CN106529546B - A kind of image-recognizing method and device - Google Patents
A kind of image-recognizing method and device Download PDFInfo
- Publication number
- CN106529546B CN106529546B CN201610881419.1A CN201610881419A CN106529546B CN 106529546 B CN106529546 B CN 106529546B CN 201610881419 A CN201610881419 A CN 201610881419A CN 106529546 B CN106529546 B CN 106529546B
- Authority
- CN
- China
- Prior art keywords
- sub
- combination
- multiple features
- regions
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of image-recognizing method and device, method includes:Obtain the multizone combination of multiple features vector of two images;Multizone combination of multiple features vector is obtained by being divided into M sub-regions by two images and extracting N number of feature vector respectively to M sub-regions;Determine the Hausdorff distances of the subregion combination of multiple features vector between each corresponding sub-region in M sub-regions;The Hausdorff distances of subregion combination of multiple features vector between each corresponding sub-region in two images M sub-regions, to determine the Hausdorff distances of multizone combination of multiple features vector;The Hausdorff distance values of multizone combination of multiple features vector are bigger, then show that two images are more different.The present invention can be used both independently of each character pair in image, the multizone that is merged without using kernel function, multi-feature recognition method, while reduce operand, improve recognition efficiency.
Description
Technical field
The present invention relates to field of image recognition, more particularly to a kind of image-recognizing method and device.
Background technology
With the development of image processing techniques, more and more fields begin to use image processing techniques, for example, in industry
Field begins to use image identification industrial component instead of method of former manual identified industrial component etc..
Conventional images identification technology is the existing recognition methods based on multizone, multiple features mostly, it usually needs composition
The higher feature vector of dimension, then be trained and predict, therefore composition of vector dimension is higher, needs to select suitable kernel function
It is trained and predicts;And the method for feature based distance of the prior art is led in the use of multiple features and multizone
Normal robustness is poor, and is strongly depend on the selection of feature and its distance, and computation complexity is higher.
The content of the invention
The embodiment of the present invention provides a kind of image-recognizing method and device, and character pair in image can be made to make independently of each other
With, the multizone that is merged without using kernel function, multi-feature recognition method, while operand can be reduced, improve identification
Efficiency.
In a first aspect, the present invention provides a kind of image-recognizing method, including:
Obtain the multizone combination of multiple features vector of two width images to be recognized;The multizone combination of multiple features vector passes through
The images to be recognized is divided into M sub-regions and extracts N number of feature vector respectively to the M sub-regions and is obtained;
Determine the subregion multiple features between each corresponding sub-region in the M sub-regions of the two width images to be recognized
The Hausdorff distances of mix vector;
The subregion multiple features group between each corresponding sub-region in the two width images to be recognized M sub-regions
The Hausdorff distances of resultant vector, to determine the Hausdorff distances of multizone combination of multiple features vector;
If the Hausdorff distance values of the multizone combination of multiple features vector are bigger, show described two it is to be identified
Image is more different.
Second aspect, the present invention provides a kind of pattern recognition device, including:
Acquisition module, for obtaining the multizone combination of multiple features of two width images to be recognized vector;How special the multizone is
Sign mix vector extracts by the way that the images to be recognized is divided into M sub-regions and respectively N number of feature to the M sub-regions
Vector obtains;
First determining module, for determining each corresponding sub-region in the M sub-regions of the two width images to be recognized
Between subregion combination of multiple features vector Hausdorff distances;
Second determining module, for each corresponding sub-region in the two width images to be recognized M sub-regions it
Between subregion combination of multiple features vector Hausdorff distances, with determine multizone combination of multiple features vector Hausdorff
Distance;
If the Hausdorff distance values of the multizone combination of multiple features vector are bigger, show described two it is to be identified
Image is more different.
As can be seen that in the scheme of the embodiment of the present invention, first, by the way that the two width images to be recognized is divided into M
Sub-regions simultaneously extract N number of feature vector respectively to the M sub-regions, how special to obtain the multizone of two width images to be recognized
Levy mix vector;Secondly, the subregion combination of multiple features vector between each corresponding region in M sub-regions is determined
Hausdorff distances;The finally subregion between each corresponding region in the two width images to be recognized M sub-regions
The Hausdorff distances of combination of multiple features vector, to determine the Hausdorff distances of multizone combination of multiple features vector;If institute
It is bigger to state the Hausdorff distance values of multizone combination of multiple features vector, then shows that the two width images to be recognized is more different.It can
See, compared to the prior art, this programme is then obtained by dividing region in the same manner respectively to two width images to be recognized
The Hausdorff distances of corresponding sub-region combination of multiple features vector are taken, the group of the final multizone multiple features for obtaining described image
The Hausdorff distances of resultant vector, the similarity of this two images is judged according to distance value size, and this method can make figure
Character pair is used both independently of each as in, the multizone that is merged without using kernel function, multi-feature recognition method, while can
To reduce operand, the recognition efficiency during image similarity degree is identified is improved.
