CN103701798A - Intelligent progressive transmission system and transmission method for three-dimensional model - Google Patents

Abstract

The invention discloses an intelligent progressive transmission system and transmission method for a three-dimensional model. The transmission system comprises a three-dimensional model preprocessing module, a server side and a client, wherein the output end of the three-dimensional model preprocessing module is connected with the server side; the serve side is communicated with the client. The transmission method mainly comprises the following steps: performing DOP-SLOD (Distribution of Precision-Smooth LOD) encoding on the three-dimensional model in a scene, and performing accuracy binding on the encoded three-dimensional model to generate a model encoding stream node file; sending a request to a server; judging the accuracy; forecasting a track to generate a model pre-downloading queue; transmitting the stored model encoding stream node file to the client, decrypting the model encoding stream node file received by the client, and finishing a dynamic real-time rendering drawing step of the model. By adopting the intelligent progressive transmission system and the transmission method, rapid transmission of the three-dimensional model can be realized by effectively using limited bandwidth on the premise of ensuring the browsing effect of a user.

Description

Threedimensional model intelligent progressive formula transmission system and transmission method

Technical field

The present invention relates to field of computer technology, particularly a kind of transmission method of threedimensional model.

Background technology

In recent years, along with the concept of three-dimensional internet more and more comes into one's own, each major company has released the three-dimensional internet application platform of oneself one after another, what cause the earliest extensive concern is the starting Second Life three-dimensional internet platform of Linden Lab, also have in addition IBM Corporation IW three-dimensional internet application platform, in look the VRPIE three-dimensional network platform of allusion quotation.Yet whichever three-dimensional internet platform, has all inevitably run into the contradiction between model data amount and finite element network bandwidth.

For solving above-mentioned contradiction, current most of three-dimensional internet application platform model based coding has all adopted detail structure (Level of Details is called for short LOD) technology.But because LOD Technology Need generates a plurality of discontinuous LOD models in advance to threedimensional model, then download as required demand correspondence and be set to a plurality of accuracy models.This mode exists defect as follows:

(1) continuity when guaranteeing that as much as possible model changes between different accuracy, need to generate the simplified model of many different accuracies, thereby greatly increase transmitted data amount to same object model;

(2) due to the corresponding a plurality of models of model different accuracy, so model is when precision changes, and is Discrete Change, and discrete.

Chinese scholars is constantly along with further investigation, progressive transmission (Progressive Transmission is proposed, PT) be to solve finite bandwidth and the effective method of large-scale virtual contextual data, the hopping sense that can avoid again discrete LOD method to produce when model rendering shows simultaneously.

Hoppe has proposed the progressive transmission method in three-dimensional virtual scene first, has realized the continuous incremental transmission of three-dimensional body, and the method has thoroughly changed the discreteness of conventional method display effect.

Teler and Lischinski have proposed the concept of three-dimensional data Streaming stream, by the qualified model data to virtual scene, transmit as a stream, each transmission user be sub-fraction model data within sweep of the eye, realize user's online displaying live view, but the method only pay the utmost attention to user within sweep of the eye model data consider, be that model comes into view and starts to download, do not come into view and do not download, this mode, when user's long period does not change position, can cause bandwidth waste unavoidably.

Cheng Zhi congruence people convection techniques elaborates, and designs stream transmission that gradual compression, the viewpoint of 3-D geometric model is relevant, has the wrong stream transmission framework of controlling function.But emphasis is still to compress by model data, reduces as much as possible transmitted data on network amount, can not fundamentally change the contradiction between finite bandwidth and extensive three-dimensional geometry data.

The people such as Li Chengying adopt the mode of progressive transmission to realize domestic first online virtual museum display platform, but the function of platform is confined to show not have interactive function.The people such as Wang Wei has provided detailed process and the explanation of model progressive transmission in large-scale virtual scene in addition, complete elaboration in progressive transmission key modules form and principle, for descendant's research provides reference.

The key component that realizes progressive transmission is model to be applicable to the coding of progressive transmission, and numerous researchers have also carried out a large amount of research on the simplification compression of threedimensional model and streaming coding techniques in recent years, have obtained larger breakthrough.

Hoppe proposed to represent model with Progressive Mesh (Progressive Mesh, PM) in 1996.PM core concept be by geometrical model by coding, formation base grid and a series of delta file, in the process of transmission, implementation model plays up demonstration continuously.When but the PM method for expressing proposing due to Hoppe merges on summit, used complicated energy method computations to determine the merging mode on summit, so model simplification efficiency is very low.

