CN108280190A - Image classification method, server and storage medium - Google Patents
Image classification method, server and storage medium Download PDFInfo
- Publication number
- CN108280190A CN108280190A CN201810071179.8A CN201810071179A CN108280190A CN 108280190 A CN108280190 A CN 108280190A CN 201810071179 A CN201810071179 A CN 201810071179A CN 108280190 A CN108280190 A CN 108280190A
- Authority
- CN
- China
- Prior art keywords
- image
- information
- default
- classified
- classification
- 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
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5862—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using texture
-
- 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
- G06F16/53—Querying
- G06F16/532—Query formulation, e.g. graphical querying
-
- 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
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- 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
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5838—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using colour
Abstract
The invention discloses a kind of image classification method, server and storage medium, this method includes:Obtain image to be classified;The characteristic information of extraction classification image predeterminable area, and characteristic information is generated into target feature vector;Target feature vector is matched with sorted index vector data in default sample library, determines the corresponding default classification type of the index vector data of successful match, and using default classification type as the target classification type of image to be classified.The present invention to image to be classified by extracting global characteristics and local feature, and compared with the index in default picture library, classification type is determined according to comparing result, it can be by the way that image be identified, realization classifies automatically to image, to improve image classification accuracy.
Description
Technical field
The present invention relates to a kind of technical field of image processing more particularly to image classification method, server and storage mediums.
Background technology
Currently, in operation system operating process, generally require that great mass of data is scanned into image and is preserved to provide at electronics
Material, facilitates inquiry and use, and in general, these data have many classification, such as identity card, contract, personal reference etc.,
Business personnel needs to carry out sorting out manually very inconvenient when the data of scanning, and possible maloperation leads to classification error, also pole
To influence efficiency.
Invention content
It is a primary object of the present invention to propose a kind of image classification method, server and storage medium, it is intended to improve figure
As the efficiency of classification and the accuracy of image classification.
To achieve the above object, the present invention provides a kind of image classification method, and described image sorting technique includes following step
Suddenly:
Obtain image to be classified;
Extract the characteristic information of the image to be classified predeterminable area, and by the characteristic information generate target signature to
Amount;
The target feature vector is matched with sorted index vector data in default sample library, determines matching
The successfully corresponding default classification type of index vector data, and using the default classification type as the image to be classified
Target classification type.
Preferably, the characteristic information of the extraction image to be classified predeterminable area, and the characteristic information is generated
Target feature vector specifically includes:
The characteristic information of the image to be classified global area is extracted according to the directionality descriptor of color and edge, is generated
Multi-dimensions histogram vector information;According to the default characteristic point for accelerating robust feature algorithm to extract the image to be classified regional area
Information, and characteristic point vector information is generated according to the default characteristic point information;It is waited for point according to described in the extraction of vision bag of words
The characteristic information of class image local area, and according to default visual word generate the corresponding visual word histogram of the characteristic information to
Measure information;
The multi-dimensions histogram vector information, characteristic point vector information and visual word histogram vectors information are generated into target
Feature vector.
Preferably, the directionality descriptor according to color and edge extracts the spy of the image to be classified global area
Reference ceases, and generates multi-dimensions histogram vector information, specifically includes:
According to the directionality descriptor of color and edge extract in the image to be classified colouring information of global area and
The colouring information and texture information are combined and generate multi-dimensions histogram vector information by texture information.
Preferably, described according to the default characteristic point for accelerating robust feature algorithm to extract the image to be classified regional area
Information, and characteristic point vector information is generated according to the default characteristic point information, it specifically includes:
Characteristic point information is preset according to accelerating robust feature algorithm to be generated according to default matrix, and according to the default feature
Point information generates characteristic point vector information.
Preferably, the characteristic information that the image to be classified regional area is extracted according to vision bag of words, and root
The corresponding visual word histogram vectors information of the characteristic information is generated according to default visual word, is specifically included:
The characteristic information that the image to be classified regional area is extracted according to vision bag of words, passes through the characteristic information
Visual word histogram vectors information corresponding with the mapping relations of the default visual word generation characteristic information.
Preferably, described by sorted index vector data progress in the target feature vector and default sample library
Match, determines the corresponding default classification type of the index vector data of successful match, and using the default classification type as described in
Before the target classification type of image to be classified, the method further includes:
According to the directionality descriptor at color and edge, robust feature algorithm and vision bag of words is accelerated to extract sample graph
Index vector data are generated as the vector information in data, and according to the vector information.
Preferably, the characteristic information of the extraction image to be classified predeterminable area, and the characteristic information is generated
After target feature vector, the method further includes:
Described eigenvector is matched with sorted index vector data in default sample library, is in matching result
When matching unsuccessful, the image to be classified is set as default classification type.
