CN108647689A - Magic square restored method and its device based on GoogLeNet neural networks - Google Patents
Magic square restored method and its device based on GoogLeNet neural networks Download PDFInfo
- Publication number
- CN108647689A CN108647689A CN201810558003.5A CN201810558003A CN108647689A CN 108647689 A CN108647689 A CN 108647689A CN 201810558003 A CN201810558003 A CN 201810558003A CN 108647689 A CN108647689 A CN 108647689A
- Authority
- CN
- China
- Prior art keywords
- magic square
- image
- training
- neural networks
- googlenet
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of magic square restored methods and its device based on GoogLeNet neural networks, and this method comprises the following steps：Step S100：Several training images for obtaining each face of magic square, as magic square training set；Step S200：GoogLeNet neural networks are trained with magic square training set, obtain training neural network, obtain each face image of parked magic square and with the distribution of color in each face of trained neural network recognization parked magic square, obtain recognition result；Step S300：Magic square reconstitution steps are calculated according to recognition result, parked magic square is restored according to magic square reconstitution steps.The present invention can quickly restore the magic square for upsetting sequence.Another aspect of the present invention additionally provides the realization device of the above method.
Description
Technical field
The present invention relates to a kind of magic square restored methods and its device based on GoogLeNet neural networks, belong to image knowledge
Other method field.
Background technology
Magic square, ' Huarongdao ' and independent diamond chess are three big inconceivable by foreign intelligence expert and referred to as riddle circle,
And the welcome degree of magic square is even more the miracle of riddle circle, is nowadays the most popular one of puzzle.With regard to current
Speech, the robot both domestic and external for restoring magic square are substantially according to simple color sensor and mechanical operation completion,
Image Acquisition and calculating restoration algorithm need to consume a large amount of time, such as a kind of quadravalence is disclosed in CN201720779468.4
Magic square restores robot, which is mainly improved by the structure of robot, and identification of the coupled computer to magic square color,
Complete the recovery to magic square.
The magic square multi-panel image storage accuracy of image-recognizing method used in existing magic square recognition methods, acquisition is poor, and
Image recognition processes are easy by environment, illumination, interference of background etc., and recognition efficiency is low, and recognition accuracy is low.
Invention content
According to the one side of the application, a kind of magic square restored method based on GoogLeNet neural networks is provided, it should
The close friend that method realizes machine and natural environment interact, color is extracted from image data, and be not necessarily to manual intervention, with compared with
Strong robustness and very fast property.
The magic square restored method based on GoogLeNet neural networks, includes the following steps：
Step S100：It obtains recovery respectively and upsets several training images in each face of magic square under state, described in set
Training image is as magic square training set；
Step S200：The GoogLeNet neural networks are trained with the magic square training set, obtain training neural network,
The distribution of color for obtaining each face image of parked magic square and each face of parked magic square described in the trained neural network recognization, obtains
To recognition result；
Step S300：Magic square reconstitution steps are calculated according to the recognition result, according to the magic square reconstitution steps to described
Parked magic square is restored.
Optionally, the step S100 further includes being pre-processed to obtain the magic square training set to the training image；
The pretreatment includes：
Step S110：The RGB image of the training image is obtained, mean value centralization method and method for normalizing are sequentially used
It handles the RGB image and obtains normalizing image；
Step S120：Amplify the normalizing picture size using zero padding, obtains enlarged drawing；
Step S130：Sequentially to the enlarged drawing carry out Gaussian Blur is smooth, after noise reduction process, then by the enlarged drawing
Color boundaries binary mask as being converted to each piece in HSV color spaces and the magic square defined in hsv color space；
Step S140：Magic square color lump in the training image is identified using the color boundaries binary mask, according to
After the magic square color lump carries out Color Segmentation processing, fleck is obtained, each stigma is filled using image expansion Processing Algorithm
Point.
Optionally, the reconstitution steps of magic square described in the step S300 are by mechanical arm according to the magic square reconstitution steps
Complete recovery operation；
The acquisition of training image described in the step S100 is clamped by the mechanical arm described in the magic square rotation acquisition
Each face image of magic square, and the magic square upset operating by the mechanical arm and obtains upsetting the magic square under state.