Compression Progressive Mesh (the Compressive Progressive Mesh that Renato Pajarola and Jarek Rossignac propose, CPM) technology, compare with PM technology, CPM method is more focused on the compression of model data, owing to being batch operation when encoding and decode, uses CPM method, model is discrete when playing up demonstration, be not continuous, when the network bandwidth has in limited time, this discreteness is more obvious.

Schmalstieg proposes a kind of smooth level of detail (Smooth LOD, SLOD) model representation method.This method for expressing principle and PM method are quite similar, but have avoided while merging on summit energy complicated in PM method to calculate, thereby have greatly improved model simplification efficiency.

From above achievement in research, can find out, the correlative study work of large-scale virtual scene progressive transmission at present has obtained a large amount of achievements.But still can find, lot of research still mainly concentrates in the processing of model its data, because geometrical model data are in continuous increase, only by transaction module data, use progressive transmission mode to transmit, eventually can not meet user's online displaying live view requirement.Therefore in band-limited situation, by a good progressive transmission strategy, will have achievement in research and make full use of, the unique channel of effectively utilizing finite bandwidth to be only to address this problem.

Summary of the invention

The technical issues that need to address of the present invention are to provide and a kind ofly can guaranteeing, under the prerequisite of user's result of browse, effectively to utilize finite bandwidth to realize a kind of system and method for threedimensional model fast transport.

For solving the problems of the technologies described above, the technical solution adopted in the present invention is:

Threedimensional model intelligent progressive formula transmission system, comprises threedimensional model pretreatment module, server end and client, and the output connection server end of described threedimensional model pretreatment module, communicates with one another between server end and client; Wherein:

Threedimensional model pretreatment module, for threedimensional model is carried out to DOP-SLOD coding, the threedimensional model encoding is carried out to precision binding, generate corresponding model based coding stream node file, and the model based coding stream node file transfer of handling well is stored to server end;

Server end, for carrying out precision and priority judgement according to the trajectory predictions result of client to meeting the model of prediction locus, thereby forms orderly pre-download queue, further by network, to client transmissions, meets the data that user requires;

Client, for completing according to user's historical movement track the following prediction of track of user and the Accuracy Matching of data streaming file, and initiates request of data to server end; The data that send according to server end are stored and precision judgement, further according to the data streaming file that meets required precision receiving, threedimensional model are carried out to real-time rendering drafting.

The concrete structure of each module of the present invention is: described server end comprises model database, data verification module and priority determination module, wherein model database is for receiving and store the model based coding stream node file of threedimensional model pretreatment module transmission, whether data verification module is for meeting required precision and verify reading in the data of surplus from model database, priority determination module is for predicting the outcome according to client movement locus, near model track is carried out to priority judgement, according to result of determination, generate pre-download queue;

Described client comprises data demand module, trajectory predictions module, precision determination module, data processing module and modeling rendering module, wherein data demand module is for sending communication request to server end, and foundation or interrupt communication, trajectory predictions module is used for according to user's historical movement trajectory predictions user's following movement locus, precision determination module is for carrying out judgement and the checking of precision to the data of downloading from server end, complete and receive mating of data and the required precision of user, data processing module carries out denoising for the model based coding stream node file to receiving, the real-time rendering that modeling rendering module has been used for threedimensional model shows, realize user's online displaying live view.

Threedimensional model intelligent progressive formula transmission method, is characterized in that specifically comprising the following steps:

A. the threedimensional model in scene is carried out to preliminary treatment DOP-SLOD coding, the threedimensional model encoding is carried out to precision binding, generation model encoding stream node file, and deposit in server end model database;

B. user is by client-access three-dimensional internet scene, and whether client is to judging for accessing first current scene, if step C is carried out in access first, if not step D is carried out in access first;

C. user end to server sends communication link request, after communication link success, to server, send data download request, server end carries out precision judgement according to the data download request receiving and priority is judged, if precision meets the demands, carry out step F, if can not meet required precision, carry out step e;

D. the local existing model data of server calls, and carry out precision judgement, carries out step e if meet required precision, if can not meet required precision, returns to step C and sends request of data to server;

E. client is carried out trajectory predictions, and trajectory predictions result is sent to server end; Server end is according to the queue of trajectory predictions result generation model pre-download, and the data of pre-download are carried out to precision judgement, if meet required precision, carries out step F, if can not meet required precision, re-starts trajectory predictions;

F. server end sends the model based coding stream node file of storage to client, and client is decoded after model based coding stream node file, and the dynamic real-time rendering that completes model is drawn.