Preferably, described by sorted index vector data progress in the target feature vector and default sample library
Match, determines the corresponding default classification type of the index vector data of successful match, and using the default classification type as described in
After the target classification type of image to be classified, the method further includes:
The default sample library is updated according to the target classification type of the image to be classified, and receives inquiry and modification refers to
It enables, the image in updated default sample library is inquired and changed.
In addition, to achieve the above object, the present invention also proposes that a kind of server, the server include:Memory, processing
Device and it is stored in the image classification program that can be run on the memory and on the processor, described image sort program is matched
It is set to the step of realizing image classification method as described above.
In addition, to achieve the above object, the present invention also proposes a kind of storage medium, and image is stored on the storage medium
The step of sort program, described image sort program realizes image classification method as described above when being executed by processor.
The image classification method that technical solution of the present invention proposes first carries out feature information extraction, feature to image to be classified
Information includes global characteristics and local feature, and characteristic information is generated target feature vector, then by target feature vector with
Index vector data in default picture library are compared, and classification type is determined according to comparing result.Technical solution of the present invention passes through
Image is identified, realization classifies automatically to image, improves the efficiency of image classification and the accuracy of image classification.
Description of the drawings
Fig. 1 is the server architecture schematic diagram for the hardware running environment that the embodiment of the present invention is related to;
Fig. 2 is the flow diagram of image classification method first embodiment of the present invention;
Fig. 3 is the structural schematic diagram of image classification system of the present invention;
Fig. 4 is the flow diagram of image classification method second embodiment of the present invention;
Fig. 5 is the structure diagram of CEDD operation principles of the present invention;
Fig. 6 is the flow diagram of image classification method 3rd embodiment of the present invention;
Fig. 7 is the flow diagram of image classification method fourth embodiment of the present invention.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific implementation mode
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
Referring to Fig.1, Fig. 1 is the server architecture schematic diagram for the hardware running environment that the embodiment of the present invention is related to.
As shown in Figure 1, the server may include:Processor 1001, such as CPU, communication bus 1002, user interface
1003, network interface 1004, memory 1005.Wherein, communication bus 1002 is for realizing the connection communication between these components.
User interface 1003 may include display screen, input unit such as keyboard, and optional user interface 1003 can also include standard
Wireline interface, wireless interface.Network interface 1004 may include optionally that (such as WI-FI connects standard wireline interface and wireless interface
Mouthful).Memory 1005 can be high-speed RAM memory, can also be nonvolatile memory, such as magnetic disk storage.Storage
Device 1005 optionally can also be the storage device independently of aforementioned processor 1001.
It will be understood by those skilled in the art that server architecture shown in Fig. 1 does not constitute the restriction to server, it can
To include either combining certain components or different components arrangement than illustrating more or fewer components.
As shown in Figure 1, as may include operating system, network communication mould in a kind of memory 1005 of storage medium
Block, Subscriber Interface Module SIM and image classification program.
In server shown in Fig. 1, network interface 1004 is mainly used for connecting network, with network into row data communication;
User interface 1003 is mainly used for connecting user terminal, with terminal into row data communication;Server of the present invention passes through processor
The image classification program stored in 1001 calling memories 1005, and execute following operation:
Obtain image to be classified;
Extract the characteristic information of the image to be classified predeterminable area, and by the characteristic information generate target signature to
Amount;
The target feature vector is matched with sorted index vector data in default sample library, determines matching
The successfully corresponding default classification type of index vector data, and using the default classification type as the image to be classified
Target classification type.
Further, processor 1001 can call the image classification program stored in memory 1005, also execute following
Operation:
The characteristic information of the image to be classified global area is extracted according to the directionality descriptor of color and edge, is generated
Multi-dimensions histogram vector information;According to the default characteristic point for accelerating robust feature algorithm to extract the image to be classified regional area
Information, and characteristic point vector information is generated according to the default characteristic point information;It is waited for point according to described in the extraction of vision bag of words
The characteristic information of class image local area, and according to default visual word generate the corresponding visual word histogram of the characteristic information to
Measure information;
The multi-dimensions histogram vector information, characteristic point vector information and visual word histogram vectors information are generated into target
Feature vector.
Further, processor 1001 can call the image classification program stored in memory 1005, also execute following
Operation:
According to the directionality descriptor of color and edge extract in the image to be classified colouring information of global area and
The colouring information and texture information are combined and generate multi-dimensions histogram vector information by texture information.
Further, processor 1001 can call the image classification program stored in memory 1005, also execute following
Operation:
Characteristic point information is preset according to accelerating robust feature algorithm to be generated according to default matrix, and according to the default feature
Point information generates characteristic point vector information.
Further, processor 1001 can call the image classification program stored in memory 1005, also execute following
Operation:
The characteristic information that the image to be classified regional area is extracted according to vision bag of words, passes through the characteristic information
Visual word histogram vectors information corresponding with the mapping relations of the default visual word generation characteristic information.
