CN104361581B - The CT scan data dividing method being combined based on user mutual and volume drawing - Google Patents
The CT scan data dividing method being combined based on user mutual and volume drawing Download PDFInfo
- Publication number
- CN104361581B CN104361581B CN201410568890.6A CN201410568890A CN104361581B CN 104361581 B CN104361581 B CN 104361581B CN 201410568890 A CN201410568890 A CN 201410568890A CN 104361581 B CN104361581 B CN 104361581B
- Authority
- CN
- China
- Prior art keywords
- data
- volume
- supervoxel
- volume data
- segmentation
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/24—Indexing scheme for image data processing or generation, in general involving graphical user interfaces [GUIs]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20192—Edge enhancement; Edge preservation
Landscapes
- Image Processing (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The present invention provides a kind of CT scan data dividing method being combined based on user mutual and volume drawing, includes five steps:Image smoothing, for obscuring details, strengthens border;Reconstructed volume data, the view data that previous step is treated recombinates the form of volumetric data, facilitates subsequent operation;Volume data obtained in the previous step is drawn, and according to drawing result adjusting parameter, noise in volume data is removed as far as possible;Obtained data are subjected to SuperVoxel divisions;Impose a condition progress SuperVoxel clusters, finally gives segmentation result.Volume drawing is combined by the present invention with the segmentation of CT scan data, with the characteristics of simple to operate, step is directly perceived, input is convenient, segmentation effect is preferable.
Description
Technical field
The present invention relates to a kind of CT scan data dividing method being combined based on user mutual and volume drawing.
Background technology
Industrial Computed Laminography technology (Industrial Computerized Tomography), abbreviation industry CT
Or ICT, it is the new imaging technique that computer technology is combined and produced with radiology.It is sent out using ray through object
Raw this property that decays, the data that sensor is obtained rebuild so as to obtain the three dimensional grey scale image of object to be detected, clearly
Internal structure, density of material and defective eigenpairs clear, accurate, that intuitively reflect object.Can efficiently it be examined using this technology
Survey the quality of highroad, it is possibility to have imitate the detection for avoiding the subjective factor of people from bringing and analytical error.
In traditional detection method, technical staff is needed by high-resolution CT equipment, obtains a large amount of of sampling road surface
CT is cut into slices, and further every CT section is handled.But road surface is mixed by pitch and building stones more, what is not only obtained cuts
Gray scale is approached in piece, it is difficult to is distinguished, and due to the presence of noise, is increased the difficulty of resolution, while single uninteresting read tablet
Work not only largely takes the times of technical staff but also easily causes to omit and judge by accident.
In order to solve the above problems, the CT scan data segmentation side that the present invention is combined based on user mutual and volume drawing
Method, from direct volume drawing of being cut into slices to CT, segmentation is set up on the basis of drafting and user mutual, with step it is clear,
The characteristics of simple to operate, algorithm is efficient, result is accurate, drastically increases the operating efficiency of technical staff, and improves inspection
The quality of survey.
The content of the invention
Present invention solves the technical problem that being:In the dividing processing to CT slice of datas, it is right in conventional method to overcome
Single section is carried out under the efficiency of processing presence, the problems such as error is larger, it is proposed that a kind of point based on volume drawing and interaction
Segmentation method, substantially increases efficiency.And by structure extraction, the technology such as bilateral filtering strengthens image, improves segmentation
As a result.
The technical solution adopted by the present invention is:A kind of CT scan data segmentation being combined based on user mutual and volume drawing
Method, including following five steps:
Step (1), image smoothing:Big, the shortcoming of obscurity boundary for CT scan data noise, is employed based on relatively total
The method that structure is extracted from texture being deteriorated, is obscured to image detail, but enhances the structure of image, makes border
Become apparent from;
Step (2), reconstructed volume data:For the ease of subsequent treatment, by enhanced image in step (1), according to certain
Rule, constitute volume data, and carry out volume data bilateral filtering enhancing border;
Step (3), the processing of volume data:The volume data obtained in volume drawing step display (2), adjusts rendering parameter, root
According to effect is drawn, the rendering parameter needed for adjustment finally gets rid of data unnecessary in volume data;
Step (4), SuperVoxel are divided:For the ease of operation and speed up processing, the body number that step (3) is obtained
According to SuperVoxel divisions are carried out, then according to effect is drawn, the SuperVoxel for representing object is extracted, and it is compiled
Number, it is easy to subsequent treatment;
Step (5), merging:On the basis of SuperVoxel, two parameters a, b are designed, a represents current block and surrounding
The adjoining degree of block is planted, b represents that the border that current block has with certain block accounts for the ratio on itself total border.Can be in real time by closing
And effect, the two parameters are adjusted, so as to obtain preferable result.