Owing to having adopted technique scheme, the technological progress that the present invention obtains is:

The present invention be take DOP-SLOD model based coding method as basis, intelligent progressive formula by network implementation model is transmitted, can be according to different user in data transmission procedure the difference to threedimensional model accuracy requirement, the download precision of preference pattern targetedly, avoid downloading the total data of useless model, guaranteeing on the basis of user's effect that model smoothing is played up in navigation process, rationally effectively utilized finite bandwidth, reduce user's downloading data amount, further reduced the network delay of transfer of data.

Accompanying drawing explanation

Fig. 1 is the general structure block diagram of transmission system of the present invention.

Fig. 2 is the workflow of transmission method of the present invention.

Fig. 3-1a to Fig. 3-1d is the whole structure figure of pretreated sculpture model under different accuracy.

Fig. 3-2a to Fig. 3-2d is that pretreated sculpture model is played up the partial result figure under different accuracy in client.

Fig. 3-3a to Fig. 3-3b is two data volume comparison diagrams that different user client arrives in experiment one.

Fig. 4 is the data volume comparison diagram that conventional method and the inventive method are transmitted different models.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in further details:

A threedimensional model intelligent progressive formula transmission system, its general structure block diagram as shown in Figure 1.Comprise threedimensional model pretreatment module, server end and client, the output connection server end of described threedimensional model pretreatment module, communicates with one another between server end and client.

Described threedimensional model pretreatment module, for threedimensional model is carried out to DOP-SLOD coding, generates corresponding model based coding stream node file, and the model based coding stream node file transfer of handling well is stored to server end.The preliminary treatment of threedimensional model is coded in client and server and carries out mutual completing before, before model starts to use in three-dimensional internet scene, model is carried out to precoding, form the corresponding stream node file that is applicable to progressive transmission, and convection current node file carries out precision binding, do not take the stand-by period that user browses, first encoding, available all the life.

The main task of server end has been the data task that transmission meets user's required precision, according to the trajectory predictions result of client, to meeting the model of prediction locus, carry out precision and priority judgement, thereby form orderly pre-download queue, further by network, to client transmissions, meet the data that user requires.

Concrete server end comprises model database, data verification module and priority determination module.Wherein model database is for receiving and store the model based coding stream node file of threedimensional model pretreatment module transmission; Whether data verification module is for meeting required precision and verify reading in the data of surplus from model database; Priority determination module, for according to the predicting the outcome of client movement locus, carries out priority judgement near model track, according to result of determination, generates pre-download queue.

The groundwork of client comprises calculating, access and plays up, and according to user's historical movement track, completes the following prediction of track of user and the Accuracy Matching of data streaming file, and initiates request of data to server end; The data that send according to server end are stored and precision judgement, further according to the data streaming file that meets required precision receiving, threedimensional model are carried out to real-time rendering drafting.

Concrete client comprises data demand module, trajectory predictions module, precision determination module, data processing module and modeling rendering module.Wherein the main completing user of data demand module is to request of data and the transmission work of server end, comprising communication request, connection setup and communication disruption; Trajectory predictions module is used for according to user's historical movement trajectory predictions user's following movement locus; Precision determination module is for carrying out judgement and the checking of precision to the data of downloading from server end, completes and receives mating of data and the required precision of user, and this module is the key that precision is controlled, the intelligent terminal of communicating by letter between control client and server; Data processing module carries out denoising for the model based coding stream node file to receiving, remove some garbages, such as precision information, now the precision information in threedimensional model stream node file has completed corresponding work, programming garbage, after these garbages are processed, be transferred in order modeling rendering module; The real-time rendering that modeling rendering module has been used for threedimensional model shows, realizes user's online displaying live view.

A threedimensional model intelligent progressive formula transmission method, as shown in Figure 2, described transmission method, based on above-mentioned threedimensional model intelligent progressive formula transmission system, specifically comprises the following steps its flow chart:

A. threedimensional model pretreatment module is carried out preliminary treatment DOP-SLOD coding to the threedimensional model in scene, and the threedimensional model encoding is carried out to precision binding, generation model encoding stream node file, and deposit in server end model database.

It is specific as follows that threedimensional model pretreatment module is carried out the method for preliminary treatment coding:

Before model is carried out to DOP-SLOD method coding, first model is carried out to space cutting, what the present invention adopted is k-d Tree method, model is divided into the region that a lot of point sets are less, when encoding, each coded object is the zonule after cutting, thereby has greatly improved code efficiency.