Description of the drawings
It in order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention, for those of ordinary skill in the art, without creative efforts, can be with
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is a kind of image-recognizing method flow diagram disclosed by the embodiments of the present invention;
Fig. 2 is a kind of multizone multi-feature extraction schematic diagram disclosed by the embodiments of the present invention;
Fig. 3 is another image-recognizing method flow diagram disclosed by the embodiments of the present invention;
Fig. 4 is a kind of pattern recognition device schematic diagram disclosed by the embodiments of the present invention.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present invention, the technical solution in the embodiment of the present invention is carried out clear, complete
Site preparation describes, it is clear that described embodiment is only part of the embodiment of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained every other without creative efforts
Embodiment belongs to the scope of protection of the invention.
It should be noted that the term used in embodiments of the present invention is only merely for the mesh of description specific embodiment
, it is not intended to limit the invention." the one of the embodiment of the present invention and singulative used in the attached claims
Kind ", " described " and "the" are also intended to including most forms, unless context clearly shows that other meanings.It is also understood that this
Term "and/or" used herein refers to and comprising one or more associated any or all possible group of list items purpose
It closes.
Fig. 1 is referred to, Fig. 1 is a kind of image-recognizing method flow diagram disclosed by the embodiments of the present invention.Such as Fig. 1 institutes
Show, a kind of image-recognizing method disclosed by the embodiments of the present invention comprises the following steps:
101st, pattern recognition device obtains the multizone combination of multiple features vector of two width images to be recognized;The multizone is more
Combinations of features vector extracts by the way that the images to be recognized is divided into M sub-regions and respectively N number of spy to the M sub-regions
Sign vector obtains.
Wherein, images to be recognized refers to the target image that needs identify, can be video camera institute the image collected, at this
In inventive embodiments, which needs for coloured image, can be the forms such as bmp or jpeg, can support CMYK or
The color modes such as RGB.
Optionally, which can be all targeted color images for needing to identify, for example, image of clothing, family
Have image, character image etc..
Wherein, above-mentioned two width images to be recognized is respectively divided into M sub-regions by above-mentioned pattern recognition device, can will be upper
The multizone combination of multiple features vector for stating two images is expressed as:
P={ P1,...,PM, Q={ Q1,...,QM}
Wherein, PM, QMThe m-th subregion combinations of features vector in above-mentioned images to be recognized is represented respectively.
Every sub-regions in above-mentioned M region include the feature of N number of type, it is above-mentioned per sub-regions combination of multiple features to
Amount is represented by:
Pm={ pm,1,...,pm,N, Qm={ qm,1,...,qm,N}
Wherein, wherein m is more than or equal to 1 and less than or equal to M.pm,N, qm,NAbove-mentioned figure to be identified is represented respectively
The feature vector of the n-th type feature in m-th of subregion combinations of features vector as in.
Position and size of every sub-regions in above-mentioned two width images to be recognized in above-mentioned M sub-regions is corresponding
's.For example, as shown in Fig. 2, Fig. 2 is a kind of multizone disclosed by the embodiments of the present invention the more feature extraction schematic diagram.Such as
Shown in figure, two width images to be recognized are respectively designated as figure P and figure Q, in P is schemed, subregion P8Among the third line of figure P
Position, and include the feature of N number of type, i.e. subregion P8Combination of multiple features vector includes N number of feature vector;In Q is schemed,
Subregion Q8Position among the third line of figure Q, and include the feature of N number of type, i.e. subregion Q8Combination of multiple features to
Amount includes N number of feature vector.
102nd, described image identification device determines the subregion feature group between each corresponding sub-region in M sub-regions
The Hausdorff distances of resultant vector.
Wherein, the son between each corresponding sub-region in the M sub-regions for determining above-mentioned two width images to be recognized
The Hausdorff distances of provincial characteristics mix vector, including:
Pass through first function dH(Pm,Qm) determine between each corresponding sub-region subregion combinations of features vector
Hausdorff distances, the dH(Pm,Qm) function representation is:
dH(Pm,Qm)=max { hk(Pm,Qm),hk(Qm,Pm)}
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, dH(Pm,Qm) it is Hausdorff distances between m-th of subregion in above-mentioned two width images to be recognized.