Further, processor 1001 can call the image classification program stored in memory 1005, also execute following
Operation:
According to the directionality descriptor at color and edge, robust feature algorithm and vision bag of words is accelerated to extract sample graph
Index vector data are generated as the vector information in data, and according to the vector information.
Further, processor 1001 can call the image classification program stored in memory 1005, also execute following
Operation:
Described eigenvector is matched with sorted index vector data in default sample library, is in matching result
When matching unsuccessful, the image to be classified is set as default classification type.
Further, processor 1001 can call the image classification program stored in memory 1005, also execute following
Operation:
The default sample library is updated according to the target classification type of the image to be classified, and receives inquiry and modification refers to
It enables, the image in updated default sample library is inquired and changed.
The present embodiment through the above scheme, by extracting global characteristics and local feature to image to be classified, and with it is pre-
If the index in picture library is compared, classification type is determined according to comparison similarity, can be by the way that image be identified, realization pair
Image is classified automatically, to improve image classification accuracy.
Based on above-mentioned hardware configuration, image classification method embodiment of the present invention is proposed.
It is the flow diagram of image classification method first embodiment of the present invention with reference to Fig. 2, Fig. 2.
In the first embodiment, described image sorting technique includes the following steps:
Step S10, server obtain image to be classified;
It should be noted that the executive agent of the present embodiment can be server, connected by mobile terminal and server
It connects, the mobile terminal can be that the terminal devices such as tablet, computer are illustrated by taking computer as an example in the present embodiment, can also be
Other can realize that the terminal device of same or similar function, the present embodiment are not restricted this.
Image to be sorted is uploaded to the mobile terminal by user by mobile terminal, is equipped on the mobile terminal
The interface of classification processing, such as application program (Application, APP) on the mobile terminal is installed, by applying journey
Image to be classified is uploaded to the server and handled by sequence, can also be other interface forms, and the present embodiment does not limit this
System.
The reference material or Copy of ID Card that the information to be sorted can be issued for personal reference report, different bank
Deng the various image informations being scanned by scanner or other scanning devices, described image information preservation can be provided at electronics
Material.
Step S20 extracts the characteristic information of the image to be classified predeterminable area, and the characteristic information is generated target
Feature vector;
It can get and wait in the present embodiment it is understood that the predeterminable area is global area and regional area
The global area of classification image and regional area, so as to carry out feature extraction by various aspects, and the feature extracted are led to
It crosses preset algorithm and generates feature vector, image to be sorted can be converted to digitized parameter information, it is easier to realize to institute
State the identification of image to be classified.
In the concrete realization, the characteristic information in global area can be first extracted, reaches pre- in the characteristic information matching degree
If when value, then extracting the characteristic information of regional area, to reduce the influence of other noise images, the accurate of feature extraction is improved
Degree, can also extract the characteristic information of the global area and regional area in the image to be classified, and the feature is believed simultaneously
Breath generates feature vector makes the identification degree higher of the image to be classified to improve the match parameter of feature vector.
In the present embodiment, described eigenvector can be the characteristic information that will be extracted, and feature ginseng is generated according to preset algorithm
Number, and the characteristic parameter is clustered, to obtain the feature vector of the image to be classified, described eigenvector can be
With directive characteristic information, to improve the accuracy of the image to be classified identification.
Step S30 matches the target feature vector with sorted index vector data in default sample library,
It determines the corresponding default classification type of the index vector data of successful match, and the default classification type is waited for point as described in
The target classification type of class image.
It should be noted that the default sample library is the sample preserved according to categorized good historical sample image
Product library, the historical sample image in the default sample database is the image pattern filled in each classification, and described image sample can
Different reference materials is can relate to, needs the reference material being filled into a sample, and according to the direction of color and edge
Property descriptor (Color and Edge DirectivityDescriptor, CEDD), accelerate robust feature (Speeded Up
Robust Features, SURF) and vision bag of words (Bag Of Words Model, BOVW), extract the spy in sample
Sign, and accordingly generate index vector data.
It is understood that the default classification type can be Copy of ID Card, the Certificate of House Property and loan certificate etc., lead to
It crosses and matches image to be classified with the classified image data in default sample library, in successful match, by described in
Target classification type of the image type of the image data of classification as the image to be classified.
In the concrete realization, the image to be classified and all sample images in the sample library can be compared,
And similarity value is generated according to the similarity of comparing result, similarity is generated according to the result being compared with all sample images
List, and the sequence of the similarity list from high to low is arranged, select the highest sample image of similarity corresponding
Target classification type of the image type as the image to be classified, such as the image to be classified and the phase in the sample library
Like degree list be filename A be 90%, filename B is 75% etc., choose similarity highest 90% corresponding point of filename A
Class Type is as the corresponding target classification type of the image to be classified.
