CN107452060A - Full angle automatic data collection generates virtual data diversity method - Google Patents
Full angle automatic data collection generates virtual data diversity method Download PDFInfo
- Publication number
- CN107452060A CN107452060A CN201710502725.4A CN201710502725A CN107452060A CN 107452060 A CN107452060 A CN 107452060A CN 201710502725 A CN201710502725 A CN 201710502725A CN 107452060 A CN107452060 A CN 107452060A
- Authority
- CN
- China
- Prior art keywords
- full angle
- data collection
- diversity method
- model
- different
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/10—Constructive solid geometry [CSG] using solid primitives, e.g. cylinders, cubes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Health & Medical Sciences (AREA)
- Computer Graphics (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention belongs to data mining technology field, discloses a kind of full angle automatic data collection generation virtual data diversity method, carries out high-precision modeling to real world object, while the real material of model is simulated by simulation algorithm；By carrying out emulation change to the environment residing for model, from different perspectives, different scenes, different photoenvironments are carried out from the substantial amounts of visual pattern of full angle rapid extraction to object；In output, classified automatically automatically according to the object module and other environmental information extracted, form data set.The present invention can relatively accurately generate the virtual sample available for high dimensional data prediction；Substantial amounts of data can be generated in a short time；Greatly reduce the manpower and materials spent by existing gathered data collection；Simultaneously because model has the plasticity of height；Simulation under different views, also make it that the general type of data set is more preferable, it is more reliable.
Description
Technical field
The invention belongs to data mining technology field, more particularly to a kind of full angle automatic data collection generation dummy data set side
Method and graphical simulation system.
Background technology
The collection of available data collection more extraction and handmarking, these disclosures from public image database and video database
Database is mostly what is taken based on human vision, so the data set of collection has larger limitation in angle, content.Simultaneously
The process of the collection and process of mark requires a great deal of time and manpower.
In summary, the problem of prior art is present be：
1st, without the method for quick gathered data collection, gathered data collection does not depart from manually yet.
2nd, existing data set majority is limited in angular content, cannot full angle extraction is observed to object.
The content of the invention
The problem of existing for prior art, the invention provides a kind of full angle automatic data collection to generate dummy data set side
Method.
The present invention is achieved in that a kind of full angle automatic data collection generation virtual data diversity method, and the full angle is certainly
Dynamic collection generation virtual data diversity method carries out high-precision modeling to real world object using graphical simulation softward, while by imitative
True algorithm simulation goes out the real material of model；Emulation change is carried out to the environment residing for model by image conversion simulation softward, from
Different angle, different scenes, different photoenvironments are carried out from the substantial amounts of visual pattern of full angle rapid extraction to object；Exporting
When, classified automatically automatically according to the object module and other environmental information extracted, form data set.
Further, the full angle automatic data collection generation virtual data diversity method comprises the following steps：
Step 1, by 3DMAX or Maya modeling softwares, to the object that needs gather by three-view diagram, make basic
Geometrical model, each several part scale and details are adjusted further according to details, model is tried one's best close to the object in reality, realized high
Accuracy modeling, or directly extract the modeled required model finished in various game the resources of movie ＆ TV from network；
Step 2, sample pattern is imported in graphical simulation softward, establishes document scene.By software to sample pattern
The attributes such as the metallicity of each several part, degree of roughness, refractive index carry out respective settings, it is possessed the phase of the material in actual life
Close attribute, such as metallic luster, mirror-reflection etc., while in the suitably local textures of model, it is possessed the line of real-world object
Manage feature；The landform of the scene is carved simultaneously, adds texture, it is simulated cement pavement, meadow, soil etc. existing
Real environment.Background picture and sky picture are set simultaneously, it with environment such as city, rural areas and ambient lighting can be added.
Step 3, is rendered by software to object module and illumination is constructed, and is set up and is looked squarely, overlooks, looking up etc. no
, can be never when these tracks need to meet that the virtual video camera in software moves in orbit with the camera tracks at visual angle
Same visual angle, different azimuth photograph target object, and target object is the main body in the visual field.Operate the virtual video camera in software
Moved on track camera and image information is extracted to object module from all angles.
Step 4, change in step 1 the landform, background picture, sky picture and the ambient lighting that set to change object
Body background information, object module is simulated in different environment, different places, the visual effect under different illumination.Not
Perform step 3 step 4 repeatedly with position.
Step 5, it is true according to object module to different files to caused different pictures classification outputs in step 4
Recognize this group of pictures Folder Name and picture name main fields, suffix is added for picture further according to the order of collection picture.It is real
The classified storage of existing pictures, forms data set.
Another object of the present invention is to provide one kind using full angle automatic data collection generation virtual data diversity method
Graphical simulation system.