Wherein hk(Pm,Qm) the Hausdorff distances that represent k ranks be all element distances in P and Q | | p-q | | middle kth
Small value.
Wherein, by reasonably selecting k values, the representativeness of the distance value improved and the robustness of match cognization.It is optional
Ground, k values are less than or equal to the quantity setting other values of 4 or the characteristic type in every sub-regions.
103rd, pair of every sub-regions of the described image identification device in the two width images to be recognized M sub-regions
The Hausdorff distances of the subregion combination of multiple features vector between subregion are answered, to determine multizone combination of multiple features vector
Hausdorff distances.
Wherein, the sub-district between each corresponding sub-region in above-mentioned two width images to be recognized M sub-regions
The Hausdorff distances of domain combination of multiple features vector, to obtain the Hausdorff distances of multizone combination of multiple features vector, bag
It includes:
Pass through second function DH(Pm,Qm) connect the Hausdorff distances of above-mentioned subregion combination of multiple features vector
Come, to determine the Hausdorff distances of above-mentioned multizone combination of multiple features vector;
Wherein, above-mentioned second function DH(Pm,Qm) be expressed as:
DH(Pm,Qm)=f (dH(P1,Q1),...,dH(PM,QM))
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, above-mentioned second function DH(Pm,Qm) it is to every sub-regions combination of multiple features vector in M sub-regions
The linear weighted function summing function of Hausdorff distances.
Wherein, the Hausdorff distances of above-mentioned subregion combination of multiple features vector are connected above by second function
Come, with determine above-mentioned multizone combination of multiple features vector Hausdorff distances may particularly denote for:
DH(Pm,Qm)=f (dH(P1,Q1),...,dH(PM,QM))=a1dH(P1,Q1)+a2dH(P2,Q2)+,...,+aMdH
(PM,QM)
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, a1, a2... aMFor weighting coefficient, every sub-regions feature in above-mentioned M sub-regions is in above-mentioned figure
The size of weighting coefficient is set as the weight in identification process.If a certain sub-regions feature in above-mentioned M sub-regions exists
Weight is big in above-mentioned image recognition processes, then increases corresponding weighting coefficient;It is on the contrary then reduce corresponding weighting coefficient.
Optionally, above-mentioned second function DH(Pm,Qm) can also be every sub-regions spy in above-mentioned M sub-regions
The weight in above-mentioned image recognition processes is levied to set the summing function of corresponding weighting coefficient.
104th, described image identification device is determined according to the Hausdorff distance values of the multizone combination of multiple features vector
The similitude of the two width images to be recognized.
Above-mentioned pattern recognition device is according to the Hausdorff distances D of above-mentioned multizone combination of multiple features vectorH(Pm,Qm)
Size determines the similarity of above-mentioned two width images to be recognized;If the Hausdorff of above-mentioned multizone combination of multiple features vector away from
From DH(Pm,Qm) value it is smaller, then it represents that the similarity of above-mentioned two width images to be recognized is bigger;It is on the contrary then represent that above-mentioned two width is treated
Identify that the similarity of image is smaller.
As can be seen that in the scheme of the embodiment of the present invention, first, by the way that the two width images to be recognized is divided into M
Sub-regions simultaneously extract N number of feature vector respectively to the M sub-regions, how special to obtain the multizone of two width images to be recognized
Levy mix vector;Secondly, the subregion combination of multiple features vector between each corresponding region in M sub-regions is determined
Hausdorff distances;The finally subregion between each corresponding region in the two width images to be recognized M sub-regions
The Hausdorff distances of combination of multiple features vector, to determine the Hausdorff distances of multizone combination of multiple features vector;If institute
It is bigger to state the Hausdorff distance values of multizone combination of multiple features vector, then shows that the two width images to be recognized is more different.It can
See, compared to the prior art, this programme is then obtained by dividing region in the same manner respectively to two width images to be recognized
The Hausdorff distances of corresponding sub-region combination of multiple features vector are taken, the group of the final multizone multiple features for obtaining described image
The Hausdorff distances of resultant vector, the similarity of this two images is judged according to distance value size, and this method can make figure
Character pair is used both independently of each as in, the multizone that is merged without using kernel function, multi-feature recognition method, while can
To reduce operand, the recognition efficiency during image similarity degree is identified is improved.
Fig. 3 is referred to, Fig. 3 is another image-recognizing method flow diagram disclosed by the embodiments of the present invention.Such as Fig. 3 institutes
Show, a kind of image-recognizing method disclosed by the embodiments of the present invention comprises the following steps:
301st, described image identification device obtains the multizone combination of multiple features vector of two images to be matched;The multi-region
Domain combination of multiple features vector extracts by the way that the image to be matched is divided into M sub-regions and respectively N to the M sub-regions
A feature vector obtains.