The structural schematic diagram of image classification system as shown in Figure 3, described image categorizing system may include Image Management
System and to scheme to search dsp subsystem, the Image Management subsystem and respectively independently deployment is run to scheme to search dsp subsystem.It is described
Image Management subsystem the image to be classified is transferred to it is described to scheme to search dsp subsystem, and to scheme to search dsp subsystem to image
Classify, and classification results are returned into Image Management subsystem.It, will be described after Image Management subsystem receives recognition result
Image to be classified categorizedly preserves.
In the concrete realization, it can determine which image classification picture library needs first to scheme to search in drawing system, determine
After classification just image pattern is filled into each classification.Such as personal reference material, the reference material sample that different bank is issued
All difference such as formula layout, so the different reference materials being related to are required for filling a sample, system can foundation
CEDD characteristics algorithms, SURF algorithm and BOVW models, extract the feature in sample, and accordingly generate index, then will wait for point
Class image scanning is at electronic image and uploads to image management system, and image management system is again by image transmitting to scheme to search figure system
System goes out feature vector to scheme to search after drawing system receives also according to CEDD characteristics algorithms, SURF algorithm and BOVW model extractions,
And compared with the aspect indexing in picture library, the classification of each image is obtained according to similarity, and return the result to Image Management
After the image management system receives recognition result, image is categorizedly preserved for system.
Scheme provided in this embodiment, server by extracting global characteristics and local feature to image to be classified, and
It is compared with the index in default picture library, classification type is determined according to comparison similarity, it can be real by the way that image is identified
Now classified automatically to image, to improve image classification accuracy.
Further, as shown in figure 4, proposing image classification method second embodiment of the present invention based on first embodiment,
In the present embodiment, the step S20 is specifically included:
Step S201 extracts the feature of the image to be classified global area according to the directionality descriptor of color and edge
Information generates multi-dimensions histogram vector information;
It should be noted that directionality descriptor (the Color and EdgeDirectivity of the color and edge
Descriptor, CEDD), it combines the color and texture information of image, generates one 144 histograms.CEDD histograms
Figure information is made of six regions, and six regions are exactly the 6 dimensional vector histograms extracted, then in the every of these texture informations
The 24 dimension colouring informations that color module extracts are added in one-dimensional, thus color and texture can effectively be combined,
The final histogram information for obtaining 6*24=144 dimensions.
Step S202 believes according to the default characteristic point for accelerating robust feature algorithm to extract the image to be classified regional area
Breath, and characteristic point vector information is generated according to the default characteristic point information;
It is understood that the acceleration robust feature algorithm (Spleeded Up Robust Features, SURF),
It is a kind of steady local feature region detection and description algorithm, the feature extraction to more be refined to image to be classified,
In the present embodiment, the feature extraction that part is carried out by SURF algorithm improves image to be classified to refine image to be classified
Identification degree.
Step S203, extracts the characteristic information of the image to be classified regional area according to vision bag of words, and according to
Default visual word generates the corresponding visual word histogram vectors information of the characteristic information;
It should be noted that the vision bag of words (Bag of Visual Word, BOVW) are by presetting visual word
The characteristic information of the image to be classified is extracted, to be treated by the weighted value of characteristic matching by the visual word of default classification
Classification image is effectively classified.
Step S204 believes the multi-dimensions histogram vector information, characteristic point vector information and visual word histogram vectors
Breath generates target feature vector.
In the present embodiment, it will be combined utilization by CEDD algorithms, SURF algorithm and BOVW models, fully extraction waits for
Characteristic information in classification image, to more be refined to the feature in image to be classified from global characteristics to local feature
Characteristic matching automatically classifies to image to be classified to realize, improves the accuracy of image classification.
Further, the step S201, specifically includes:
Step S205 extracts the face of global area in the image to be classified according to the directionality descriptor of color and edge
The colouring information and texture information are combined and generate multi-dimensions histogram vector information by color information and texture information.
In the concrete realization, the colouring information in image to be classified is extracted first, in order to realize in image to be classified
The extraction of colouring information generates additive mixture model (Red Green Blue, RGB), is tone by the RGB model conversions
Saturation degree lightness model (Hue Saturation Value, HSV), by 10 Interval Fuzzy filters by the HSV models
Information generates 10 fuzzy histogram informations, each color area that 10 Interval Fuzzy filters export then is passed through 24 section moulds
Paste filter is further divided into 3 H values regions, inputs 10 dimensional vectors and S, V channel value, exports 24 dimensional vectors.