Advantages of the present invention and good effect are：, can be rapidly from 360 degree of full-shapes by being modeled in graphical simulation softward
The angle and the data set of content that degree rapid extraction largely needs；Realizing relatively accurately generation can be used for high dimensional data to predict
Virtual sample；Substantial amounts of data can be generated in a short time；Greatly reduce the people spent by existing gathered data collection
Power material resources；Simultaneously because model has the plasticity of height, this, which allows the technology to gather in more actual lives, to adopt
The data set collected, for example, mountain fire, blast etc..Simulation under different views, also make it that the general type of data set is more preferable, it is more reliable.
Contrast traditional data collection acquisition mode：Traditional data collection needs to spend time several weeks to obtain related resource on network, needs simultaneously
A large amount of manpowers are wanted to be labelled to the data that these are collected classification.Often the data set of 1,000,000 orders of magnitude makes and needs to spend
Take one-month period.And it is proposed that new technology, under equal conditions modeling collection only need spend 2 time-of-weeks can complete
The making of data set, its time cost reduce by more than 50%.Simultaneously the data set made by us possess more fully visual angle and
General type.So that the data set trains the neutral net come with accurate.
Brief description of the drawings
Fig. 1 is full angle automatic data collection generation dummy data set method flow diagram provided in an embodiment of the present invention.
Fig. 2 is the data set figure provided in an embodiment of the present invention by classification.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
The embodiment of the present invention by establishing real scene in the graphical simulation softward by taking illusory 4 engine as an example, and
Object model, constantly converted by automated procedures residing for complexity, illumination condition, cameras capture visual angle and the model of scene
Position.Output image classified storage is realized to classification under the corresponding file of model.
The application principle of the present invention is explained in detail below in conjunction with the accompanying drawings.
As illustrated, the virtual data diversity method of full angle automatic data collection provided in an embodiment of the present invention generation car include with
Lower step：
Step 1, the car model of 20 two high emulation is extracted from the resources of movie ＆ TV.
Step 2, auto model being imported into illusory 4 engine, it is painted, vehicle window, the part such as tire assigns material respectively,
Make it have and identical material properties in actual life.
Step 3, by 5 track cameras set up, shooting, collecting is carried out from multi-angle.
Step 4, be collected one group of picture is stored in tab file folder
Step 4, change surrounding enviroment and repeat step 3 step 4, until data set capacity reaches expectation.
The present invention carries out high-precision modeling using graphical simulation softward to real world object, while is simulated by simulation algorithm
Go out the real material of model, for example, the metallic luster of steel, the fold of clothes, mirror-reflection etc..Simulated simultaneously by image conversion
Software carries out emulation change to the environment residing for model, from different perspectives, different scenes, different photoenvironments to object carry out from
The substantial amounts of visual pattern of full angle rapid extraction.Simultaneously in output, automatically according to the object module extracted and other rings
Environment information is classified automatically.Form data set.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement made within refreshing and principle etc., should be included in the scope of the protection.
Claims (4)
1. a kind of full angle automatic data collection generates virtual data diversity method, it is characterised in that the full angle automatic data collection generation
Virtual data diversity method is generated by ambient image analogue simulation where target object and object in full angle automatic data collection
Virtual image data diversity method；High-precision modeling is carried out to real world object using graphical simulation softward, while calculated by emulating
Method, the factors such as reflective body surface, degree of roughness, metal-like are calculated, simulate the real material of model；Pass through image conversion
Simulation softward carries out emulation change to the environment residing for model, and from different perspectives, different scenes, different photoenvironments enter to object
Go from the substantial amounts of visual pattern of full angle rapid extraction；In output, automatically according to the object module extracted and other rings
Environment information carries out file designation, while is stored in different files, realizes automatic classification, forms data set.
2. full angle automatic data collection as claimed in claim 1 generates virtual data diversity method, it is characterised in that the full angle
The virtual data diversity method of automatic data collection generation car comprises the following steps：
Step 1, the car model of 20 two high emulation is extracted from video display game resource；
Step 2, auto model is imported into illusory 4 engine, it is painted, vehicle window, the part such as tire assigns material respectively, makes it
With with identical material properties in actual life；
Step 3, by 5 track cameras set up, shooting, collecting is carried out from multi-angle；
Step 4, be collected one group of picture is stored in tab file folder；
Step 5, by changing background model element at random, photoenvironment repeats step 3 step to update surrounding enviroment
Four, until data set capacity reaches desired 1,000,000 pictures.
