CN108108494A - A kind of picture classification intelligent terminal - Google Patents
A kind of picture classification intelligent terminal Download PDFInfo
- Publication number
- CN108108494A CN108108494A CN201810049884.8A CN201810049884A CN108108494A CN 108108494 A CN108108494 A CN 108108494A CN 201810049884 A CN201810049884 A CN 201810049884A CN 108108494 A CN108108494 A CN 108108494A
- Authority
- CN
- China
- Prior art keywords
- subregion
- picture
- tag along
- along sort
- intelligent terminal
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The invention discloses a kind of picture classification labeling methods of zoned diffustion, including automatically configuring corresponding tag along sort respectively for the picture pre-saved, are obtained wherein the tag along sort configured is chosen from preset at least one tag along sort.And it is with the process of accelerator key words sorting by way of cutting combination is carried out to picture it also offers a kind of improved embodiment.
Description
Technical field
The invention belongs to field of image recognition, and specifically, the present invention relates to a kind of picture classification intelligent terminals.
Background technology
With the development of society, between people interaction it is more and more, people can by mobile phone chats, send information with
And send picture etc..Simultaneously because cell-phone function is increasingly powerful, user passes through mobile phone photograph, all kinds of social networking applications and webpage
Browsing can obtain various photos and picture, so cause to store a large amount of different types and interior in the mobile phone of user
The picture of appearance.When user browses the picture in mobile phone, can only be obtained substantially according to shooting order or picture the date into
Row browsing, can not meet the needs of user easily browses particular picture.Simultaneously when needing to carry out taxonomic revision to picture, use
Family can not easily complete the taxonomic revision operation of picture on mobile phone, can only by after in picture bulk transfer to computer, then into
Row taxonomic revision operates;So cause more low to the efficiency of picture classification and browsing.And present existing picture classification
The recognizer that method is taken is roughly the same but they have the problem of a similary, and being exactly cannot be effectively right when picture is larger
Picture quickly identified, and when the too big picture of pixel is too big, it is very low for the very big processing speed of the consumption of hardware.
Therefore present applicant proposes picture is effectively split, and the new skill that multithreading identifies progress tag along sort simultaneously is carried out
Art scheme to solve the problems of prior art, that is, provides a kind of intelligence that can effectively solve more than technical problem
Terminal.
The content of the invention
It is contemplated that at least solve one of technical problem in the prior art.For this purpose, the present invention provides a kind of figure
Piece classification intelligent terminal, the intelligent terminal are mobile terminal, including:Setting unit, for according to the instruction with input, if
Surely it is used at least one tag along sort of the picture;Memory, for pre-saving the figure that the intelligent terminal acquires
Piece and the tag along sort information;It is characterized in that, the intelligent terminal is also equipped with tag along sort determination unit, for traveling through
The picture in the memory divides each picture each tag along sort that setting unit configures to carry out image knowledge according to
Not, corresponding tag along sort is automatically configured for the picture according to image recognition result.
Further, it is described corresponding tag along sort is automatically configured to picture to specifically include, the picture is uniformly drawn
It is divided into AxB subregion, image identification is then carried out to each subregion to obtain the tag along sort.
Further, it is described to automatically configure corresponding tag along sort and further include:
The image of the subregion adjacent to the AxB subregion progress divided is combined identification and obtains the tag along sort;
Wherein described combination identification includes, and selectes one of subregion, that is, xi,j, wherein i ∈ (1, A), i ∈ (1, A), then
From the subregion xi,jIt sets out, extends the subregion x to multiple directionsi,jThe colourity of the adjacent subregion expanded to is detected simultaneously
And/or gray scale, until the colourity or gray scale then stop continuing peritropous subregion extension more than a threshold value.
Further, the threshold value is the subregion xi,jWith being averaged for the colourity on the boundary line of adjacent sectors or gray scale
Value.
Description of the drawings
From following description with reference to the accompanying drawings it will be further appreciated that the present invention.Component in figure is not drawn necessarily to scale,
But it focuses on and shows in the principle of embodiment.In the figure in different views, identical reference numeral specifies correspondence
Part.