DOP-SLOD algorithm steps is as follows:

(1) be written into model, initialization correlated variables;

(2) model point set being carried out to this data structure of k-d Tree carries out space and cuts apart;

(3) in the region after cutting apart, find fast two nearest points of Euler;

(4) delete limit, the summit of deletion and corresponding precision p are pressed into stack spiltStack, the summit of renewal is pressed into stack updateStack;

(5) until only during remaining M0, end-of-encode.

After end-of-encode, stack spiltStack information is deposited in and upgrades list (update list), deposit stack updateStack information in folding list (collapsed list), in whole like this cataloged procedure, the change information of model just preserves, can be according to the content in above-mentioned two tables when model rendering shows, corresponding backward is recovered one by one.

In the three-dimensional applications scene creating, the model encoding is passed through to [O _id, (x _id, y _id, z _id)]: carry out precision binding, the model based coding stream node file of generating scene, wherein O _idfor model TD, (x _id, y _id, z _id) be that in three-dimensional applications scene, model is written into coordinate figure a little.And then by the model based coding stream node file distribution of scene to the model database of server end in, for user's download access.

B. user is by client-access three-dimensional internet scene, and whether client is to judging for accessing first current scene, if step C is carried out in access first, if not step D is carried out in access first.

C. the data demand module of client sends communication link request to server, after communication link success, to server, send data download request, the data verification module of server end is carried out precision judgement according to the data download request receiving, the priority determination module of server end carries out priority judgement according to the data download request receiving, if precision meets the demands, carry out step F, if can not meet required precision, carry out step e.

When carrying out precision judgement, first define precision base value p _base: precision base value refers to that model can meet the lowest accuracy value that user browses demand, in model database, by a precision base value table, realizes the corresponding one by one of different models and precision base value, and precision base value is static, once set, no longer changes.

The present embodiment adopts the sculpture model of 3DS form for the archetype of testing, file name is chenshe036.3DS, and archetype dough sheet number is 31398, and file size is 737KB.

Weighing model based coding effect is mainly two aspects: tri patch quantity and the visual effect of simplification.The present embodiment is stored in sculpture model in the server end model database whole structure under different accuracy as shown in Fig. 3-1a to Fig. 3-1d after the preliminary treatment of DOP-SLOD method, and sculpture model is being played up partial result under different accuracy as shown in Fig. 3-2a to Fig. 3-2d after by client downloads.In experimentation, the preliminary treatment of model and client being played up is all dynamic processes, and the variation of model accuracy is a continuous process, is not discrete, shows for convenience result, only captures the model result of 4 different accuracies here.

Fig. 3-1a shown when precision p be 0.171, while receiving 126KB data, the whole structure figure of models show, unilateral quantity is 5336; And Fig. 3-1b has shown while receiving 230KB data when precision p is 0.312, the whole structure figure of models show, unilateral quantity is 9732; Fig. 3-1c has shown while receiving 465KB data when precision p is 0.631, the design sketch of models show, and unilateral quantity is 19780; Fig. 3-1d provides while receiving 737KB partial data when precision p is 1, the design sketch of models show, and unilateral quantity is 31398.

Fig. 3-2a to Fig. 3-2d has provided respectively DOP-SLOD to the rear partial result comparison diagram of sculpture model coding, is respectively from left to right the statue model of 5336 dough sheets, 9732 dough sheets, 19780 dough sheets and 31398 dough sheets.

By Fig. 3-1b, can be found out, when precision is 0.312, while receiving 9732 dough sheets, model has met visual demand substantially, basic precision p that can be using current precision as model, user first during access scenario, when the data that receive can reach basic precision, download and stop immediately, carry out immediately precision judgement state, if basic precision can meet user while requiring, bandwidth is given to pre-download process, if and basic precision can not meet user while requiring, continue download model data, until the precision of model can reach user's requirement.When user leaves current view point, downloading process also stops at once, and bandwidth is given pre-download process.

In experiment, by different models being carried out to the processing of DOP-SLOD method, find that Most models is in precision at 0.3 o'clock, substantially can meet user and initially browse needs, therefore, in the accuracy table of model basis, unified by the basic p of the precision of model in model database _basebe set to 0.300.

Definition mean accuracy p _average: after every user finishes the access of model at every turn, the client feedback precision that downloading data reaches during to this Access Model of this user of server end, server end is according to a certain model accessed accuracy value collection within a period of time, calculate mean value, as the mean accuracy of model "current" model.