3. full angle automatic data collection as claimed in claim 1 generates virtual data diversity method, it is characterised in that the full angle
Automatic data collection generation virtual data diversity method further comprises the steps：
Step 1, by 3DMAX or Maya modeling softwares, carry out high-precision modeling to the object that needs gather, or from net
In network required model is extracted in various game the resources of movie ＆ TV；
Step 2, sample pattern is imported in graphical simulation softward, pass through the simulation algorithm defined in graphical simulation softward
Function pair sample pattern material is set, and it is possessed the association attributes of the material in actual life；
Step 3, set up different camera tracks and image information is extracted from all angles to object module；
Step 4, called by intrinsic function, produce random position point, step 3 step 4 is performed repeatedly in diverse location；
Step 5, to caused different pieces of information collection classified storage in step 4.
A kind of 4. figure that virtual data diversity method is generated using full angle automatic data collection described in claim 1~2 any one
Change simulation system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710502725.4A CN107452060A (en) | 2017-06-27 | 2017-06-27 | Full angle automatic data collection generates virtual data diversity method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710502725.4A CN107452060A (en) | 2017-06-27 | 2017-06-27 | Full angle automatic data collection generates virtual data diversity method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107452060A true CN107452060A (en) | 2017-12-08 |
Family
ID=60487106
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710502725.4A Pending CN107452060A (en) | 2017-06-27 | 2017-06-27 | Full angle automatic data collection generates virtual data diversity method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107452060A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108509855A (en) * | 2018-03-06 | 2018-09-07 | 成都睿码科技有限责任公司 | A kind of system and method generating machine learning samples pictures by augmented reality |
| CN108596259A (en) * | 2018-04-27 | 2018-09-28 | 济南浪潮高新科技投资发展有限公司 | A method of the artificial intelligence training dataset for object identification generates |
| CN109190674A (en) * | 2018-08-03 | 2019-01-11 | 百度在线网络技术（北京）有限公司 | The generation method and device of training data |
| CN109615655A (en) * | 2018-11-16 | 2019-04-12 | 深圳市商汤科技有限公司 | A kind of method and device, electronic equipment and the computer media of determining gestures of object |
| CN109871859A (en) * | 2018-09-28 | 2019-06-11 | 北京矩视智能科技有限公司 | One kind automatically generating training set of images system |
| CN109934907A (en) * | 2019-02-14 | 2019-06-25 | 深兰科技（上海）有限公司 | A kind of sample generating method, device, medium and equipment |
| CN110059724A (en) * | 2019-03-20 | 2019-07-26 | 东软睿驰汽车技术（沈阳）有限公司 | A kind of acquisition methods and device of visual sample |
| CN111739137A (en) * | 2020-05-26 | 2020-10-02 | 复旦大学 | Method for generating three-dimensional attitude estimation data set |
| CN112257731A (en) * | 2019-07-05 | 2021-01-22 | 杭州海康威视数字技术股份有限公司 | Virtual data set generation method and device |
| CN112686988A (en) * | 2020-12-31 | 2021-04-20 | 北京北信源软件股份有限公司 | Three-dimensional modeling method, three-dimensional modeling device, electronic equipment and storage medium |
| CN113362466A (en) * | 2021-06-07 | 2021-09-07 | 浙江大学 | Free type collection method for high-dimensional material |
| CN113554045A (en) * | 2020-04-23 | 2021-10-26 | 国家广播电视总局广播电视科学研究院 | Data set manufacturing method, device, equipment and storage medium |
| CN114454325A (en) * | 2022-01-21 | 2022-05-10 | 四川农业大学 | Maintenance system and production method of prefabricated concrete component for assembly type building |
| CN112257731B (en) * | 2019-07-05 | 2024-06-28 | 杭州海康威视数字技术股份有限公司 | Virtual data set generation method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104103081A (en) * | 2014-07-14 | 2014-10-15 | 西安电子科技大学 | Virtual multi-camera target tracking video material generation method |
| CN105046731A (en) * | 2015-06-25 | 2015-11-11 | 西安理工大学 | Accident site virtual multi-attention-perspective conversion three-dimensional dynamic reproducing method |
-
2017
- 2017-06-27 CN CN201710502725.4A patent/CN107452060A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104103081A (en) * | 2014-07-14 | 2014-10-15 | 西安电子科技大学 | Virtual multi-camera target tracking video material generation method |
| CN105046731A (en) * | 2015-06-25 | 2015-11-11 | 西安理工大学 | Accident site virtual multi-attention-perspective conversion three-dimensional dynamic reproducing method |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108509855B (en) * | 2018-03-06 | 2021-11-23 | 成都睿码科技有限责任公司 | System and method for generating machine learning sample picture through augmented reality |