Optionally, the mechanical arm is connected with photographic device control, and the photographic device obtains the magic square one side
Image after, control the mechanical arm and rotate the magic square or operation arbitrarily is upset to magic square progress.
Optionally, each face image of the training image and the parked magic square is by Microsoft's kinect depth cameras
It obtains.
Optionally, described " calculating magic square reconstitution steps according to the recognition result " is completed using two-phase algorithm.
Optionally, the step S200 is further comprising the steps of：
Step S210：The magic square training set is inputted into the GoogLeNet neural networks, using supervised learning to described
GoogLeNet neural networks are iterated training, obtain the first training neural network, and to meeting clarity and distinguishing rate requirement
Image output label, obtain label image；
Step S220：After being pre-processed to the label image, pretreatment image collection is obtained, is schemed using the pretreatment
Image set, which exercises supervision to the first training neural network by back-propagation algorithm, learns repetitive exercise, obtains the training god
Through network.
Another aspect of the present invention additionally provides a kind of magic square restoring means based on GoogLeNet neural networks, including：
Training set acquisition module, for obtaining several training figures for restoring and upsetting each face of magic square under state respectively
Picture gathers the training image as magic square training set；
Identification module is used to train GoogLeNet neural networks with the magic square training set, obtains training neural network,
The distribution of color for obtaining each face image of parked magic square and each face of parked magic square described in the trained neural network recognization, obtains
To recognition result；
Restoration module, for calculating magic square reconstitution steps according to the recognition result, according to the magic square reconstitution steps pair
The parked magic square is restored.
Optionally, the training set acquisition module includes：
Preprocessing module obtains the magic square training set for being pre-processed to the training image.
Optionally, the identification module includes：First repetitive exercise module, for will the magic square training set input described in
GoogLeNet neural networks are iterated training to the GoogLeNet neural networks using supervised learning, obtain the first instruction
Practice neural network, and to meeting clarity and distinguishing the image output label of rate requirement, obtains label image；
Secondary iteration training module obtains pretreatment image collection after being pre-processed to the label image, uses
The pretreatment image collection, which exercises supervision to the first training neural network by back-propagation algorithm, learns repetitive exercise, obtains
To the trained neural network.
Beneficial effects of the present invention include but not limited to：
(1) magic square restored method and its device provided by the present invention based on GoogLeNet neural networks, this method are logical
It crosses GoogLeNet neural networks acquired image is identified, speed is fast, accuracy is high, adapts to all kinds of environment and uses, tool
There is stronger natural environment learning ability.
(2) magic square restored method and its device provided by the present invention based on GoogLeNet neural networks, use
GoogLeNet neural networks, this neural network use Inception structures, not only further improve the accurate of prediction classification
Rate, and considerably reduce parameter amount；Robot carries out deep learning by merging for camera and mechanical arm, can be fast
Algorithm of magic square is restored in the acquisition image of speed and calculating, and stability is high, and does not need manual intervention.
(3) magic square restored method and its device provided by the present invention based on GoogLeNet neural networks are schemed from RGB
Learn and extract feature as in, there is stronger robustness and real-time；It can realize that all out of order magic squares restore, adaptability
By force, accuracy rate is high, and speed is exceedingly fast.
Description of the drawings
Fig. 1 is the magic square restored method flow schematic block based on GoogLeNet neural networks in the preferred embodiment of the present invention
Figure；
Fig. 2 is the structural representation of the magic square restoring means based on GoogLeNet neural networks in the preferred embodiment of the present invention
Figure.
Specific implementation mode
The present invention is described in detail with reference to embodiment, but the invention is not limited in these embodiments.
Referring to Fig. 1, the magic square restored method based on GoogLeNet neural networks includes the following steps：
Step S100：It obtains recovery respectively and upsets several training images in each face of magic square under state, described in set
Training image is as magic square training set；
Step S200：GoogLeNet neural networks are trained with the magic square training set, obtains training neural network, obtain
The distribution of color in each face image of parked magic square and each face of parked magic square described in the trained neural network recognization, is known
Other result；
Step S300：Magic square reconstitution steps are calculated according to the recognition result, according to the magic square reconstitution steps to described
Parked magic square is restored.