Image to be classified is sent into 10 Interval Fuzzy filters, institute by the structure diagram of CEDD operation principles as shown in Figure 5
It states and generates 10 Interval Fuzzy filters according to 10 dimension histogram informations of the colouring information of image to be classified generation, then will be described
10 dimension histogram informations are sent into 24 Interval Fuzzy filters and generate 24 dimension histogram informations, to realize the face to image to be classified
Color identifies, and the color characteristic information of color extraction image to be classified that finer basis identifies.
In order to realize the identification and extraction to textural characteristics, in the present embodiment, the CEDD algorithms also pass through digital mistake
Texture marginal information in image to be classified described in filter rapid extraction, and according to edge histogram descriptor (Edge
Histogram Descriptor, EHD) one 6 dimension histogram of extraction, it is known that, CEDD histogram informations are made of 6 regions,
It is exactly the 6 dimensional vector histograms taken out by six regions that texture information extracts, then by texture information per one-dimensional addition
24 dimension colouring informations in color extraction, and colouring information and texture information are combined, obtain 144 dimension histogram informations.
Continue as shown in Figure 5 being filtered in image to be classified feeding digital filter to extract in image to be classified
Texture information, then by the texture information by edge histogram carry out judge generate 6 dimension histogram informations, described 6 are tieed up
Histogram information is quantified with the 24 dimension histogram informations extracted, CEDD features is generated, to realize to image to be classified
Global characteristics extraction.
Further, the step S202, specifically includes:
Step S206 presets characteristic point information according to accelerating robust feature algorithm to be generated according to default matrix, and according to institute
It states default characteristic point information and generates characteristic point vector information.
In the concrete realization, all characteristic points are generated by Hessian matrix first, is used for feature extraction, then passes through
The multilayer of the octave group of division, which forms, builds dimensional space, and the size of image is consistent between different groups, by being used not between different groups
Disconnected increased box filter, then each pixel that Hessian matrix is handled and 26 in two-dimensional space and dimensional space field
A point is compared, and orients initial key point, and pass through the weaker key point of filter energy and the mistake of location of mistake
Point filters out the characteristic point of final stabilization, and the characteristic point is carried out to the Ha Er in the border circular areas that statistics acquisition counts
Wavelet character, in the circle shaped neighborhood region of characteristic point, horizontal, the vertical Haar wavelet transform feature for counting all the points in 60 degree of sectors is total
With, it is then fan-shaped to be rotated and counted again in the region after Haar wavelet transform characteristic value at certain intervals, will finally it be worth most
Principal direction of the fan-shaped direction of that big as this feature point.
Take the rectangular area block of a 4*4 around characteristic point, the rectangular area direction along characteristic point principal direction,
The Haar wavelet transform feature horizontally and vertically of 25 pixels is counted per sub-regions, which is characterized as level
4 direction value, vertical direction value, horizontal direction absolute value and vertical direction absolute value directions, using this 4 values as per height
The feature vector in block region, so description vectors characteristic information of the shared 4*4*4=64 dimensional vectors as Surf features.
Further, the step S203, specifically includes:
Step S207 extracts the characteristic information of the image to be classified regional area according to vision bag of words, passes through institute
State characteristic information visual word histogram vectors information corresponding with the mapping relations of the default visual word generation characteristic information.
In the specific implementation, then the feature extracted first in image to be classified is described according to data set selected characteristic, formed special
Levy data;Again by learning bag of words, all merged using the characteristic handled well, then Feature Words are divided into the method for cluster
If Ganlei, each class is equivalent to a visual word, finally utilizes vision bag of words quantized image feature, straight to form visual word
Square figure vector information.
Scheme provided in this embodiment obtains color and the relevant global characteristics information of texture, then by CEDD algorithms
By SURF algorithm obtain preset characteristic point 64 dimensional feature vector information, then by BOVW models obtain visual word histogram to
Information is measured, to by combining the multi-direction characteristic information extracted to greatest extent in image to be classified, improve image recognition
Accuracy.
Further, as shown in fig. 6, proposing image classification method of the present invention the based on first embodiment and second embodiment
Three embodiments are illustrated based on first embodiment in the present embodiment, and before the step S30, the method further includes:
Step S301 according to the directionality descriptor at color and edge, accelerates robust feature algorithm and vision bag of words
The vector information in sample image data is extracted, and index vector data are generated according to the vector information.
In the concrete realization, in order to make image to be classified carry out Accurate classification, first of all, it is necessary to establish a variety of sorted pre-
If sample library, and the image data that completion of having classified will be preserved in the default sample library, then according to the image of classification completion
Data carry out feature vector matching with image to be classified, thus according to matching result, it will corresponding point of most matched image data
Target classification type of the Class Type as image to be classified.
It should be noted that in order to carry out consistent Data Matching, before matching, can to classified image data according to
CEDD algorithms, SURF algorithm and BOVW model extractions go out the vector data information for image data of having classified, and according to the vector
Data information establishes index, so as to faster find corresponding classification type according to matched image data.