Fig. 1 is the key words sorting flow chart of one embodiment of the present of invention.
Specific embodiment
In order to enable the objectives, technical solutions, and advantages of the present invention are more clearly understood, below in conjunction with attached drawing and its implementation
Example, the present invention will be described in further detail;It should be appreciated that specific embodiment described herein is only used for explaining this hair
It is bright, it is not intended to limit the present invention.To those skilled in the art, after access is described in detail below, the present embodiment
Other systems, method and/or feature will become obvious.It is intended to all such additional systems, method, feature and advantage
It is included in this specification, is included within the scope of the invention, and protected by the appended claims.In detailed below
Describe the other feature of the disclosed embodiments, and these characteristic roots according to it is described in detail below will be aobvious and easy
See.
Embodiment one.
The present embodiment provides a kind of picture classification intelligent terminal, the intelligent terminal is mobile terminal, such as mobile phone, tablet
Computer, camera, video camera, laptop etc. can carry out the picture terminal that photo obtains in other words, including:It sets single
Member, for according to the instruction with input, setting to be used at least one tag along sort of the picture;Memory, for protecting in advance
Deposit picture and the tag along sort information that the intelligent terminal acquires;It is characterized in that, the intelligent terminal also has
Standby tag along sort determination unit, for traveling through the picture in the memory, sets up each picture separately according to and puts list
Each tag along sort of member configuration carries out image identification, and corresponding contingency table is automatically configured for the picture according to image recognition result
Label.
Further, it is described corresponding tag along sort is automatically configured to picture to specifically include, the picture is uniformly drawn
It is divided into AxB subregion, image identification is then carried out to each subregion to obtain the tag along sort.
Further, it is described to automatically configure corresponding tag along sort and further include:
The image of the subregion adjacent to the AxB subregion progress divided is combined identification and obtains the tag along sort;
Wherein described combination identification includes, and selectes one of subregion, that is, xi,j, wherein i ∈ (1, A), i ∈ (1, A), then
From the subregion xi,jIt sets out, extends the subregion x to multiple directionsi,jThe colourity of the adjacent subregion expanded to is detected simultaneously
And/or gray scale, until the colourity or gray scale then stop continuing peritropous subregion extension more than a threshold value.
The threshold value is the subregion xi,jWith the colourity or the average value of gray scale on the boundary line of adjacent sectors.
Picture recognition therein or the algorithm that uses of specific algorithm of combination identification for:
foreach imageXndo
The theme distribution of labels θ~Dirichlet (α) is sampled;θ is to tie up Dirichlet by the K of alpha parameter
Distribution
foreach labelYniofimageXndo
Z is distributed to themei~Multinominal (θ) is sampled
Always from theme zi'sIn take a label
For XnCalculate label priori:WhereinIt is YnMiddle yi
Quantity;ξ in training processni=0, η > 0;ξ in test processni> 0, η > 0.
To labels θ '~Dirichlet (| α 'n) theme distribution sampled;θ ' is by α 'nThe L dimensions of parametrization
Dirichlet is distributed
for each instancexni of Xndo
V is distributed to labeli~Multinominal (θ ') is sampled
FromIn take an example;Label viC dimension multinomial
for each tag tni in Tn of image Xndo
G is distributed to labeli~Multinominal (θ ') is sampled
From label gi'sIn take a mark
In algorithm above, y={ y1,y2,...,yL, Y represents a set for having L label, with T={ t1,
t2,...,tTRepresent and have the set of T user identifier.With D={ ([X1,T1],Y1) ..., ([XN,TN],YN) represent one have
The training set of N number of sample, whereinBeing one has MnThe bag of a example, Being one has GnThe set of a user identifier, andIt is the L in Y setn
The set of a label.Can generate above one based on image (or image-region) instance X and user identifier T (if there is
If) learning machine that is labeled carries out cluster in visual signature space and establish a prototype set C={ c1,c2,...,
cC}.Wherein xiBe a size be C vector, wherein xi,cIt is that prototype c appears in xiIn number.Certainly should be noted that
Above recognizer is only a kind of successful example, and in the practical application of method, it can be by those skilled in the art again
Other recognition methods are replaced, and the either primary object of the present embodiment or innovative point of the invention are the knowledge of zoned diffustion
The framework synchronously identified of other method, i.e. multithreading.