Wherein, above-mentioned two width images to be recognized is respectively divided into M sub-regions by above-mentioned pattern recognition device, can will be upper
The multizone combination of multiple features vector for stating two images is expressed as:
P={ P1,...,PM, Q={ Q1,...,QM}
Wherein, PM, QMThe m-th subregion combinations of features vector in above-mentioned images to be recognized is represented respectively.
Every sub-regions in above-mentioned M region include the feature of N number of type, it is above-mentioned per sub-regions combination of multiple features to
Amount is represented by:
Pm={ pm,1,...,pm,N, Qm={ qm,1,...,qm,N}
Wherein, wherein m is more than or equal to 1 and less than or equal to M.pm,N, qm,NAbove-mentioned figure to be identified is represented respectively
The feature vector of the n-th type feature in m-th of subregion combinations of features vector as in.
Position and size of every sub-regions in above-mentioned two width images to be recognized in above-mentioned M sub-regions is corresponding
's.For example, as shown in Fig. 2, Fig. 2 is a kind of multizone disclosed by the embodiments of the present invention the more feature extraction schematic diagram.Such as
Shown in figure, two width images to be recognized are respectively designated as figure P and figure Q, in P is schemed, subregion P8Among the third line of figure P
Position, and include the feature of N number of type, i.e. subregion P8Combination of multiple features vector includes N number of feature vector;In Q is schemed,
Subregion Q8Position among the third line of figure Q, and include the feature of N number of type, i.e. subregion Q8Combination of multiple features to
Amount includes N number of feature vector.
302nd, described image identification device is determined by first function in the M sub-regions of the two width images to be recognized
The Hausdorff distances of subregion combination of multiple features vector between each corresponding sub-region.
Wherein, above-mentioned first function is expressed as:
dH(Pm,Qm)=max { hk(Pm,Qm),hk(Qm,Pm)}
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, dH(Pm,Qm) it is Hausdorff distances between m-th of subregion in above-mentioned two width images to be recognized.
Wherein hk(Pm,Qm) the Hausdorff distances that represent k ranks be all element distances in P and Q | | p-q | | middle kth
Small value.
Wherein, by reasonably selecting k values, the representativeness of the distance value improved and the robustness of match cognization.It is optional
Ground, k values are less than or equal to the quantity setting other values of 4 or the characteristic type in every sub-regions.
303rd, described image identification device by second function by the two width images to be recognized M sub-regions multiple features groups
The Hausdorff distances of resultant vector connect, to obtain the Hausdorff distances of the multizone combination of multiple features vector.
Wherein, above-mentioned second function DH(Pm,Qm) be expressed as:
DH(Pm,Qm)=f (dH(P1,Q1),...,dH(PM,QM))
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, above-mentioned second function DH(Pm,Qm) it is to every sub-regions combination of multiple features vector in M sub-regions
The linear weighted function summing function of Hausdorff distances.
Wherein, the Hausdorff distances of above-mentioned subregion combination of multiple features vector are connected above by second function
Come, with determine above-mentioned multizone combination of multiple features vector Hausdorff distances may particularly denote for:
DH(Pm,Qm)=f (dH(P1,Q1),...,dH(PM,QM))=a1dH(P1,Q1)+a2dH(P2,Q2)+,...,+aMdH
(PM,QM)
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, a1, a2... aMFor weighting coefficient, every sub-regions feature in above-mentioned M sub-regions is in above-mentioned figure
The size of weighting coefficient is set as the weight in identification process.If a certain sub-regions feature in above-mentioned M sub-regions exists
Weight is big in above-mentioned image recognition processes, then increases corresponding weighting coefficient;It is on the contrary then reduce corresponding weighting coefficient.
Optionally, above-mentioned second function DH(Pm,Qm) can also be every sub-regions spy in above-mentioned M sub-regions
The weight in above-mentioned image recognition processes is levied to set the summing function of corresponding weighting coefficient.
304th, described image identification device is determined according to the Hausdorff distance values of the multizone combination of multiple features vector
The similitude of the two width images to be recognized.
Above-mentioned pattern recognition device is according to the Hausdorff distances D of above-mentioned multizone combination of multiple features vectorH(Pm,Qm)
Size determines the similarity of above-mentioned two width images to be recognized;If the Hausdorff of above-mentioned multizone combination of multiple features vector away from
From DH(Pm,Qm) value it is smaller, then it represents that the similarity of above-mentioned two width images to be recognized is bigger;It is on the contrary then represent that above-mentioned two width is treated
Identify that the similarity of image is smaller.