The image data for completion of having classified is passed through CEDD algorithms, SURF algorithm and BOVW by scheme provided in this embodiment
Model extraction outgoing vector data information, and indexed by vector data information foundation, so as to faster pass through rope
Draw and find the corresponding classification type of the highest image data of matching degree, improves the classification effectiveness to image to be classified.
Further, as shown in fig. 7, proposing image classification method of the present invention the based on first embodiment and second embodiment
Four embodiments are illustrated based on first embodiment in the present embodiment, and after the step S20, the method further includes:
Step S208 matches described eigenvector with sorted index vector data in default sample library,
Matching result is that the image to be classified is set as default classification type when matching unsuccessful.
It should be noted that the default classification type can prove for the classification type for completion of having classified, such as identity card
Equal classification types can be also classification type undetermined, for storing unfiled successful image to be classified.
In the concrete realization, the image to be classified that also can extract non-successful match, according to the characteristic information of image to be classified
Sorted index vector data in sample library are preset in update, to keep the index in default sample library more rich, are met more
The classification of different image datas.
Further, after the step S30, the method further includes:
Step S40 updates the default sample library according to the target classification type of the image to be classified, and receives inquiry
It is instructed with modification, the image in updated default sample library is inquired and changed.
It should be noted that the inquiry and modification instruction can be user carries out interface operation execution by mobile terminal
Instruction, for example, the click commands that user is executed by the inquiry shown in serve port or modification button, can also pass through its other party
The inquiry or modification instruction, the present embodiment that formula is sent are not restricted this.
The image to be classified is set as by scheme provided in this embodiment for matching unsuccessful image to be classified
Default classification type can also classify hence for unsuccessful image to be classified is matched, and improve server for image point
The integrality of class, while inquiry being provided and changes interface and is effectively managed presetting the sample image in sample library.
It should be noted that herein, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that process, method, article or system including a series of elements include not only those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or system institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including this
There is also other identical elements in the process of element, method, article or system.
The embodiments of the present invention are for illustration only, can not represent the quality of embodiment.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical scheme of the present invention substantially in other words does the prior art
Going out the part of contribution can be expressed in the form of software products, which is stored in one as described above
In computer readable storage medium (such as ROM/RAM, magnetic disc, CD), including some instructions are used so that a station terminal equipment (can
To be mobile phone, computer, server, air conditioner or the network equipment etc.) execute method described in each embodiment of the present invention.
It these are only the preferred embodiment of the present invention, be not intended to limit the scope of the invention, it is every to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of image classification method, which is characterized in that described image sorting technique includes the following steps:
Obtain image to be classified;
The characteristic information of the image to be classified predeterminable area is extracted, and the characteristic information is generated into target feature vector;
The target feature vector is matched with sorted index vector data in default sample library, determines successful match
The corresponding default classification type of index vector data, and using the default classification type as the target of the image to be classified
Classification type.
2. image classification method as described in claim 1, which is characterized in that the extraction image to be classified predeterminable area
Characteristic information, and by the characteristic information generate target feature vector, specifically include:
The characteristic information of the image to be classified global area is extracted according to the directionality descriptor of color and edge, generates multidimensional
Histogram vectors information;According to the default characteristic point letter for accelerating robust feature algorithm to extract the image to be classified regional area
Breath, and characteristic point vector information is generated according to the default characteristic point information;It is extracted according to vision bag of words described to be sorted
The characteristic information of image local area, and the corresponding visual word histogram vectors of the characteristic information are generated according to default visual word
Information;
The multi-dimensions histogram vector information, characteristic point vector information and visual word histogram vectors information are generated into target signature
Vector.
3. image classification method as claimed in claim 2, which is characterized in that described to be described according to color and the directionality at edge
Symbol extracts the characteristic information of the image to be classified global area, generates multi-dimensions histogram vector information, specifically includes:
The colouring information and texture of global area in the image to be classified are extracted according to the directionality descriptor of color and edge
The colouring information and texture information are combined and generate multi-dimensions histogram vector information by information.
4. image classification method as claimed in claim 2, which is characterized in that described to extract institute according to acceleration robust feature algorithm
The default characteristic point information of image to be classified regional area is stated, and characteristic point vector letter is generated according to the default characteristic point information
Breath, specifically includes:
Characteristic point information is preset according to accelerating robust feature algorithm to be generated according to default matrix, and is believed according to the default characteristic point
Breath generates characteristic point vector information.
5. image classification method as claimed in claim 2, which is characterized in that described to be waited for according to described in the extraction of vision bag of words
The characteristic information of classification image local area, and the corresponding visual word histogram of the characteristic information is generated according to default visual word
Vector information specifically includes:
The characteristic information that the image to be classified regional area is extracted according to vision bag of words, by the characteristic information and in advance
If the mapping relations of visual word generate the corresponding visual word histogram vectors information of the characteristic information.
