CN104715495A - Three-dimensional model transmission method based on user designated error precision - Google Patents

Abstract

The invention discloses a kind of three-dimensional model transmitting methods based on user's specification error precision. The collimation error and low frequency aberration of mode needed for the present invention allows user specified in client first. Secondly the geometric data of the model is done Laplacian conversion by server-side, and the coordinate after conversion is known as δ-coordinates coordinate. Then, the collimation error selected according to user, does corresponding quantization for δ-coordinates coordinate. Pass through Selecting All Parameters again The anchor point of two different values comes tectonic model low frequency aberration and parameter Between functional relation, and the anchor point quantity of addition needed for choosing the model according to the low frequency aberration that this function and user are specified. Finally the low frequency aberration of the model is finely adjusted, so that it is met the requirement of user, and the model is sent to client. Method according to the present invention, the threedimensional model quality that user is transmitted needed for being selected according to current network quality and actual demand.

Description

Based on the three-dimensional model transmitting method of user's specification error precision

Technical field

The invention belongs to the multimedia technology field relevant to Internet Transmission, particularly a kind of three-dimensional model compression transmitting method of user's specification error precision.

Background technology

Along with development and the industrial widespread use to three-dimensional model of computer graphics, three-dimensional model is just towards future development that is complicated, that become more meticulous.By the dimensional Modeling Technology of maturation, some large-scale virtual scenes such as 3D museum, three-dimensional virtual city, landform lifelikely can be presented in us at the moment.These scenes are usually in large scale, dough sheet One's name is legion, and resolution is high, and visual effect is splendid.Therefore whole model is extremely complicated and model data amount is large.In addition, in modern industrial design, the structure of industrial design product becomes increasingly complex, and the parts of composition get more and more, and each size zero parts need to be expressed by three-dimensional model.Therefore, industrial product structure is more complicated, and the structure of three-dimensional model is also more complicated.

Another approach obtaining three-dimensional model is by 3-D scanning technology, and along with improving constantly of scanning technique, the cost being obtained three-dimensional model by scanning is more and more lower, and the precision of model is more and more higher.Its local detail feature rich of true for some, complicated material object, the curvature of surface mesh is large, the face of such as human body, the artworks etc., if obtain its three-dimensional model by three-dimensional modeling approach, the manufacture craft of art designing will be extremely huge, often seem unable to do what one wishes, and by 3-D scanning technology, can easily the volume coordinate of object, color-values be swept in computing machine rapidly, obtain and natural scale, shape, the living three-dimensional model of color, work efficiency is improved.In video display special technology making field, 3-D scanning technology is widely used in the entertainment medium such as film and tv industry, advertising industry, and the utilization of special effects has brought unprecedented visual enjoyment, has now become the criterion of high-quality production of film and TV.In order to show each details in kind, 3-D scanning technology is just towards the future development become more meticulous ultra high-definition.

Because the scale of three-dimensional model is more and more huger, mould shapes becomes increasingly complex, and the precision of model is more and more higher, causes the data volume of the required process of three-dimensional model and transmission sharply to increase.Meanwhile, universal along with PC and intelligent movable equipment, increasing domestic consumer to the browsing of three-dimensional model, the demand such as to share and also sharply increase.Especially in e-commerce field, in order to attract client, businessman needs the internal information of displaying merchandise in all directions, as structure, function information, shopper can be roamed in dummy market, carries out interactively operation to three-dimensional model.

In three-dimensional model transmitting, by carrying out compression process to the data volume of three-dimensional model, significantly can reduce the data volume of required transmission, thus save the limited network bandwidth and reduce the data processing amount of client.According to the difference of network quality and user to the difference of three-dimensional model demand, the multiresolution progressive compression transmission method of three-dimensional model is a solution wherein.But, for drawing effect in the middle of client realization the best, under given bit rate, progressive compression transmission method need be weighed between the two in the compressibility of model and drafting distortion rate, and this calculating needs the calculation cost of at substantial.In addition, in some Mobile solution, as moving game and mobile virtual roaming, three-dimensional model in complex scene all adopts the compression method of single bit rate usually, so that three-dimensional model can be transferred to client from service end fast, and the compression method of single resolution is adopted to have higher compressibility.