Mean accuracy p _averageaccording to formula one, calculate:

$p_{\mathrm{average}}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{p}_{(a_{\mathrm{id}},m_{\mathrm{id}})}}{n}$ Formula one

Wherein: n be sometime in access "current" model number of users;

A _idfor user ID,

M _idfor current accessed model ID.

In formula one, by user ID and accessed model ID value is bound, can uniquely determine current the obtain precision of all users to a certain model, by calculating mean value, obtain the mean accuracy requirement of user's "current" model, the scope of mean accuracy is p _average(p _base≤ p _average≤ 1).

The present invention, when model being carried out to priority judgement, except consider, distance and visual field factor, to consider the interest-degree of user to model simultaneously, uses the interest-degree that the required precision of model is embodied to user in the present invention.Owing to having ageing to the required precision of model, it is dynamic change, therefore the priority of same model is just likely not different in the same time, can make like this model priority that in a period of time, user interest degree is high high, come the prostatitis of pre-download queue, while having avoided only considering distance and the visual field, priority sequence is single, priority is temporal evolution and changing not, thereby improves the validity of pre-download data.

The concrete deterministic process of priority is as follows.

At client and server, set up communication and start after transfer of data, client is by the mean accuracy p of the maximum of p in the data flow of receiving and "current" model _averagecompare, under the prerequisite not changing in viewpoint, work as p=p _averagetime, "current" model transfer of data stops, and enters pre-download process.When user's viewpoint changes, when entering pre-download process, client feeds back to server end by the maximum of p in the current data flow receiving, for the mean accuracy renewal of this model.

Priority is calculated according to formula two:

$p_{m_{\mathrm{id}}}=p_{\mathrm{average}}/(R_{(a_{\mathrm{id}},m_{\mathrm{id}})}*\tan \mathrm{\θ})$ Formula two

In formula: p _averagefor "current" model mean accuracy, go after decimal point three to calculate,

for the distance of user and "current" model,

θ is the vector angle between prediction locus and user model, 0 < θ < pi/2.

If a certain model mean accuracy p _averagelarger, nearest a period of time is described, in scene, user wants high to this model attention rate compared with other models, and user is interested in it, and the probability of accessing it is also just higher.Between prediction locus and user model, vector angle θ is larger, and visual field outer ledge part when model moves on prediction locus the closer to user is described, user is relatively low to this model attention rate, so θ and priority is inversely proportional to, and replaces with tan θ herein.User and modal distance

larger, illustrate that both distances are far away, user is in the process of access, and the meeting of first paying close attention to is the model of close together, and therefore distance is far away, and the download priority of model is corresponding lower.By three factors constraints, the priority of model is judged, can better react the demand of user to model, the model data can more autotelic download user in the process of download needing.

D. existing model data in invoking server end model database, and carry out precision judgement, if meet required precision, carries out step e, if can not meet required precision, returns to step C and sends request of data to server.

E. the trajectory predictions module of client is carried out trajectory predictions, and trajectory predictions result is sent to server end; Server end is according to the queue of trajectory predictions result generation model pre-download, and the data of pre-download are carried out to precision judgement, if meet required precision, carries out step F, if can not meet required precision, re-starts trajectory predictions.

In the present invention, trajectory predictions adopts W-EWMA trajectory predictions method, W-EWMA method is that the historical track vector in the window ranges to setting carries out exponent-weighted average method, by setting window threshold values, in the process of calculating, only consider the historical movement vector in window ranges, this has just reduced the irrelevant vectorial stand-by period of calculating, has improved computational speed.

W-EWMA method step is as described below:

(1) determine window threshold values n;

(2) determine vectorial weights factor alpha;

(3) access local historical track storage file, read historical track vectorial coordinate in window;

(4) by calculating prediction locus, determine future position coordinate, and pass to server end.

Trajectory predictions vector through type three calculates:

${\stackrel{\&RightArrow;}{m}}_{n+1}=\underset{i=0}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}^i\&CenterDot;{\stackrel{\&RightArrow;}{m}}_{n-i}$ Formula three

In formula:

for trajectory predictions vector.

for historical movement track vector,

α is historical movement track vector weights coefficients, 0 < α < 1.