Herein, magic square used can be that international magic square standard color-mixture is combined by red, yellow, blue, green, white, orange six kinds of colors.
Or color scale is higher than 0.95 magic square between each lattice.Each face of magic square used includes multiple units of square permutation arrangement
Lattice, cell herein are known as color lump, fleck also according to the difference of processing step.The present invention is fast using deep neural network
The color characteristic in each face of magic square in speed extraction image, you can calculate the algorithm of magic square recovery, robot passes through restoration algorithm
All magic squares for upsetting sequence can be restored.
The present invention carries out a series of calculating (convolution using neural network to the image in each face of parked magic square of input
Layer, pond layer, ReLU activation primitives layer, full articulamentum and classification layer), show that preferably some parameters identify color, by this
A little parameters save as model.When restoring magic square, then call the parameter in this model that color is identified.
Preferably, the step S100 further includes being pre-processed to obtain the magic square training set to the training image；
The pretreatment includes：
Step S110：The RGB image of the training image is obtained, mean value centralization method and method for normalizing are sequentially used
It handles the RGB image and obtains normalizing image；
Step S120：Amplify the normalizing picture size using zero padding, obtains enlarged drawing；
Step S130：Sequentially to the enlarged drawing carry out Gaussian Blur is smooth, after noise reduction process, then by the enlarged drawing
Color boundaries binary mask as being converted to each piece in HSV color spaces and the magic square defined in hsv color space；
Step S140：Magic square color lump in the training image is identified using the color boundaries binary mask, according to
After the magic square color lump carries out Color Segmentation processing, fleck is obtained, each stigma is filled using image expansion Processing Algorithm
Point.
The amplified normalizing image meets the GoogLeNet neural networks input size requirement.Stigma herein
Point is that each unit lattice are formed by color difference block after above-mentioned processing on each face of magic square in the training image.Preferably, described
The size of convolution kernel used is 3 × 3 during Gaussian Blur is smooth.
The present invention captures the high resolution R GB images of magic square training image, and the information of reservation color lump that can be more complete, is next
The color data processing of step is prepared.The identification accuracy of gained training image can be improved by pretreatment, improved and passed through the instruction
Practice identification accuracy and recognition efficiency of the neural network to magic square image of collection training.Gained training image, which is concentrated, includes at least place
The image in each face of magic square under reset condition, while acquiring the image in each face under a large amount of out of order state.Out of order state
Under each face amount of images it is more, training gained GoogLeNet neural networks identification accuracy and efficiency it is higher.
In one embodiment, described pre-process includes：It is captured using mean value centralization and method for normalizing processing
Picture size is amplified to the size of deep neural network input requirements using zero padding by image, is smoothly schemed using Gaussian Blur
Picture simultaneously reduces high frequency noise, then converts the image into HSV color spaces, uses the magic square block face defined in hsv color space
Color boundary binary mask identifies image magic square color lump.After Color Segmentation processing, it is small that some can be left in the target image
Spot can be very good the fleck of filling target image with image expansion Processing Algorithm.
Preferably, the reconstitution steps of magic square described in the step S300 are by mechanical arm according to the magic square reconstitution steps
Complete recovery operation；The acquisition of training image described in the step S100 is clamped by mechanical arm described in the magic square rotation acquisition
Each face image of magic square, and the magic square upset operating by the mechanical arm and obtains upsetting the magic square under state.
Preferably, the mechanical arm is connected with photographic device control, and the photographic device obtains the magic square one side
Image after, control the mechanical arm and rotate the magic square or operation arbitrarily is upset to magic square progress.
Preferably, the training image and each face image of the parked magic square are obtained by Microsoft's kinect depth cameras
It takes.
The embodiment of the present invention coordinates depth god using Microsoft's kinect depth cameras monitoring magic square surface, depth camera
The description etc. that can quickly identify all colours through network, build cube.Depth camera can obtain RGB image.