Summary of the invention

The present invention is directed to the deficiency of prior art on single resolution model Compression Transmission Technology, propose a kind of three-dimensional model transmitting method based on user's specification error precision.

The technical solution adopted for the present invention to solve the technical problems is as follows:

First, user is allowed to specify the collimation error and the low frequency aberration of required mode in client.Secondly, the geometric data (Cartesian coordinates) of this model is done Laplacian conversion by service end, and the coordinate after conversion is called δ-coordinates coordinate.Then, according to the collimation error that user selects, δ-coordinates coordinate is done corresponding quantification.Again by the anchor point of Selecting All Parameters υ (parameter υ represents anchor point dense degree) two different values, carry out the funtcional relationship between tectonic model low frequency aberration and parameter υ, and choose according to the low frequency aberration that this function and user are specified the anchor point quantity that this model need add.Finally, the low frequency aberration of this model is finely tuned, make it meet the requirement of user, and this model is sent to client.After client receives data, a demand solution least square solution, just can recover the geometric data of this model.

Beneficial effect of the present invention: the present invention is a kind of three-dimensional model compression transmitting method based on user's specification error precision, user can select the required three-dimensional model quality transmitted according to the demand of current network quality and reality.

Accompanying drawing explanation

Fig. 1 is anchor point distribution and path.

Embodiment

Below in conjunction with accompanying drawing, the invention will be further described:

First the present invention allows user specify the collimation error and the low frequency aberration of required mode in client.

Collimation error review extraction is: wherein: $S_q\left(v_i\right)=\left|\right|S\left(v_i\right)-S\left(Q\left(v_i\right)\right)\left|\right|,S\left(v_i\right)=v_i-\frac{\underset{j\∈N\left(i\right)}{\mathrm{\Σ}}{l}_{\mathrm{ij}}^{-1}{v}_j}{\underset{j\∈N\left(i\right)}{\mathrm{\Σ}}{l}_{\mathrm{ij}}^{-1}},$ L in above-mentioned formula _ijrepresent the distance of summit i to j, N (i) represents the set of the abutment points on i summit, Q (v _i) be the vertex v that reconstruction model is corresponding _i.

Low frequency aberration review extraction is: wherein: v _ifor original geometric coordinate, Q (v _i) be the geometric coordinate of reconstruction model.

Service end upon receiving a request, will do following process:

Step 1: be relative coordinate form from Cartesian coordinates formal transformation by the geometric data of three-dimensional model.Definition V={v ₁, v ₂,, v _nbe the vertex set of grid M, wherein v _i=(x _i, y _i, z _i).δ-coordinates the coordinate defining each summit is wherein d _irepresent the number (being the degree on summit) of the direct abutment points on summit, for x _i, y _i, z _icorresponding δ-coordinates coordinate.Definition Laplacian matrix is L=D-A, and wherein D is main diagonal matrix, and the element on principal diagonal is: D _ii=d _i, A is the adjacency matrix of grid M.It has following relation: Lx=D δ ^(x), Ly=D δ ^(y), Lz=D δ ^(z).Can by the x of grid, y by above three relational expressions, z coordinate is converted into δ-coordinates coordinate.

Step 2: δ-coordinates coordinate is quantized.After the Cartesian coordinates of three-dimensional model being converted into δ-coordinates relative coordinate, real-coded GA can not directly with based on the coding of dictionary or entropy code, and must be by quantifying.The visual level of mode needed for selecting according to client, select corresponding bits to quantize in service end, quantizing range is 3-8bits.