Through type three can calculate trajectory predictions vector future position coordinate is:

(x _n+1, y _n+1, z _n+1)=(x _n+ a, y _n+ b, z _n+ c) formula four

For example establishing window threshold values n is 2, namely during predicted motion track, only gets nearest 2 movement locus vectors, and three points of component movement track are respectively A (x ₀, y ₀, z ₀), B (x ₁, y ₁, z ₁), C (x ₂, y ₂, z ₂), future position coordinate is D (x ₃, y ₃, z ₃), historical track vector is:

${\stackrel{\&RightArrow;}{m}}_1=(x_1-{\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="HDA0000445176530000037.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,y_1-{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="HDA0000445176530000037.png"\; state="\{\{state\}\}">y</figure-callout>}}_0,z_1-z_0)$ Formula five

${\stackrel{\&RightArrow;}{m}}_2=(x_2-{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="HDA0000445176530000037.png"\; state="\{\{state\}\}">x</figure-callout>}}_1,y_2-{\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="HDA0000445176530000037.png"\; state="\{\{state\}\}">y</figure-callout>}}_1,z_2-z_1)$ Formula six

And then can try to achieve trajectory predictions vectorial coordinate

for:

${\stackrel{\&RightArrow;}{m}}_3=(x_3-x_2,y_3-y_2,z_3-z_2)$ Formula seven

According to formula three, can obtain:

${\stackrel{\&RightArrow;}{m}}_3={\stackrel{\text{\&RightArrow;}}{m}}_2+\mathrm{\&alpha;}\&times;{\stackrel{\&RightArrow;}{m}}_1$ Formula eight

If vectorial weights coefficient is α, α can artificially set, and α value is larger, illustrates that historical track vector is more obvious on the impact of predicted vector.

Can be in the hope of following equation group according to formula five to formula eight.

$\left\{\begin{array}{c}{x}_3-x_2=(x_2-x_1)+\mathrm{\&alpha;}\&times;(x_1-x_0)\\ y_3-y_2=(y_2-y_1)+\mathrm{\&alpha;}\&times;(y_1-y_0)\\ z_3-z_2=(z_2-z_1)+\mathrm{\&alpha;}\&times;(z_1-z_0)\end{array}\right\}$ Formula nine

By formula nine, can be derived the prediction coordinate (x of future position D ₄, y ₄, z ₄) be:

$\left\{\begin{array}{c}{x}_3={2x}_2-(1-\mathrm{\&alpha;})\&times;x_1-{\mathrm{\&alpha;x}}_0\\ y_3={2y}_2-(1-\mathrm{\&alpha;})\&times;y_1-{\mathrm{\&alpha;y}}_0\\ z_3={2z}_2-(1-\mathrm{\&alpha;})\&times;z_1-{\mathrm{\&alpha;<figure-callout\; id="0"\; label="z"\; filenames="HDA0000445176530000037.png"\; state="\{\{state\}\}">z</figure-callout>}}_0\end{array}\right\}$

Predicted vector coordinate

for:

$\begin{array}{c}{\stackrel{\&RightArrow;}{m}}_3=(x_3-x_2,y_3-y_2,z_3-z_2)\\ =(x_2-(1-\mathrm{\&alpha;})\&times;x_1-{\mathrm{\&alpha;x}}_0,y_2-(1-\mathrm{\&alpha;})\&times;y_1-{\mathrm{\&alpha;<figure-callout\; id="0"\; label="y"\; filenames="HDA0000445176530000037.png"\; state="\{\{state\}\}">y</figure-callout>}}_0,z_2-(1-\mathrm{\&alpha;})\&times;z_1-{\mathrm{\&alpha;<figure-callout\; id="0"\; label="z"\; filenames="HDA0000445176530000037.png"\; state="\{\{state\}\}">z</figure-callout>}}_0\end{array}$

When window threshold value n > 2, also can calculate successively predicted vector and future position coordinate according to formula three and formula four.

When client completes after the calculating of predicted vector, to predict the outcome and send to server end, carry out request of data, server end finds out the model in predicted vector compound condition around according to predicted vector, then according to priority determination methods, Binding distance, the visual field and interest-degree three factor constraintss are the priority index of computation model respectively

and according to priority index, calculate the download priority of compound condition model, and finally according to priority generation model pre-download queue from high to low, and according to queue sequence, the model data of certain precision is passed to client, for client, play up.In theory, n is larger for window threshold values, and it is more accurate to predict.In Forecasting Methodology, weights α can free assignment, and assignment is more close to 1, and vectorial weight is larger, also larger to track predicted impact, otherwise impact is less.Because weight is to reduce successively by the form of exponential depth, when power number reaches some, it is very little that vectorial weight becomes, and almost close to 0, therefore, in practical operation, getting window threshold values is 20, to facilitate experiment.