The principle of the present invention is:
(1) by applying CT data Direct Volume Rendering Techniques, single section segmentation is changed into body segmentation, come in quantity
Say, greatly reduce workload, effectively reduce the error caused by artifact;
(2), can be effective by the Transfer Fuction parameters in interaction adjustment volume drawing according to volume drawing result
The redundant data in volume data is removed, amount of calculation is not only reduced, and eliminates influence of this partial data to final result,
Improve segmentation effect;
(3) volume data is subjected to SuperVoxel divisions, the form clustered with SuperVoxel is gone to realize and split;Pass through
The trend of adjusting parameter Real Time Effect cluster, it is ensured that the correctness of final effect.
The advantage of the present invention compared with prior art is:
1st, to CT tomography direct volume drawings, compared to single slicing treatment, the time is greatlyd save;
2nd, target data block is chosen by adjusting rendering parameter Transfer Function, is operated more directly perceived;
3rd, segmentation is realized by the way of SuperVoxel is clustered, it is simple to operate, and can be closed by parameter Real Time Effect
And trend so that result is controllable, more accurately.
Brief description of the drawings
Fig. 1 is the overall process flow for the CT scan data dividing method being combined based on user mutual and volume drawing;
Fig. 2 is comparison diagram after image enhaucament;Wherein, a is original, and b is after strengthening.
Fig. 3 is interaction forebody-afterbody data comparison schematic diagram;Wherein, a is initial data, after b is interactive.
Fig. 4 is SLIC algorithm search scope schematic diagrames;A represents that the hunting zone of commonsense method is whole figure, and b represents SLIC
Hunting zone be 2S*2S neighborhoods.
Fig. 5 is Supervoxel schematic diagrames;Wherein a is the schematic diagram of initial data, after b is interactive, is kept as far as possible
The schematic diagram data of target's center.
After data clusters terminate centered on Fig. 6, legacy data schematic diagram is extended.Bulk for cluster data, fritter
For growth data.
Fig. 7 is cluster result schematic diagram;Wherein a is target's center's data clusters schematic diagram, and b is final result.
Embodiment
Fig. 1 gives the overall process stream for the CT scan data dividing method being combined based on user mutual and volume drawing
Journey.
The present invention provides a kind of CT scan data dividing method being combined based on user mutual and volume drawing, key step
It is described below:
1st, related total variance image enchancing method
Total variance (Total Variation, TV) minimization method is that one kind can keep side well in image reconstruction
The effective ways of edge.Method is mainly used in recovering from observed image f the approximate image u of a Piecewise Smooth, and v=u-f quilts
It is considered noise or small repeat pattern, is removed from observed image.However, in some cases, composition v is also critically important, it is special
It is not when it expresses texture.Observed image is decomposed into ideal image and Complex-valued additive random noise by TV methods, by ideograph
The solution of picture, available for image denoising, its basic model is:F=u+w wherein f are observed image, and u is ideograph to be restored
Picture, w is additive noise.
This patent strengthens image using the method (RTV) of extraction structure from texture based on related total variance.The party
Method includes the general total variance for carrying out calculation window in units of pixel, and formula is:
Wherein, R (p) is the Delta Region centered on pixel p, and q belongs to R (p);gp,qIt is a weighting function, is defined asWherein σ controls the bulk of window.In order to significantly distinguish knot from texture
Structure, re-define one it is novel be used to describe the intrinsic difference of window, form is:
The formula results in the difference in whole space.For above formula, there is an important properties:Generally,
Result window in, only the situation comprising texture is fewer than the situation comprising texture and structure side simultaneously.Can intuitively it explain
For in complex situations, in local window, the contribution of the similarity direction gradient on main side is more some larger than texture.
In order to further enhance the contrast of structure and texture, particularly significant region, by combiningWithObtain
One more effective structural texture decomposed form:
Wherein (Sp-Ip)2Control input and output can not deviate too much.New itemIt is referred to as
With respect to population variance (RTV), texture can be effectively removed from image.λ is weight, and ε is the positive number of a very little, is prevented by 0
Remove.RTV is combinedWithCharacteristic, be it is a kind of simple and it is highly effective extract structure method.Effect is as schemed
2b。
2nd, the interaction process of volume data
This part is divided into two parts, carries out bilateral filtering processing to volume data first, then interactive selection object again
Body.
Bilateral filtering (Bilateral filter) is a kind of can to protect the wave filter of side denoising.Why this can be reached
Denoising effect, is because wave filter is made up of two functions:One function is to determine filter coefficient by geometric space distance,
Another determines filter coefficient by pixel value difference.