6. the image classification method as described in any one of claim 1 to 5, which is characterized in that described by the target signature
It is vectorial to be matched with sorted index vector data in default sample library, determine that the index vector data of successful match correspond to
Default classification type, and using the default classification type as the target classification type of the image to be classified before, it is described
Method further includes:
According to the directionality descriptor at color and edge, robust feature algorithm and vision bag of words is accelerated to extract sample image number
Vector information in, and index vector data are generated according to the vector information.
7. the image classification method as described in any one of claim 1 to 5, which is characterized in that the extraction is described to be sorted
The characteristic information of image predeterminable area, and by after characteristic information generation target feature vector, the method further includes:
Described eigenvector is matched with sorted index vector data in default sample library, is matching in matching result
When unsuccessful, the image to be classified is set as default classification type.
8. the image classification method as described in any one of claim 1 to 5, which is characterized in that described by the target signature
It is vectorial to be matched with sorted index vector data in default sample library, determine that the index vector data of successful match correspond to
Default classification type, and using the default classification type as the target classification type of the image to be classified after, it is described
Method further includes:
The default sample library is updated according to the target classification type of the image to be classified, and receives inquiry and modification instruction,
Image in updated default sample library is inquired and changed.
9. a kind of server, which is characterized in that the server includes:It memory, processor and is stored on the memory
And the image classification program that can be run on the processor, described image sort program are arranged for carrying out such as claim 1 to 8
Any one of described in image classification method the step of.
10. a kind of storage medium, which is characterized in that be stored with image classification program, described image classification on the storage medium
It is realized such as the step of image classification method described in any item of the claim 1 to 8 when program is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810071179.8A CN108280190A (en) | 2018-01-24 | 2018-01-24 | Image classification method, server and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810071179.8A CN108280190A (en) | 2018-01-24 | 2018-01-24 | Image classification method, server and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108280190A true CN108280190A (en) | 2018-07-13 |
Family
ID=62804942
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810071179.8A Pending CN108280190A (en) | 2018-01-24 | 2018-01-24 | Image classification method, server and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108280190A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109558823A (en) * | 2018-11-22 | 2019-04-02 | 北京市首都公路发展集团有限公司 | A kind of vehicle identification method and system to scheme to search figure |
CN110516712A (en) * | 2019-08-01 | 2019-11-29 | 仲恺农业工程学院 | Insect pest image-recognizing method, insect pest monitoring method, device, equipment and medium |
CN111046933A (en) * | 2019-12-03 | 2020-04-21 | 东软集团股份有限公司 | Image classification method and device, storage medium and electronic equipment |
CN111061898A (en) * | 2019-12-13 | 2020-04-24 | Oppo(重庆)智能科技有限公司 | Image processing method, image processing device, computer equipment and storage medium |
WO2020186777A1 (en) * | 2019-03-16 | 2020-09-24 | 平安科技(深圳)有限公司 | Image retrieval method, apparatus and device, and computer-readable storage medium |
CN112215225A (en) * | 2020-10-22 | 2021-01-12 | 北京通付盾人工智能技术有限公司 | KYC certificate verification method based on computer vision technology |
CN112484242A (en) * | 2020-11-30 | 2021-03-12 | 珠海格力电器股份有限公司 | Air conditioner type selection method based on augmented reality and related device |
CN110097082B (en) * | 2019-03-29 | 2021-08-27 | 广州思德医疗科技有限公司 | Splitting method and device of training set |
CN113688268A (en) * | 2021-08-31 | 2021-11-23 | 中国平安人寿保险股份有限公司 | Picture information extraction method and device, computer equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102708373A (en) * | 2012-01-06 | 2012-10-03 | 香港理工大学 | Method and device for classifying remote images by integrating space information and spectral information |
CN103927387A (en) * | 2014-04-30 | 2014-07-16 | 成都理想境界科技有限公司 | Image retrieval system, method and device |
CN105488527A (en) * | 2015-11-27 | 2016-04-13 | 小米科技有限责任公司 | Image classification method and apparatus |
CN106557728A (en) * | 2015-09-30 | 2017-04-05 | 佳能株式会社 | Query image processing and image search method and device and surveillance |
CN106919920A (en) * | 2017-03-06 | 2017-07-04 | 重庆邮电大学 | Scene recognition method based on convolution feature and spatial vision bag of words |
-
2018