F. server end sends the model based coding stream node file of storage to client, client is after model based coding stream node file, precision determination module carries out judgement and the checking of precision to the data of downloading from server end, completes and receives mating of data and the required precision of user; Then data processing module carries out denoising to the model based coding stream node file receiving, and last modeling rendering module is to carrying out model based coding stream node file decoding, and the dynamic real-time rendering that completes model is drawn.

It is the gradual decode procedure of model that the real-time rendering of client models is drawn essence.Model is after preliminary treatment, and the least unit of transmitting in network is called stream node Node.

Node is defined as follows:

Node _i＝{i _index，x _i，y _i，z _i，Triangles _delete，Triangles _change，precision _i}

Wherein, i _index: be node index number, x _i, y _i, z _i. be apex coordinate information, Triangles _deletefor the triangle of deleting after edge contraction, Triangles _deletefor the triangle changing after edge contraction, precision _i. be model accuracy corresponding to present node, wherein triangle is represented by vertex index.

Decode procedure is the inverse operation of cataloged procedure, take Node as least unit, adopts edge joint to receive the method for limit decoding, and because the data volume of each node Node is very little, real-time rendering that therefore can implementation model is drawn.Model download with procedure for displaying in be a complete continual process, user receives in Node file when precision meets user's needs, i.e. precision _i=p _averagetime, just can stop downloading, play up simultaneously and show the file of all having downloaded.Whole process is continuous just as flowing water, so progressive transmission is also referred to as stream transmission.This method has solved that layer mode bring comprises the problem that layer that detailed information is more more needs problem that the long period waits for and model rendering effect to jump.

Adopt the present invention to carry out the intelligent progressive formula transmission of threedimensional model, the data variance that can receive according to displaying live view effect and the different user of client carries out the embodiment of effect.Respectively different user being received to the volume of transmitted data that same experimental model and same user adopt different transmission methods to receive different experiments models below tests.

Experiment one

The present embodiment chooses test0001 and two users of test0002 show as effect, and the data volume receiving according to two subscription clients analyzes contrast, and its result is as shown in Fig. 3-3a to Fig. 3-3b.

Fig. 3-3a scene be take user test0001 as main perspective, according to the demand of user test0001, and the effect showing in real time, test0001 client is 307KB to model A data volume, corresponding precision is 0.415, and the data volume that receives Model B is 457KB, and corresponding precision is 0.620.Fig. 3-3b scene is to take user test0002 as main perspective, according to the demand of user test0002, and the effect showing in real time, test0001 client is 425KB to model A data volume, corresponding precision is 0.575, and the data volume that receives Model B is 364KB, and corresponding precision is 0.493.

According to client to model data amount can find out, different clients, because trajectory predictions result is different, therefore also different to the demand of same model, even conduct interviews in Same Scene simultaneously, different user is also not quite similar to the requirement of same model, and the data that client arrives are also not exclusively the same.Different to the demand of model, the data precision receiving is also variant.Therefore the present invention can reach the effect of different user difference transmission, make different user make full use of limited bandwidth, the data that transmission needs, effectively reduce transmission volume, reduce period of reservation of number, avoid the delay causing due to network bandwidth limitations, effectively raised the real-time that user browses.

Experiment two

This experiment adopts traditional progressive transmission method to receive different model datas from intelligent progressive formula transmission method of the present invention for same user and compares analysis.

Experimental model A, B, C, D priority reduce successively, and data volume is respectively 737KB, 584KB, 377KB, 964KB.For embodying pre-download effect, in experiment, customer location remains unchanged, and can keep like this user before a pre-download process does not complete, and can not enter another one pre-download process, more can experience experiment effect.

Same user is adopted respectively to traditional progressive transmission strategy and intelligent progressive formula transmission policy Access Model A, B, C, D, elapsed time T ₂after, can monitor the model data that subscriber's local receives, the data volume comparison diagram that conventional method and intelligent method transmit different models as shown in Figure 4, the volume of transmitted data line that in figure, heavy line is conventional method, the volume of transmitted data line that fine line is intelligent method.

In Fig. 4, heavy line has shown under traditional progressive transmission method, model progressive download, after model A download is complete, bandwidth is given Model B and is downloaded, download model C again after Model B completes, with this, sequentially carry out, the condition that the model that priority is low starts to download is that the whole downloads of model data that priority is high are complete, and accuracy value is 1.Through T ₂after time, adopt traditional progressive transmission strategy, the data of model A, B, C are all downloaded, and accuracy value is 1, and the data division of model D is downloaded, and precision is 0.653.