Embodiment two.
The present embodiment provides a kind of picture classification intelligent terminal, the intelligent terminal is configured to obtain or in advance
Picture is preserved, can be also used for:
Picture to pre-save automatically configures corresponding tag along sort respectively, wherein the tag along sort configured is from advance
It chooses and obtains at least one tag along sort of setting.
Further, the described pair of picture pre-saved automatically configures corresponding tag along sort and specifically includes, by the figure
Piece is uniformly divided into AxB subregion, and image identification is then carried out to each subregion to obtain the tag along sort, the subregion
Quantity carries out the distribution of ratio according to the aspect ratio of picture, and particular number is matched somebody with somebody according to the hardware for the hardware device for implementing this method
It puts and is configured.
Further, it is described to automatically configure corresponding tag along sort and further include:
The image of the subregion adjacent to the AxB subregion progress divided is combined identification and obtains the tag along sort;
Wherein described combination identification includes, and selectes one of subregion, that is, xi,j, wherein i ∈ (1, A), i ∈ (1, A), then
From the subregion xi,jIt sets out, extends the subregion xi,jThe colourity and/or gray scale of the adjacent subregion expanded to are detected simultaneously, directly
Extremely the colourity or gray scale then stop continuing peritropous subregion extension more than a threshold value, and the method for extension can use counterclockwise
The method of rotary expansion or simultaneously to subregion described in eight Directional Extensions.The selected of primary partition can be most intermediate by selecting
Subregion set either simultaneously from four or multiple subregions being distributed on picture, so can synchronously be combined point
With making, the process that scoring area identifies is more more, improves the processing speed of this method.
Further, the threshold value is the subregion xi,jWith being averaged for the colourity on the boundary line of adjacent sectors or gray scale
Value can select threshold point in the way of " key point " in general image processing method.
Picture recognition therein or the algorithm that uses of specific algorithm of combination identification for:
foreach image Xndo
The theme distribution of labels θ~Dirichlet (α) is sampled;θ is to tie up Dirichlet by the K of alpha parameter
Distribution
for each label Yni of image Xndo
Z is distributed to themei~Multinominal (θ) is sampled
Always from theme zi'sIn take a label
For XnCalculate label priori:WhereinIt is YnMiddle yi
Quantity;ξ in training processni=0, η > 0;ξ in test processni> 0, η > 0.
To labels θ '~Dirichlet (| α 'n) theme distribution sampled;θ ' is by α 'nThe L dimensions of parametrization
Dirichlet is distributed
for each instance xni of Xndo
V is distributed to labeli~Multinominal (θ ') is sampled
FromIn take an example;Label viC dimension multinomial
for each tag tni in Tn of image Xndo
Label distribution gi~Multinominal (θ ') is sampled
From label gi'sIn take a mark
Embodiment three.
The present embodiment provides a kind of picture classification intelligent terminal, the intelligent terminal is configured to obtain or store
Then picture carries out picture tag along sort, tag along sort includes each tag along sort for traveling through configuration respectively first, to current
Target Photo under the tag along sort traversed carries out image identification, the classification currently traversed according to image recognition result
Marker feature under label, when some marker feature is comprised in multiple Target Photos, to multiple described target figures
The feature for the marker that each pictures in piece are included is weighted or screens, under the tag along sort that is currently traversed
Marker feature, wherein, the tag along sort is associated with the marker feature without image, and the marker feature includes
Gray value at the key point position of marker and the key point position;
Object feature recognition is identified to acquired picture;The marker feature that acquired picture includes is calculated respectively
In key point position and gray value at key point position and the pass of the marker feature under each tag along sort of setting
The distance value of gray value at key point position and key point position determines the mark that the picture includes according to the distance value
The similarity of object feature and the marker feature under each tag along sort of setting, the marker that will be included with acquired picture
The similarity of feature meets the tag along sort of given threshold condition, is allocated to acquired picture, to complete to acquired figure
The classification of piece;All pictures got under same category label are stored in same file folder, while in acquisition
The thumbnail acceptance of the bid note tag along sort of picture.