Preferably, described " calculating magic square reconstitution steps according to the recognition result " is completed using two-phase algorithm.
The calculating of magic square recovering step herein is referred to as using two-phase algorithm (two-phase algorithm)
" min2phase " is carried out.
The present invention utilizes magic square in deep neural network rapid extraction image by the RGB image in each face of crawl magic square
The color characteristic in each face, you can calculate magic square recovery algorithm, robot by restoration algorithm can by it is all upset it is suitable
The magic square of sequence is restored.
Preferably, the step S200 is further comprising the steps of：
Step S210：The magic square training set is inputted into the GoogLeNet neural networks, using supervised learning to described
GoogLeNet neural networks are iterated training, obtain the first training neural network, and to meeting clarity and distinguishing rate requirement
Image output label, obtain label image；
Step S220：After being pre-processed to the label image, pretreatment image collection is obtained, is schemed using the pretreatment
Image set, which exercises supervision to the first training neural network by back-propagation algorithm, learns repetitive exercise, obtains the training god
Through network.
Clarity and rate of distinguishing require to be preset as needed, or meet the figure that international magic square color matching requires
Picture.The present invention is trained GoogLeNet neural networks using the training image for largely upsetting sequence magic square, records and compares
The algorithm that training uses every time, to obtain more quick magic square restoration algorithm.
Referring to Fig. 2, another aspect of the present invention additionally provides the magic square restoring means based on GoogLeNet neural networks,
Including：
Training set acquisition module 100, for obtaining several training restored and upset each face of magic square under state respectively
Image gathers the training image as magic square training set；
Identification module 200 obtains training nerve net for training GoogLeNet neural networks with the magic square training set
Network obtains the color point in each face image of parked magic square and each face of parked magic square described in the trained neural network recognization
Cloth obtains recognition result；
Restoration module 300, for calculating magic square reconstitution steps according to the recognition result, according to the magic square reconstitution steps
The parked magic square is restored.
Preferably, the training set acquisition module 100 includes：
Preprocessing module obtains the magic square training set for being pre-processed to the training image.
Preferably, the identification module 200 includes：
First repetitive exercise module, for the magic square training set to be inputted the GoogLeNet neural networks, using prison
Educational inspector practises and is iterated training to the GoogLeNet neural networks, obtains the first training neural network, and to meeting clarity
Image output label with rate requirement is distinguished, obtains label image；
Secondary iteration training module obtains pretreatment image collection after being pre-processed to the label image, uses
The pretreatment image collection, which exercises supervision to the first training neural network by back-propagation algorithm, learns repetitive exercise, obtains
To the trained neural network.
The above is only several embodiments of the present invention, not any type of limitation is done to the present invention, although this hair
It is bright to be disclosed as above with preferred embodiment, however not to limit the present invention, any person skilled in the art is not taking off
In the range of technical solution of the present invention, makes a little variation using the technology contents of the disclosure above or modification is equal to
Case study on implementation is imitated, is belonged in technical proposal scope.
Claims (10)
1. a kind of magic square restored method based on GoogLeNet neural networks, which is characterized in that include the following steps：
Step S100：Several training images for restoring and upsetting each face of magic square under state are obtained respectively, gather the training
Image is as magic square training set；
Step S200：The GoogLeNet neural networks are trained with the magic square training set, obtains training neural network, obtain
The distribution of color in each face image of parked magic square and each face of parked magic square described in the trained neural network recognization, is known
Other result；
Step S300：Magic square reconstitution steps are calculated according to the recognition result, are waited for again described according to the magic square reconstitution steps
Former magic square is restored.
2. the magic square restored method according to claim 1 based on GoogLeNet neural networks, which is characterized in that described
Step S100 further includes being pre-processed to obtain the magic square training set to the training image；
The pretreatment includes：
Step S110：The RGB image of the training image is obtained, mean value centralization method and method for normalizing is sequentially used to handle
The RGB image obtains normalizing image；
Step S120：Amplify the normalizing picture size using zero padding, obtains enlarged drawing；
Step S130：Sequentially carry out that Gaussian Blur is smooth, after noise reduction process, then the enlarged drawing is turned to the enlarged drawing
It is changed to each piece in HSV color spaces and the magic square defined in hsv color space of color boundaries binary mask；
Step S140：Magic square color lump in the training image is identified using the color boundaries binary mask, according to described
After magic square color lump carries out Color Segmentation processing, fleck is obtained, each fleck is filled using image expansion Processing Algorithm.