In two-sided filter, the value of output pixel depends on the weighted array of the value of neighborhood territory pixel,
Weight coefficient ω (i, j, k, l) depends on domain of definition core:
With codomain core
Product
For interactive selection object function, mainly by adjusting the Transfer Function parameters in volume drawing, control
The observability of voxel processed is realized.Concrete principle is exactly, in voxel, and different positions and different structures have different
Transparency, we are by regulation parameter, to change the transparency of diverse location, in the case where ensureing not lose target object,
Remainder data is removed as much as possible.When preferable drawing result is reached, a weight threshold is defined, further according to now
Parameter, the weight of each voxel of backwards calculation when less than threshold value, abandons this voxel, finally gives and only includes target
The volume data of object.Effect such as Fig. 3.
3rd, the super voxel segmentation based on SLIC
Super voxel is to carry out over-segmentation to volume data, and around set region, the content of homogeneity in volume data is returned
For a class, as one surpass voxel.The segmentation of super voxel generally will not coverage goal and background borderline region, i.e. a volume data divides
After a series of super voxels, rarely super voxel crosses over the border of target and background, otherwise most of super voxels are distributed in target
Region, otherwise it is distributed in background area.Super voxel pattern can simplify the feature description of volume data, be a kind of effective acceleration work
Tool.
SLIC core algorithms are described:
(1) initialization procedure:Respectively k grid,It is that color range is worth minimum body in Grid Edge length, grid
Element is k-means initial center, and all voxels do not sort out, thus ownership class apart from dis (i)=∞, i represents a certain picture
Element.
(2) per j class iterative process:Centered on such center of gravity, (such as Fig. 4) is calculated in 2s*2s neighborhoods one by one individual
Apart from D between voxel and such central point, if D<Dis (i), then i temporarily sort out such j, D=dis (i);Readjust such.
Wherein:
SuperVoxel segmentation results such as Fig. 5 a.
4th, SuperVoxel is clustered
Cluster process is broadly divided into 3 parts, and Part I at utmost chooses target's center;Part II cluster is previous
Walk the SuperVoxel chosen;The class that Part III obtains the remaining SuperVoxel clusters of Part I to Part II
In.
(1) interactive mode similar with step 2 is used, on the basis of target data, further reduces consolidation scope, to the greatest extent
Amount extracts the centre data (such as Fig. 5 b) of object, like this, and the data volume of merging can be reduced to the full extent, improves and calculates
The efficiency of method.
(2) for the class i for being currently needed for handling, its overall circumference S (SuperVoxel phases with how many non-classes are counted
It is adjacent), its neighboring extent m with all classes aroundi,j(the SuperVoxel number adjacent with every kind of non-i classes).First, two are set
Individual threshold value initaAnd initb, the class j maximum with class i neighboring extents is then searched, ifAnd
Then class i is merged into class j, if being unsatisfactory for conditions above, the processing to class i is abandoned.Iteration merges process, until
To satisfied result (such as Fig. 7 a), threshold value init can be adjusted during mergingaAnd initb, control the trend merged.
(3) result that Part II is obtained, it is believed that be that our classifications finally to be split are total.Now, for every
The data (such as Fig. 6) that one quilt (1) is ignored, perform the operation in (2), are incorporated into certain class, can obtain final conjunction
And effect (such as Fig. 7 b).
The technology contents that the present invention is not elaborated belong to the known technology of those skilled in the art.
Although illustrative embodiment of the 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 is in the spirit and scope of the present invention that appended claim is limited and is determined, these changes
Change is it will be apparent that all utilize the innovation and creation of present inventive concept in the row of protection.