Example IV.
The present embodiment provides a kind of picture classification label terminals, can be used for obtaining and preserving picture, including:
Picture to pre-save automatically configures corresponding tag along sort respectively, wherein the tag along sort configured is from advance
It chooses and obtains at least one tag along sort of setting.
Further, the described pair of picture pre-saved automatically configures corresponding tag along sort and specifically includes, by the figure
Piece is uniformly divided into AxB subregion, and image identification is then carried out to each subregion to obtain the tag along sort, the subregion
Quantity carries out the distribution of ratio according to the aspect ratio of picture, and particular number is matched somebody with somebody according to the hardware for the hardware device for implementing this method
It puts and is configured.
Further, it is described to automatically configure corresponding tag along sort and further include:
The image of the subregion adjacent to the AxB subregion progress divided is combined identification and obtains the tag along sort;
Wherein described combination identification includes, and selectes one of subregion, that is, xi,j, wherein i ∈ (1, A), i ∈ (1, A), then
From the subregion xi,jIt sets out, extends the subregion xi,jThe colourity and/or gray scale of the adjacent subregion expanded to are detected simultaneously, directly
Extremely the colourity or gray scale then stop continuing peritropous subregion extension more than a threshold value, and detection here can detect whole expansions
The whole colourity and/or gray scale for the subregion opened up, in the present embodiment by setting multiple detections on the propagation direction of subregion
Point, such as when being extended from a subregion to another subregion, uniformly detected along the direction of extension multiple points colourity and/
Or gray value is as detected value.
The method of extension can use the method for rotary expansion counterclockwise or simultaneously to subregion described in eight Directional Extensions.
Primary partition it is selected can by select most intermediate subregion set or simultaneously from four be distributed on picture or
The multiple subregions of person, so can synchronously be combined distribution makes the process of scoring area identification more more, improves the processing of this method
Speed.
Further, the threshold value is the subregion xi,jWith being averaged for the colourity on the boundary line of adjacent sectors or gray scale
Value can select threshold point in the way of " key point " in general image processing method.
Although the present invention is described by reference to various embodiments, but it is to be understood that do not departing from the present invention's above
In the case of scope, many changes and modifications can be carried out.Therefore, be intended to foregoing detailed description be considered as it is illustrative and
Nonrestrictive, and it is to be understood that following claims is intended to limit (including all equivalents) spirit and model of the present invention
It encloses.The above embodiment is interpreted as being merely to illustrate the present invention rather than limit the scope of the invention.It is reading
After the content of the record of the present invention, technical staff can make various changes or modifications the present invention, these equivalence changes and
Modification equally falls into the scope of the claims in the present invention.
Claims (4)
1. a kind of picture classification intelligent terminal, the intelligent terminal is mobile terminal, including:Setting unit, for according to
The instruction of input, setting are used at least one tag along sort of the picture;Memory obtains for preserving the intelligent terminal
Obtained picture and the tag along sort information;It is characterized in that, the intelligent terminal is also equipped with tag along sort determination unit,
For traveling through the picture in the memory, each picture is divided according to setting unit configure each tag along sort into
Row image identifies, corresponding tag along sort is automatically configured for the picture according to image recognition result.
2. intelligent terminal as described in claim 1, which is characterized in that the tag along sort determination unit automatically configures picture
Corresponding tag along sort specifically includes, and the picture is uniformly divided into AxB subregion, then carries out image to each subregion
It identifies to obtain the tag along sort.
3. intelligent terminal as claimed in claim 2, which is characterized in that described to automatically configure corresponding tag along sort and further include:
The image of the subregion adjacent to the AxB subregion progress divided is combined identification and obtains the tag along sort;
Wherein described combination identification includes, and selectes one of subregion, that is, xi,j, wherein i ∈ (1, A), i ∈ (1, A), then from institute
State subregion xi,jIt sets out, extends the subregion x to multiple directionsi,jDetect simultaneously the adjacent subregion expanded to colourity and/or
Gray scale, until the colourity or gray scale then stop continuing peritropous subregion extension more than a threshold value.