3. the magic square restored method according to claim 1 based on GoogLeNet neural networks, which is characterized in that described
The reconstitution steps of magic square described in step S300 are to complete recovery operation according to the magic square reconstitution steps by mechanical arm；
The acquisition of training image described in the step S100 is clamped the magic square rotation by the mechanical arm and obtains the magic square
Each face image, and the magic square upset operating by the mechanical arm and obtains upsetting the magic square under state.
4. the magic square restored method according to claim 3 based on GoogLeNet neural networks, which is characterized in that described
Mechanical arm is connected with photographic device control, after the photographic device obtains the image of the magic square one side, controls the machine
Tool arm rotates the magic square or arbitrarily upsets operation to magic square progress.
5. the magic square restored method according to claim 1 based on GoogLeNet neural networks, which is characterized in that described
Each face image of training image and the parked magic square is obtained by Microsoft's kinect depth cameras.
6. the magic square restored method according to claim 1 based on GoogLeNet neural networks, which is characterized in that described
" calculating magic square reconstitution steps according to the recognition result " is completed using two-phase algorithm.
7. the magic square restored method according to claim 1 based on GoogLeNet neural networks, which is characterized in that described
Step S200 is further comprising the steps of：
Step S210：The magic square training set is inputted into the GoogLeNet neural networks, using supervised learning to described
GoogLeNet neural networks are iterated training, obtain the first training neural network, and to meeting clarity and distinguishing rate requirement
Image output label, obtain label image；
Step S220：After being pre-processed to the label image, pretreatment image collection is obtained, using the pretreatment image collection
It is exercised supervision to the first training neural network by back-propagation algorithm and learns repetitive exercise, obtain the trained nerve net
Network.
8. a kind of magic square restoring means based on GoogLeNet neural networks, which is characterized in that including：
Training set acquisition module, for obtaining several training images for restoring and upsetting each face of magic square under state, collection respectively
The training image is closed as magic square training set；
Identification module is obtained training neural network, be obtained for training GoogLeNet neural networks with the magic square training set
The distribution of color in each face image of parked magic square and each face of parked magic square described in the trained neural network recognization, is known
Other result；
Restoration module, for calculating magic square reconstitution steps according to the recognition result, according to the magic square reconstitution steps to described
Parked magic square is restored.
9. the magic square restoring means according to claim 8 based on GoogLeNet neural networks, which is characterized in that described
Training set acquisition module includes：
Preprocessing module obtains the magic square training set for being pre-processed to the training image.
10. the magic square restoring means according to claim 8 based on GoogLeNet neural networks, which is characterized in that described
Identification module includes：First repetitive exercise module, for the magic square training set to be inputted the GoogLeNet neural networks,
Training is iterated to the GoogLeNet neural networks using supervised learning, obtains the first training neural network, and to meeting
Clarity and the image output label for distinguishing rate requirement, obtain label image；
Secondary iteration training module obtains pretreatment image collection, using described after being pre-processed to the label image
Pretreatment image collection, which exercises supervision to the first training neural network by back-propagation algorithm, learns repetitive exercise, obtains institute