Step 3: the increment anchor point based on BFS (breadth first traversal) is chosen.Add the low frequency aberration that anchor point can reduce reconstruction model, it is as follows that the anchor point of BFS chooses process: first random selecting summit is as anchor point, then using this point as root node, breadth first traversal (BFS) is done to the topology of model, record the degree of depth on every one deck summit, the namely topology distance of distance root node (anchor point), selects that the darkest for degree of depth point as the 2nd anchor point as.Again with the 2nd anchor point for root node, start to do breadth first traversal, if the degree of depth recorded before current depth ratio is little, revise it.When the darkest degree of depth υ=15, all anchor points selected are before put into container (υ ₁₅).Continue to select anchor point, until the darkest degree of depth is υ=14, and the anchor point additional anchor point Δ newly increased again when anchor point number during υ=14 is υ=15 ₁₄, this part anchor point newly increased is put into container (Δ ₁₄).The rest may be inferred, until anchor point increment during υ=5 is put into container (Δ ₅).Figure 1 shows that anchor point distribution during parameter υ=3.Black color dots is anchor point, and white point is general point, dotted line be in figure non-anchor apart from the maximum path value of its nearest anchor point.

Step 4: structure reconstruction model low frequency aberration M _qand the funtcional relationship M between parameter υ _q=f (v), it can be divided into three kinds of situations by the difference of δ-coordinates coordinate quantified precision, and obtains required anchor point quantity according to the low frequency aberration precision that user selects.

Wherein function M _qthree kinds of concrete conditions of=f (v) are as follows:

1) quantize if δ-coordinates coordinate presses 3bits, 4bits or 5bits, now low frequency aberration M _qand the function M between parameter υ _q=f (v) is linear: M _q=av+b.2 different values of Selecting All Parameters υ, by solving least square method reconstruction model, obtain reconstruction model error amount M respectively _q1, M _q2.((υ is worth by this two couple ₁, M _q1), (υ ₂, M _q2)) unknown number a, b can be tried to achieve, thus obtain low frequency aberration M _qand functional relation concrete between parameter υ, according to the low frequency aberration M of client input _qvalue, by inverse function υ=f ^-1(M _q)=(M _q-b)/a can try to achieve the parameter υ general proportions of anchor point (namely required).

2) quantize if δ-coordinates coordinate presses 7bits or 8bits, now low frequency aberration M _qand the function M between parameter υ _q=f (v) is form exponentially: M _q=ae ^{b υ}.Similar to choosing two different parameter υ, also can try to achieve a, the value of b.Again by inverse function υ=f ^-1(M _q)=(lnM _q-lna)/b and client input M _qvalue, tries to achieve parameter υ.

3) quantize, for the model low frequency aberration M had if δ-coordinates coordinate presses 6bits _qand the function M between parameter υ _q=f (v) is linear, some models then exponentially functional form.Therefore, we carry out matching curve now by the method for weighting, i.e. υ=α υ ₁+ (1-α) υ ₂, wherein: υ ₁=f ^-1(M _q)=(M _q-b)/a, υ ₂=f ^-1(M _q)=(lnM _q-lna)/b

Step 5: the low frequency aberration of three-dimensional model reality is finely tuned.Due to the M' of real income _qthe low frequency aberration M selected with user _qthere is certain deviation, employing Greedy anchor point choosing method finely tunes low frequency aberration now, until || M _q-M' _q||≤δ, wherein δ is family of power and influence's value.The way of Greedy anchor point choosing method is: calculate M in reconstruction model _qbe worth maximum summit, and this summit is fixed as anchor point, reconstruction model selects M again _qbe worth maximum summit, until meet the requirement of client.