4. intelligent terminal as claimed in claim 3, which is characterized in that the threshold value is the subregion xi,jWith adjacent sectors
The average value of colourity or gray scale on boundary line.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810049884.8A CN108108494A (en) | 2018-01-18 | 2018-01-18 | A kind of picture classification intelligent terminal |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810049884.8A CN108108494A (en) | 2018-01-18 | 2018-01-18 | A kind of picture classification intelligent terminal |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108108494A true CN108108494A (en) | 2018-06-01 |
Family
ID=62219463
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810049884.8A Pending CN108108494A (en) | 2018-01-18 | 2018-01-18 | A kind of picture classification intelligent terminal |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108108494A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1340178A (en) * | 1999-08-17 | 2002-03-13 | 皇家菲利浦电子有限公司 | System and method for performing region-based image retrieval using color-based segmentation |
CN103995889A (en) * | 2014-06-03 | 2014-08-20 | 广东欧珀移动通信有限公司 | Method and device for classifying pictures |
JP2017162025A (en) * | 2016-03-07 | 2017-09-14 | 株式会社東芝 | Classification label allocation device, classification label allocation method, and program |
CN107256216A (en) * | 2017-04-17 | 2017-10-17 | 捷开通讯(深圳)有限公司 | Mobile terminal, the method and storage device for managing picture |
-
2018
- 2018-01-18 CN CN201810049884.8A patent/CN108108494A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1340178A (en) * | 1999-08-17 | 2002-03-13 | 皇家菲利浦电子有限公司 | System and method for performing region-based image retrieval using color-based segmentation |
CN103995889A (en) * | 2014-06-03 | 2014-08-20 | 广东欧珀移动通信有限公司 | Method and device for classifying pictures |
JP2017162025A (en) * | 2016-03-07 | 2017-09-14 | 株式会社東芝 | Classification label allocation device, classification label allocation method, and program |
CN107256216A (en) * | 2017-04-17 | 2017-10-17 | 捷开通讯(深圳)有限公司 | Mobile terminal, the method and storage device for managing picture |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bell et al. | Material recognition in the wild with the materials in context database | |
Kumar et al. | A visual-numeric approach to clustering and anomaly detection for trajectory data | |
CN103995889B (en) | Picture classification method and device | |
Hu et al. | Angular pattern and binary angular pattern for shape retrieval | |
CN108549870A (en) | A kind of method and device that article display is differentiated | |
Huang et al. | Clothing landmark detection using deep networks with prior of key point associations | |
Lim et al. | Scene recognition with camera phones for tourist information access | |
Joshi et al. | Automatic photo orientation detection with convolutional neural networks | |
US9542619B2 (en) | Automatic image piling | |
CN109753962A (en) | Text filed processing method in natural scene image based on hybrid network | |
Fare et al. | Ancient roman coin retrieval: A systematic examination of the effects of coin grade | |
Kuang et al. | Automatic privacy prediction to accelerate social image sharing | |
Lorenzo-Navarro et al. | Evaluation of LBP and HOG descriptors for clothing attribute description | |
Lienhart et al. | Filtering adult image content with topic models | |
Lee et al. | Moving shadow detection from background image and deep learning | |
CN108170849B (en) | Image classification marking method for partition diffusion | |
CN108108494A (en) | A kind of picture classification intelligent terminal | |
CN109948577A (en) | A kind of cloth recognition methods, device and storage medium | |
Paul et al. | An innovative approach for automatic genre-based fine art painting classification | |
El Sayed et al. | Efficient 3D point clouds classification for face detection using linear programming and data mining | |
Xu et al. | Analysis of clothing image classification models: a comparison study between traditional machine learning and deep learning models | |
Bouyerbou et al. | Hybrid image representation methods for automatic image annotation: A survey | |
El Sayed et al. | 3D face detection based on salient features extraction and skin colour detection using data mining | |
Singh et al. | Comparison of different fuzzy clustering algorithms: a replicated case study | |
Xiong et al. | A multi-scale triplet deep convolutional neural network for person re-identification |
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: 20180601 |
|
RJ01 | Rejection of invention patent application after publication |