State trained neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810558003.5A CN108647689A (en) | 2018-06-01 | 2018-06-01 | Magic square restored method and its device based on GoogLeNet neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810558003.5A CN108647689A (en) | 2018-06-01 | 2018-06-01 | Magic square restored method and its device based on GoogLeNet neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108647689A true CN108647689A (en) | 2018-10-12 |
Family
ID=63759178
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810558003.5A Pending CN108647689A (en) | 2018-06-01 | 2018-06-01 | Magic square restored method and its device based on GoogLeNet neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108647689A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109350958A (en) * | 2018-11-29 | 2019-02-19 | 北京小米移动软件有限公司 | The control method of terminal and terminal, computer readable storage medium |
| CN109684971A (en) * | 2018-12-18 | 2019-04-26 | 北京理工大学珠海学院 | Magic cube-solving robot algorithm executes method |
| CN111939553A (en) * | 2019-05-17 | 2020-11-17 | 汕头市澄海区科梦智能科技有限公司 | Magic cube operation monitoring, training and blind twisting method, system, medium and equipment |
| CN113435454A (en) * | 2021-05-21 | 2021-09-24 | 厦门紫光展锐科技有限公司 | Data processing method, device and equipment |
| CN115946134A (en) * | 2022-11-30 | 2023-04-11 | 长春大学 | Magic cube robot is separated to two fork arms based on image processing |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1361503A (en) * | 2000-12-29 | 2002-07-31 | 南开大学 | Color multi-objective fusion identifying technology and system based on neural net |
| CN101436302A (en) * | 2008-12-10 | 2009-05-20 | 南京大学 | Method for sorting colors of colorful three-dimensional model based on neural network |
| CN103207992A (en) * | 2013-04-07 | 2013-07-17 | 布法罗机器人科技（苏州）有限公司 | Character and color combined recognition method of license plates |
| CN103394191A (en) * | 2013-05-13 | 2013-11-20 | 重庆大学 | Electronic three-order magic cube game system |
| CN203634799U (en) * | 2013-11-20 | 2014-06-11 | 安徽省科普产品工程研究中心有限责任公司 | Electronic magic cube |
| CN105354530A (en) * | 2015-09-22 | 2016-02-24 | 浙江宇视科技有限公司 | Vehicle body color identification method and apparatus |
| CN206764789U (en) * | 2017-05-11 | 2017-12-19 | 武汉云云天下信息科技有限公司 | A kind of programmable robot of automatic solution magic square |
| CN107618038A (en) * | 2017-09-20 | 2018-01-23 | 东南大学 | The solution magic square device of low degree-of-freedom |
| CN107909067A (en) * | 2017-11-06 | 2018-04-13 | 泉州市雄蓝教育科技有限公司 | The examination system of intelligence generation answering card and mobile terminal Automated Assessment |
| CN107909580A (en) * | 2017-11-01 | 2018-04-13 | 深圳市深网视界科技有限公司 | A kind of pedestrian wears color identification method, electronic equipment and storage medium clothes |
| CN108079581A (en) * | 2017-12-15 | 2018-05-29 | 北京小米移动软件有限公司 | Intelligent magic cube, magic square control method, equipment and storage medium |
-
2018
- 2018-06-01 CN CN201810558003.5A patent/CN108647689A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1361503A (en) * | 2000-12-29 | 2002-07-31 | 南开大学 | Color multi-objective fusion identifying technology and system based on neural net |
| CN101436302A (en) * | 2008-12-10 | 2009-05-20 | 南京大学 | Method for sorting colors of colorful three-dimensional model based on neural network |
| CN103207992A (en) * | 2013-04-07 | 2013-07-17 | 布法罗机器人科技（苏州）有限公司 | Character and color combined recognition method of license plates |
| CN103394191A (en) * | 2013-05-13 | 2013-11-20 | 重庆大学 | Electronic three-order magic cube game system |
| CN203634799U (en) * | 2013-11-20 | 2014-06-11 | 安徽省科普产品工程研究中心有限责任公司 | Electronic magic cube |
| CN105354530A (en) * | 2015-09-22 | 2016-02-24 | 浙江宇视科技有限公司 | Vehicle body color identification method and apparatus |
| CN206764789U (en) * | 2017-05-11 | 2017-12-19 | 武汉云云天下信息科技有限公司 | A kind of programmable robot of automatic solution magic square |
| CN107618038A (en) * | 2017-09-20 | 2018-01-23 | 东南大学 | The solution magic square device of low degree-of-freedom |
| CN107909580A (en) * | 2017-11-01 | 2018-04-13 | 深圳市深网视界科技有限公司 | A kind of pedestrian wears color identification method, electronic equipment and storage medium clothes |
| CN107909067A (en) * | 2017-11-06 | 2018-04-13 | 泉州市雄蓝教育科技有限公司 | The examination system of intelligence generation answering card and mobile terminal Automated Assessment |