Claims (3)
1. the CT scan data dividing method being combined based on user mutual and volume drawing, including step:Image smoothing, for mould
Details is pasted, strengthens border;Reconstructed volume data, the view data that previous step is treated is recombinated after the form of volumetric data, convenience
Continuous operation;Volume data obtained in the previous step is drawn, and according to drawing result adjusting parameter, noise in volume data is removed as far as possible;Will
Obtained data carry out SuperVoxel divisions;Impose a condition progress SuperVoxel merging, finally gives segmentation result, specifically
Each step is as follows:
Step (1), image smoothing:Big, the shortcoming of obscurity boundary for CT scan data noise, is employed based on relative total variance
The method that structure is extracted from texture, image detail is obscured, but enhance the structure of image, makes border more
Clearly;
Step (2), reconstructed volume data:For the ease of subsequent treatment, will it is smooth in step (1) after image, according to certain rule
Then, volume data is constituted, and carries out the bilateral filtering enhancing border of volume data;
Step (3), the processing of volume data:Bilateral filtering processing is carried out to volume data first, then interactive selection target object again;
For interactive selection object function, by adjusting the Transfer Function parameters in volume drawing, the observability of voxel is controlled
Come what is realized;The volume data obtained in volume drawing step display (2), adjust rendering parameter, according to draw effect, adjustment needed for
Rendering parameter, finally gets rid of data unnecessary in volume data;
Step (4), SuperVoxel are divided:Super voxel is to carry out over-segmentation to volume data, around set region, by body
The content of homogeneity is classified as a class in data, as one surpasses voxel;For the ease of operation and speed up processing, step (3) is obtained
The volume data arrived carries out the SuperVoxel based on SLIC and divided, and then according to effect is drawn, extracts and represents object
SuperVoxel, and it is numbered, it is easy to subsequent treatment;
Step (5), SuperVoxel clusters:Cluster process is divided into 3 parts, and Part I chooses target's center;Part II
The SuperVoxel that cluster back is chosen;Part III obtains the remaining SuperVoxel clusters of Part I to Part II
To class in;
On the basis of SuperVoxel, two parameters a, b are designed, a represents the adjoining degree of certain block of current block Yu surrounding, b
Represent that the current block border shared with certain block accounts for the ratio on itself total border;In real time by merging effect, the two ginsengs are adjusted
Number, so as to obtain preferable result;Detailed process is:If the classification number in result is more than number of targets, or in order to accelerate to merge speed
Degree, then reduce parameter a and b;When total classification, which is counted to, to be reached to a certain degree, it should increase a and b in time, it is to avoid incoherent piece
It is merged together;According to the meaning of parameter, the scope of adjustment is between 0.2-0.4;
The step (3):Due to containing substantial amounts of noise in CT data, the influence segmentation effect that can be had a strong impact on, in order to carry
Height segmentation quality, according to the effect of volume drawing, adjusts rendering parameter, only retains target data to be split, thus remove including
Non-targeted data including noise;
The step (4):Similar data point is merged as SuperVoxel in advance, pending data can be substantially reduced
Scale;
Step (5) the parameter a and b describe the state of current index block and surrounding neighbours, by the merging effect of previous step,
Alternatively adjusting parameter, controls amalgamation result.
2. the CT scan data dividing method according to claim 1 being combined based on user mutual and volume drawing, it is special
Levy and be step (1):Using the method that structure is extracted from texture based on relative total variance, strengthen image, this method energy
Enough architectural features that image is extracted preferably from complex environment.
3. the CT scan data dividing method according to claim 1 being combined based on user mutual and volume drawing, it is special
Levy and be step (2):For the segmentation of CT data, according to by CT scan order, data are extracted one by one, according to the orientation of object,
Data are alignd according to unified form, volume data is constituted;In order to further enhance boundary characteristic, using the bilateral filter of volume data
Ripple.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410568890.6A CN104361581B (en) | 2014-10-22 | 2014-10-22 | The CT scan data dividing method being combined based on user mutual and volume drawing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410568890.6A CN104361581B (en) | 2014-10-22 | 2014-10-22 | The CT scan data dividing method being combined based on user mutual and volume drawing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104361581A CN104361581A (en) | 2015-02-18 |
CN104361581B true CN104361581B (en) | 2017-07-21 |
Family
ID=52528839