Claims (4)

1., based on the three-dimensional model transmitting method of user's specification error precision, it is characterized in that:

Step 1, allows user specify the collimation error and the low frequency aberration of required mode in client;

Step 2, the geometric data of this model is done Laplacian conversion by service end, and the coordinate after conversion is called δ-coordinates coordinate;

Step 3, according to the collimation error that user selects, does corresponding quantification by δ-coordinates coordinate;

Step 4, by the anchor point of Selecting All Parameters υ two different values, carry out the funtcional relationship between tectonic model low frequency aberration and parameter υ, and choose according to the low frequency aberration that this function and user are specified the anchor point quantity that this model need add, wherein parameter υ represents anchor point dense degree;

Step 5, finely tunes the low frequency aberration of this model, makes it meet the requirement of user, and this model is sent to client; After client receives data, solve least square solution, just can recover the geometric data of this model.

2. the three-dimensional model transmitting method based on user's specification error precision according to claim 1, is characterized in that: step 2 specifically:

Definition V={v ₁, v ₂,, v _nbe the vertex set of grid M, wherein v _i=(x _i, y _i, z _i); δ-coordinates the coordinate defining each summit is wherein d _irepresent the number of the direct abutment points on summit, be the degree on summit, for x _i, y _i, z _icorresponding δ-coordinates coordinate; Definition Laplacian matrix is L=D-A, and wherein D is main diagonal matrix, and the element on principal diagonal is: D _ii=d _i, A is the adjacency matrix of grid M, and it has following relation: Lx=D δ ^(x), Ly=D δ ^(y), Lz=D δ ^(z); Can by the x of grid, y by above three relational expressions, z coordinate is converted into δ-coordinates coordinate.

3. the three-dimensional model transmitting method based on user's specification error precision according to claim 1, is characterized in that: step 4 specifically:

Tectonic model low frequency aberration M _qand the funtcional relationship M between parameter υ _q=f (v), it is divided into three kinds of situations by the difference of δ-coordinates coordinate quantified precision, and obtains required anchor point quantity according to the low frequency aberration precision that user selects;

Wherein function M _qthree kinds of concrete conditions of=f (v) are as follows:

1) quantize if δ-coordinates coordinate presses 3bits, 4bits or 5bits, now low frequency aberration M _qand the function M between parameter υ _q=f (v) is linear: M _q=av+b; Two different values of Selecting All Parameters υ, by solving least square method reconstruction model, obtain reconstruction model error amount M respectively _q1, M _q2; ((υ is worth by this two couple ₁, M _q1), (υ ₂, M _q2)) unknown number a, b can be tried to achieve, thus obtain low frequency aberration M _qand functional relation concrete between parameter υ, according to the low frequency aberration M of client input _qvalue, by inverse function υ=f ^-1(M _q)=(M _q-b)/a can try to achieve parameter υ;

2) quantize if δ-coordinates coordinate presses 7bits or 8bits, now low frequency aberration M _qand the function M between parameter υ _q=f (v) is form exponentially: M _q=ae ^{b υ}; Similar to choosing two different parameter υ, also can try to achieve a, the value of b; Again by inverse function υ=f ^-1(M _q)=(lnM _q-lna)/b and client input M _qvalue, tries to achieve parameter υ;

3) quantize, for the model low frequency aberration M had if δ-coordinates coordinate presses 6bits _qand the function M between parameter υ _q=f (v) is linear, some models then exponentially functional form; Therefore, matching curve is now carried out by the method for weighting, i.e. υ=α υ ₁+ (1-α) υ ₂, wherein: υ ₁=f ^-1(M _q)=(M _q-b)/a, υ ₂=f ^-1(M _q)=(lnM _q-lna)/b.

4. the three-dimensional model transmitting method based on user's specification error precision according to claim 1, is characterized in that: step 5 adopts Greedy anchor point choosing method to finely tune low frequency aberration now, until || M _q-M' _q||≤δ, wherein δ is family of power and influence's value; Greedy anchor point choosing method is: calculate M in reconstruction model _qbe worth maximum summit, and this summit is fixed as anchor point, reconstruction model selects M again _qbe worth maximum summit, until meet the requirement of client.