| CN108079581A (en) * | 2017-12-15 | 2018-05-29 | 北京小米移动软件有限公司 | Intelligent magic cube, magic square control method, equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 张雪娇: "智能魔方机器人的视觉感知与复原算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109350958A (en) * | 2018-11-29 | 2019-02-19 | 北京小米移动软件有限公司 | The control method of terminal and terminal, computer readable storage medium |
| CN109350958B (en) * | 2018-11-29 | 2022-03-11 | 北京小米移动软件有限公司 | Terminal, control method of terminal, and computer-readable storage medium |
| CN109684971A (en) * | 2018-12-18 | 2019-04-26 | 北京理工大学珠海学院 | Magic cube-solving robot algorithm executes method |
| CN111939553A (en) * | 2019-05-17 | 2020-11-17 | 汕头市澄海区科梦智能科技有限公司 | Magic cube operation monitoring, training and blind twisting method, system, medium and equipment |
| CN111939553B (en) * | 2019-05-17 | 2023-08-01 | 汕头市澄海区魔域文化有限公司 | Magic cube operation monitoring, training and blind screwing method and system, medium and equipment |
| CN113435454A (en) * | 2021-05-21 | 2021-09-24 | 厦门紫光展锐科技有限公司 | Data processing method, device and equipment |
| CN115946134A (en) * | 2022-11-30 | 2023-04-11 | 长春大学 | Magic cube robot is separated to two fork arms based on image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108647689A (en) | Magic square restored method and its device based on GoogLeNet neural networks | |
| CN106204779B (en) | Check class attendance method based on plurality of human faces data collection strategy and deep learning | |
| CN107844795B (en) | Convolutional neural network feature extraction method based on principal component analysis | |
| CN104601964B (en) | Pedestrian target tracking and system in non-overlapping across the video camera room of the ken | |
| CN105335725B (en) | A kind of Gait Recognition identity identifying method based on Fusion Features | |
| CN107507172A (en) | Merge the extra high voltage line insulator chain deep learning recognition methods of infrared visible ray | |
| CN107480730A (en) | Power equipment identification model construction method and system, the recognition methods of power equipment | |
| CN109045676B (en) | Chinese chess recognition learning algorithm and robot intelligent system and method based on algorithm | |
| CN109086754A (en) | A kind of human posture recognition method based on deep learning | |
| CN111027464A (en) | Iris identification method for convolutional neural network and sequence feature coding joint optimization | |
| CN110929687B (en) | Multi-user behavior recognition system based on key point detection and working method | |
| CN108596102A (en) | Indoor scene object segmentation grader building method based on RGB-D | |
| CN110032932B (en) | Human body posture identification method based on video processing and decision tree set threshold | |
| CN115816460B (en) | Mechanical arm grabbing method based on deep learning target detection and image segmentation | |
| CN110046574A (en) | Safety cap based on deep learning wears recognition methods and equipment | |
| CN107633229A (en) | Method for detecting human face and device based on convolutional neural networks | |
| CN111178121B (en) | Pest image positioning and identifying method based on spatial feature and depth feature enhancement technology | |
| CN108537196B (en) | Human behavior identification method of space-time distribution diagram generated based on motion history point cloud | |
| CN114972208B (en) | YOLOv 4-based lightweight wheat scab detection method | |
| CN110827304A (en) | Traditional Chinese medicine tongue image positioning method and system based on deep convolutional network and level set method | |
| CN113139497B (en) | System and method for identifying object on water surface and application based on 5G MEC | |
| CN111080574A (en) | Fabric defect detection method based on information entropy and visual attention mechanism | |
| CN115830652B (en) | Deep palm print recognition device and method | |
| CN111709365A (en) | Automatic human motion posture detection method based on convolutional neural network | |
| CN110009011A (en) | A kind of solder joint recognition methods based on image procossing |
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 |
Application publication date: 20181012 |
|
| RJ01 | Rejection of invention patent application after publication |