- 2018-01-24 CN CN201810071179.8A patent/CN108280190A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102708373A (en) * | 2012-01-06 | 2012-10-03 | 香港理工大学 | Method and device for classifying remote images by integrating space information and spectral information |
CN103927387A (en) * | 2014-04-30 | 2014-07-16 | 成都理想境界科技有限公司 | Image retrieval system, method and device |
CN106557728A (en) * | 2015-09-30 | 2017-04-05 | 佳能株式会社 | Query image processing and image search method and device and surveillance |
CN105488527A (en) * | 2015-11-27 | 2016-04-13 | 小米科技有限责任公司 | Image classification method and apparatus |
CN106919920A (en) * | 2017-03-06 | 2017-07-04 | 重庆邮电大学 | Scene recognition method based on convolution feature and spatial vision bag of words |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109558823A (en) * | 2018-11-22 | 2019-04-02 | 北京市首都公路发展集团有限公司 | A kind of vehicle identification method and system to scheme to search figure |
WO2020186777A1 (en) * | 2019-03-16 | 2020-09-24 | 平安科技(深圳)有限公司 | Image retrieval method, apparatus and device, and computer-readable storage medium |
CN110097082B (en) * | 2019-03-29 | 2021-08-27 | 广州思德医疗科技有限公司 | Splitting method and device of training set |
CN110516712A (en) * | 2019-08-01 | 2019-11-29 | 仲恺农业工程学院 | Insect pest image-recognizing method, insect pest monitoring method, device, equipment and medium |
CN110516712B (en) * | 2019-08-01 | 2023-04-07 | 仲恺农业工程学院 | Insect pest image recognition method, insect pest monitoring method, insect pest image recognition device, insect pest monitoring equipment and insect pest image recognition medium |
CN111046933A (en) * | 2019-12-03 | 2020-04-21 | 东软集团股份有限公司 | Image classification method and device, storage medium and electronic equipment |
CN111046933B (en) * | 2019-12-03 | 2024-03-05 | 东软集团股份有限公司 | Image classification method, device, storage medium and electronic equipment |
CN111061898A (en) * | 2019-12-13 | 2020-04-24 | Oppo(重庆)智能科技有限公司 | Image processing method, image processing device, computer equipment and storage medium |
CN112215225A (en) * | 2020-10-22 | 2021-01-12 | 北京通付盾人工智能技术有限公司 | KYC certificate verification method based on computer vision technology |
CN112215225B (en) * | 2020-10-22 | 2024-03-15 | 北京通付盾人工智能技术有限公司 | KYC certificate verification method based on computer vision technology |
CN112484242A (en) * | 2020-11-30 | 2021-03-12 | 珠海格力电器股份有限公司 | Air conditioner type selection method based on augmented reality and related device |
CN113688268A (en) * | 2021-08-31 | 2021-11-23 | 中国平安人寿保险股份有限公司 | Picture information extraction method and device, computer equipment and storage medium |
CN113688268B (en) * | 2021-08-31 | 2024-04-02 | 中国平安人寿保险股份有限公司 | Picture information extraction method, device, computer equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108280190A (en) | Image classification method, server and storage medium | |
Zhou et al. | Edge-based structural features for content-based image retrieval | |
US8750573B2 (en) | Hand gesture detection | |
CN105144239B (en) | Image processing apparatus, image processing method | |
CN108090511B (en) | Image classification method and device, electronic equipment and readable storage medium | |
US20120027252A1 (en) | Hand gesture detection | |
US8379990B2 (en) | Object recognition apparatus, computer readable medium storing object recognition program, and image retrieval service providing method | |
CN110717366A (en) | Text information identification method, device, equipment and storage medium | |
CN109409377B (en) | Method and device for detecting characters in image | |
JP2004361987A (en) | Image retrieval system, image classification system, image retrieval program, image classification program, image retrieval method, and image classification method | |
CN103353881B (en) | Method and device for searching application | |
CN107633205A (en) | lip motion analysis method, device and storage medium | |
CN109284613B (en) | Method, device, equipment and storage medium for identification detection and counterfeit site detection | |
Yu et al. | A rule based technique for extraction of visual attention regions based on real-time clustering | |
Reta et al. | Color uniformity descriptor: An efficient contextual color representation for image indexing and retrieval | |
CN106446889B (en) | A kind of local recognition methods of logo and device | |
CN112883827B (en) | Method and device for identifying specified target in image, electronic equipment and storage medium | |
CN106682670A (en) | Method and system for identifying station caption | |
Silva et al. | A SOM combined with KNN for classification task | |
JP4967045B2 (en) | Background discriminating apparatus, method and program | |
CN112926601A (en) | Image recognition method, device and equipment based on deep learning and storage medium | |
CN106056575B (en) | A kind of image matching method based on like physical property proposed algorithm | |
CN108174091B (en) | Image processing method, image processing device, storage medium and electronic equipment | |
Murshed et al. | Enhanced colour image retrieval with cuboid segmentation | |
CN109657083B (en) | Method and device for establishing textile picture feature library |
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: 20180713 |
|
RJ01 | Rejection of invention patent application after publication |