| CN108509855A (en) * | 2018-03-06 | 2018-09-07 | 成都睿码科技有限责任公司 | A kind of system and method generating machine learning samples pictures by augmented reality |
| CN108596259A (en) * | 2018-04-27 | 2018-09-28 | 济南浪潮高新科技投资发展有限公司 | A method of the artificial intelligence training dataset for object identification generates |
| CN109190674B (en) * | 2018-08-03 | 2021-07-20 | 百度在线网络技术（北京）有限公司 | Training data generation method and device |
| CN109190674A (en) * | 2018-08-03 | 2019-01-11 | 百度在线网络技术（北京）有限公司 | The generation method and device of training data |
| CN109871859A (en) * | 2018-09-28 | 2019-06-11 | 北京矩视智能科技有限公司 | One kind automatically generating training set of images system |
| CN109615655A (en) * | 2018-11-16 | 2019-04-12 | 深圳市商汤科技有限公司 | A kind of method and device, electronic equipment and the computer media of determining gestures of object |
| CN109934907A (en) * | 2019-02-14 | 2019-06-25 | 深兰科技（上海）有限公司 | A kind of sample generating method, device, medium and equipment |
| CN110059724A (en) * | 2019-03-20 | 2019-07-26 | 东软睿驰汽车技术（沈阳）有限公司 | A kind of acquisition methods and device of visual sample |
| CN112257731A (en) * | 2019-07-05 | 2021-01-22 | 杭州海康威视数字技术股份有限公司 | Virtual data set generation method and device |
| CN112257731B (en) * | 2019-07-05 | 2024-06-28 | 杭州海康威视数字技术股份有限公司 | Virtual data set generation method and device |
| CN113554045A (en) * | 2020-04-23 | 2021-10-26 | 国家广播电视总局广播电视科学研究院 | Data set manufacturing method, device, equipment and storage medium |
| CN113554045B (en) * | 2020-04-23 | 2024-04-09 | 国家广播电视总局广播电视科学研究院 | Data set manufacturing method, device, equipment and storage medium |
| CN111739137A (en) * | 2020-05-26 | 2020-10-02 | 复旦大学 | Method for generating three-dimensional attitude estimation data set |
| CN112686988A (en) * | 2020-12-31 | 2021-04-20 | 北京北信源软件股份有限公司 | Three-dimensional modeling method, three-dimensional modeling device, electronic equipment and storage medium |
| CN113362466A (en) * | 2021-06-07 | 2021-09-07 | 浙江大学 | Free type collection method for high-dimensional material |
| CN113362466B (en) * | 2021-06-07 | 2022-06-21 | 浙江大学 | Free type collection method for high-dimensional material |
| CN114454325A (en) * | 2022-01-21 | 2022-05-10 | 四川农业大学 | Maintenance system and production method of prefabricated concrete component for assembly type building |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107452060A (en) | Full angle automatic data collection generates virtual data diversity method | |
| CN109934914B (en) | Embedded city design scene simulation method and system | |
| US20080310707A1 (en) | Virtual reality enhancement using real world data | |
| CN109887084A (en) | A method of urban planning is used for using Immersed Virtual Reality Technology | |
| CN103226830A (en) | Automatic matching correction method of video texture projection in three-dimensional virtual-real fusion environment | |
| CN103679204A (en) | Image identification and creation application system and method based on intelligent mobile device platform | |
| CN105139445A (en) | Scenario reconstruction method and apparatus | |
| CN104915978A (en) | Realistic animation generation method based on Kinect | |
| CN111915726B (en) | Construction method of three-dimensional scene of overhead transmission line | |
| CN108280873A (en) | Model space position capture and hot spot automatically generate processing system | |
| Yang et al. | VR panoramic technology in urban rail transit vehicle engineering simulation system | |
| Zhao et al. | Autonomous driving simulation for unmanned vehicles | |
| Li et al. | OmniCity: Omnipotent city understanding with multi-level and multi-view images | |
| CN116129043A (en) | Universal three-dimensional model for fusing reality scene and construction method thereof | |
| Oke et al. | Use of augmented reality in the preservation of architectural heritage: Case of the Aqueduct Kuru Kopru (Kayseri, Turkey) | |
| Zhang et al. | Benchmarking the Robustness of Object Detection Based on Near‐Real Military Scenes | |
| Han et al. | The application of augmented reality technology on museum exhibition—a museum display project in Mawangdui Han dynasty tombs | |
| GEa | Intelligent Design Mode and Simulation of Public Environment Ecology Based on 3D Digital Technology | |
| Li et al. | Research on Landscape Architecture Modeling Simulation System Based on Computer Virtual Reality Technology | |
| CN116612256B (en) | NeRF-based real-time remote three-dimensional live-action model browsing method | |
| Lee | Case study on emerging trends in geospatial technologies for study of urban form | |
| CN112734924B (en) | Visual modeling method for roadbed slope live-action three-dimensional model | |
| Yu et al. | Object Detection Based on Digital Twin-Based Road Scenarios | |
| Campos et al. | Aerial Fire Image Synthesis and Detection. | |
| Diao et al. | Research on the key techniques on 3D virtual campus system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171208 |