the loading distance range of model Different L OD is as follows:

$\left\{\begin{array}{cc}\mathrm{LOD}2=\{M1;T2\}& s\&GreaterEqual;s1\\ \mathrm{LOD}1=\{M1,M2;T2\}& s1>s\&GreaterEqual;s2\\ \mathrm{LOD}0=\{M1,M2,M3;T1\}& s2>s\end{array}\right.;---\left(9\right)$

S is the distance of modal position apart from current view point, s1, and s2 is that L0D switches distance, the parameter during in conjunction with the generation of LOD model, the switching of setting when the obtaining of P1 value is apart from the value as s1 and s2.Equally given switch distance after, also can adjust in real time clustering, obtain new real-time adaptive LOD model.

In the time of s >=s1, only load M1, texture uses T2, and (T2 is for simplifying texture, T1 is fine textures, is generated and is optimized the two-layer LOD texture of acquisition by a upper joint texture LOD), as s1 > s >=s2, increase M2, now formed LOD1, texture still uses T2; In the time of s2 > s, further M3 is increased to scene, form refined model LOD0, use fine textures T1 simultaneously.So, form laddering model LOD and loaded system.

The present invention utilizes model LOD technology, obtains the model node of one group of different level of detail.LOD model comprises two parts, shape LOD and texture LOD.The LOD model texture that the present embodiment relates to mainly comprises the simplification texture of fine textures and one-level resolution decreasing, two-layer altogether.And shape LOD comprises refined model and multistage simplified model equally, jointly form LOD model.As 2 grades of LOD models, comprise refined model and one-level simplified model, 3 grades of LOD models, comprise refined model and 2 grades of simplified models.Refined model is the three-dimensional model having obtained, and can obtain by three-dimensional manual modeling or automatic modeling.

The face of the model that the present embodiment relates to, comprises the set that model all converts the triangular facet after triangular grid to and comprises polygon facet and triangular facet.

More than describe preferred embodiment of the present invention in detail.Should be appreciated that those of ordinary skill in the art just can design according to the present invention make many modifications and variations without creative work.Therefore, all technician in the art, all should be in by the determined protection domain of claims under this invention's idea on the basis of existing technology by the available technical scheme of logical analysis, reasoning, or a limited experiment.

Claims (5)

1. the self-adaptation LOD model building method based on face cluster, is characterized in that carrying out according to the following steps:

Step 1, generation shape LOD; Generate and optimize texture LOD;

Described generation shape LOD carries out according to the following steps:

A1, obtain the important factor of each;

A2, the segmentation threshold of determining important factor partitioning model;

A3, generation simplified model, build shape LOD;

Step 2, employing accumulate mode load shape LOD, utilize traditional switching mode to load texture LOD.

2. the self-adaptation LOD model building method based on face cluster as claimed in claim 1, is characterized in that: the important factor that obtains each described in step 1 carries out according to the following steps:

For each face of BUILDINGS MODELS defines its important factor for whole BUILDINGS MODELS outward appearance percentage contribution, the summation of setting the important factor of all is 1, and the important factor summation of setting simplified model is E, 0<E≤1; For onesize important factor summation E, obtain the combination that uses the minimum face of number.

3. the self-adaptation LOD model building method based on face cluster as claimed in claim 1, is characterized in that: segmentation threshold the partitioning model of described in step 1, determining important factor carry out according to the following steps:

All faces, according to the large minispread of important factor, by setting the segmentation threshold of two or more important factors, are cut apart the face of BUILDINGS MODELS, obtained face cluster result, thereby model partition is become to multiple parts; Again the part of these divisions is combined to the shape LOD that obtains different stage.

4. the self-adaptation LOD model building method based on face cluster as claimed in claim 1, is characterized in that: described in step 1, generate simplified model, build shape LOD and carry out according to the following steps:

The simplified model of all these ranks of formation in threshold value; By different segmentation thresholds, divide the simplified model that obtains different stage, construct shape LOD by the simplified model combination of different stage.

5. the self-adaptation LOD model building method based on face cluster as claimed in claim 1, is characterized in that: described in step 1, carry out the optimization of LOD model texture and carry out according to the following steps:

The fine textures of setting LOD model texture is T1, and the simplification texture of setting LOD model texture is T2;

For fine textures T1, optimize refined model dimension of picture:

Setting for the texture standard resolution of pinup picture is R ₁(r ₁x,r ₁y), r ₁x and r ₁y is respectively texture standard for the pinup picture resolution in coordinate axis x direction and y direction; Setting refined model photo resolution is R ₂(r ₂x,r ₂y), r ₂x and r ₂y is respectively the resolution of refined model picture in coordinate axis x direction and y direction; The reduction multiple of setting refined model picture is N (nx, ny), and nx and ny are respectively the reduction multiple of fine textures in coordinate axis x direction and y direction; Calculate

obtain the reduction multiple of refined model picture; Be original 1/N (nx, ny) by the length of refined model picture and width reduction;

The render to texture technology that uses 3Dmax, is merged into a figure by all simplification texture T2, and utilizes method for resampling to reduce according to actual needs and simplify the resolution after texture T2 merges.

Images