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410568890.6A Active CN104361581B (en) | 2014-10-22 | 2014-10-22 | The CT scan data dividing method being combined based on user mutual and volume drawing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104361581B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105427283B (en) * | 2015-11-06 | 2018-01-19 | 北京航空航天大学 | A kind of interactive body dividing method based on Hessian implicit functions |
CN105590100B (en) * | 2015-12-23 | 2018-11-13 | 北京工业大学 | Surpass the human motion recognition method of voxel based on identification |
CN105701860A (en) * | 2016-02-29 | 2016-06-22 | 江苏美伦影像系统有限公司 | Volume rendering method |
CN109884090B (en) * | 2019-03-07 | 2021-06-29 | 重庆大学 | CT spatial resolution measurement method for improving disk card method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE3107170A1 (en) * | 1981-02-26 | 1982-09-09 | Philips Patentverwaltung Gmbh, 2000 Hamburg | Device for reducing ring artefacts in CT images |
CN101819679A (en) * | 2010-04-19 | 2010-09-01 | 李楚雅 | Three-dimensional medical image segmentation method |
CN102402788A (en) * | 2011-12-22 | 2012-04-04 | 华南理工大学 | Method for segmenting three-dimensional ultrasonic image |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102008053110A1 (en) * | 2008-06-27 | 2009-12-31 | Siemens Aktiengesellschaft | Method for the reconstruction of CT image data |
CA2705731A1 (en) * | 2009-02-23 | 2010-08-23 | Sunnybrook Health Sciences Centre | Method for automatic segmentation of images |
-
2014
- 2014-10-22 CN CN201410568890.6A patent/CN104361581B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE3107170A1 (en) * | 1981-02-26 | 1982-09-09 | Philips Patentverwaltung Gmbh, 2000 Hamburg | Device for reducing ring artefacts in CT images |
CN101819679A (en) * | 2010-04-19 | 2010-09-01 | 李楚雅 | Three-dimensional medical image segmentation method |
CN102402788A (en) * | 2011-12-22 | 2012-04-04 | 华南理工大学 | Method for segmenting three-dimensional ultrasonic image |
Non-Patent Citations (4)
Title |
---|
A Hybrid Method for Haemorrhage Segmentation in Trauma Brain CT;Mohammadreza Soltaninejad et al;《MIUA 2014》;20140701;第2节 * |
SLIC Superpixels Compared to State-of-the-Art Superpixel Methods;R. Achanta et al.;《IEEE Transactions on Pattern Analysis and Machine Intelligence》;20121231;第34卷(第11期);2274-2282 * |
Structure Extraction from Texture via Relative Total Variation;Li Xu et al.;《ACM Transactions on Graphics》;20121130;第31卷(第6期);第3节,图2-9 * |
三维医疗影像数据自动分割算法研究;王正文 等;《信息技术》;20111231(第10期);47-49 * |
Also Published As
Publication number | Publication date |
---|---|
CN104361581A (en) | 2015-02-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Qin et al. | Pulmonary nodule segmentation with CT sample synthesis using adversarial networks | |
US8224089B2 (en) | Information processing device and information processing method, image processing device and image processing method, and computer program | |
CN104361581B (en) | The CT scan data dividing method being combined based on user mutual and volume drawing | |
CN103337074B (en) | A kind of method based on active contour model segmentation mammary gland DCE-MRI focus | |
Liu et al. | A CADe system for nodule detection in thoracic CT images based on artificial neural network | |
JP2016523154A (en) | How to use image noise information | |
Helen et al. | Segmentation of pulmonary parenchyma in CT lung images based on 2D Otsu optimized by PSO | |
CN110211139A (en) | Automatic segmentation Radiotherapy of Esophageal Cancer target area and the method and system for jeopardizing organ | |
CN105513014B (en) | A kind of multi-frame image super-resolution reconstruction method and its reconstructing system | |
WO2020238819A1 (en) | Systems and methods for image processing | |
CN103035009A (en) | Pulmonary nodule edge rebuilding and partitioning method based on computed tomography (CT) image | |
CN106600584A (en) | Tsallis entropy selection-based suspected pulmonary nodule detection method | |
US11776130B2 (en) | Progressively-trained scale-invariant and boundary-aware deep neural network for the automatic 3D segmentation of lung lesions | |
Zhu et al. | Metal artifact reduction for X-ray computed tomography using U-net in image domain | |
Nazir et al. | Efficient pre-processing and segmentation for lung cancer detection using fused CT images | |
Shi et al. | Spectral CT reconstruction via low-rank representation and region-specific texture preserving Markov random field regularization | |
Jiang et al. | Synthesize mammogram from digital breast tomosynthesis with gradient guided cGANs | |
Patel et al. | Moment curves | |
Xia et al. | Deep residual neural network based image enhancement algorithm for low dose CT images | |
Shen et al. | RETRACTED: A novel Gauss-Laplace operator based on multi-scale convolution for dance motion image enhancement [EAI Endorsed Scal Inf Syst (2022), Online First] | |
Zhang et al. | Simultaneous lung field detection and segmentation for pediatric chest radiographs | |
Chen et al. | Sketch-based Volumetric Seeded Region Growing. | |
CN102743185B (en) | Four-dimensional-computed tomography (4D-CT) image data interlayer interpolation method for lung | |
CN105427283B (en) | A kind of interactive body dividing method based on Hessian implicit functions | |
Zheng et al. | A lower false positive pulmonary nodule detection approach for early lung cancer screening |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20211105 Address after: 101300 the 7 floor of seat A of Shunyi District Shun Ke mansion, Beijing Patentee after: BEIJING UNIDRAW VR TECHNOLOGY RESEARCH INSTITUTE Co.,Ltd. Address before: 100191 No. 37, Haidian District, Beijing, Xueyuan Road Patentee before: BEIHANG University |