Adopt intelligent progressive formula transmission method of the present invention, through T ₂after time, model A, B, C have only downloaded respectively and have met user's precision p _averagepartial data, do not download completely, as shown in Fig. 4 fine line, the precision of model A is approximately 0.753, and the precision of Model B is approximately 0.8 left and right, and the precision of MODEL C only has 0.6 left and right, model D, because user's precision prescribed is higher, has downloaded most of data, time T ₂time accuracy value be approximately 0.9.

Elapsed time T ₂after, by Fig. 4, can be found, the data of lower 4 models of traditional approach are downloaded total amount and are greater than the data download total amount of employing the present invention to 4 models.That is to say, traditional approach has been wasted time overhead in the download data useless to user, and hash occupies bandwidth overlong time, causes period of reservation of number long, and access real-time is poor; And adopt transmission means of the present invention, 4 model data amounts obviously decline, through after a period of time, model D reaches user's permissible accuracy, stop downloading, this Time Bandwidth has been given other models that need to download, enters another one pre-download process, and the time that pre-download has been carried out is t=T ₂-T '.Visible, adopt method of the present invention to carry out the transfer of data of threedimensional model, can guarantee under the prerequisite of user's required precision, make client to data obviously reduce, effectively reduced user's stand-by period, reduced the delay of network.

Claims (3)

1. threedimensional model intelligent progressive formula transmission system, is characterized in that: comprise threedimensional model pretreatment module, server end and client, the output connection server end of described threedimensional model pretreatment module, communicates with one another between server end and client; Wherein:

Threedimensional model pretreatment module, for threedimensional model is carried out to DOP-SLOD coding, the threedimensional model encoding is carried out to precision binding, generate corresponding model based coding stream node file, and the model based coding stream node file transfer of handling well is stored to server end;

Server end, for carrying out precision and priority judgement according to the trajectory predictions result of client to meeting the model of prediction locus, thereby forms orderly pre-download queue, further by network, to client transmissions, meets the data that user requires;

Client, for completing according to user's historical movement track the following prediction of track of user and the Accuracy Matching of data streaming file, and initiates request of data to server end; The data that send according to server end are stored and precision judgement, further according to the data streaming file that meets required precision receiving, threedimensional model are carried out to real-time rendering drafting.

2. threedimensional model intelligent progressive formula transmission system according to claim 1, it is characterized in that: described server end comprises model database, data verification module and priority determination module, wherein model database is for receiving and store the model based coding stream node file of threedimensional model pretreatment module transmission, whether data verification module is for meeting required precision and verify reading in the data of surplus from model database, priority determination module is for predicting the outcome according to client movement locus, near model track is carried out to priority judgement, according to result of determination, generate pre-download queue,

Described client comprises data demand module, trajectory predictions module, precision determination module, data processing module and modeling rendering module, wherein data demand module is for sending communication request to server end, and foundation or interrupt communication, trajectory predictions module is used for according to user's historical movement trajectory predictions user's following movement locus, precision determination module is for carrying out judgement and the checking of precision to the data of downloading from server end, complete and receive mating of data and the required precision of user, data processing module carries out denoising for the model based coding stream node file to receiving, the real-time rendering that modeling rendering module has been used for threedimensional model shows, realize user's online displaying live view.

3. threedimensional model intelligent progressive formula transmission method, is characterized in that specifically comprising the following steps:

A. the threedimensional model in scene is carried out to preliminary treatment DOP-SLOD coding, the threedimensional model encoding is carried out to precision binding, generation model encoding stream node file, and deposit in server end model database;

B. user is by client-access three-dimensional internet scene, and whether client is to judging for accessing first current scene, if step C is carried out in access first, if not step D is carried out in access first;

C. user end to server sends communication link request, after communication link success, to server, send data download request, server end carries out precision judgement according to the data download request receiving and priority is judged, if precision meets the demands, carry out step F, if can not meet required precision, carry out step e;

D. the local existing model data of server calls, and carry out precision judgement, carries out step e if meet required precision, if can not meet required precision, returns to step C and sends request of data to server;

E. client is carried out trajectory predictions, and trajectory predictions result is sent to server end; Server end is according to the queue of trajectory predictions result generation model pre-download, and the data of pre-download are carried out to precision judgement, if meet required precision, carries out step F, if can not meet required precision, re-starts trajectory predictions;

F. server end sends the model based coding stream node file of storage to client, and client is decoded after model based coding stream node file, and the dynamic real-time rendering that completes model is drawn.

Images