As can be seen that in the scheme of the embodiment of the present invention, first, by the way that the two width images to be recognized is divided into M
Sub-regions simultaneously extract N number of feature vector respectively to the M sub-regions, how special to obtain the multizone of two width images to be recognized
Levy mix vector;Secondly, the subregion combination of multiple features vector between each corresponding region in M sub-regions is determined
Hausdorff distances;The finally subregion between each corresponding region in the two width images to be recognized M sub-regions
The Hausdorff distances of combination of multiple features vector, to determine the Hausdorff distances of multizone combination of multiple features vector;If institute
It is bigger to state the Hausdorff distance values of multizone combination of multiple features vector, then shows that the two width images to be recognized is more different.It can
See, compared to the prior art, this programme is then obtained by dividing region in the same manner respectively to two width images to be recognized
The Hausdorff distances of corresponding sub-region combination of multiple features vector are taken, the group of the final multizone multiple features for obtaining described image
The Hausdorff distances of resultant vector, the similarity of this two images is judged according to distance value size, and this method can make figure
Character pair is used both independently of each as in, the multizone that is merged without using kernel function, multi-feature recognition method, while can
To reduce operand, the recognition efficiency during image similarity degree is identified is improved.
Fig. 4 is referred to, Fig. 4 is a kind of structure diagram of pattern recognition device 400 disclosed by the embodiments of the present invention.Such as figure
Shown in 4, a kind of pattern recognition device 400 disclosed by the embodiments of the present invention, including:
Acquisition module 401, for obtaining the multizone combination of multiple features of two width images to be recognized vector;The multizone is more
Combinations of features vector extracts by the way that the images to be recognized is divided into M sub-regions and respectively N number of spy to the M sub-regions
Sign vector obtains.
Wherein, above-mentioned two width images to be recognized is respectively divided into M sub-regions by above-mentioned pattern recognition device, can will be upper
The multizone combination of multiple features vector for stating two images is expressed as:
P={ P1,...,PM, Q={ Q1,...,QM}
Wherein, PM, QMThe m-th subregion combinations of features vector in above-mentioned images to be recognized is represented respectively.
Every sub-regions in above-mentioned M region include the feature of N number of type, it is above-mentioned per sub-regions combination of multiple features to
Amount is represented by:
Pm={ pm,1,...,pm,N, Qm={ qm,1,...,qm,N}
Wherein, wherein m is more than or equal to 1 and less than or equal to M.pm,N, qm,NAbove-mentioned figure to be identified is represented respectively
The feature vector of the n-th type feature in m-th of subregion combinations of features vector as in.
Position and size of every sub-regions in above-mentioned two width images to be recognized in above-mentioned M sub-regions is corresponding
's.For example, as shown in Fig. 2, Fig. 2 is a kind of multizone disclosed by the embodiments of the present invention the more feature extraction schematic diagram.Such as
Shown in figure, two width images to be recognized are respectively designated as figure P and figure Q, in P is schemed, subregion P8Among the third line of figure P
Position, and include the feature of N number of type, i.e. subregion P8Combination of multiple features vector includes N number of feature vector;In Q is schemed,
Subregion Q8Position among the third line of figure Q, and include the feature of N number of type, i.e. subregion Q8Combination of multiple features to
Amount includes N number of feature vector.
First determining module 402, for determining each corresponding sub-district in the M sub-regions of the two width images to be recognized
The Hausdorff distances of subregion combination of multiple features vector between domain.
Wherein, the son between each corresponding sub-region in the M sub-regions for determining above-mentioned two width images to be recognized
The Hausdorff distances of provincial characteristics mix vector, including:
Pass through first function dH(Pm,Qm) determine between each corresponding sub-region subregion combinations of features vector
Hausdorff distances, the dH(Pm,Qm) function representation is:
dH(Pm,Qm)=max { hk(Pm,Qm),hk(Qm,Pm)}
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, dH(Pm,Qm) it is Hausdorff distances between m-th of subregion in above-mentioned two width images to be recognized.
Wherein hk(Pm,Qm) the Hausdorff distances that represent k ranks be all element distances in P and Q | | p-q | | middle kth
Small value.
Wherein, by reasonably selecting k values, the representativeness of the distance value improved and the robustness of match cognization.It is optional
Ground, k values are less than or equal to the quantity setting other values of 4 or the characteristic type in every sub-regions.
Second determining module 403, for each corresponding sub-region in the two width images to be recognized M sub-regions
Between subregion combination of multiple features vector Hausdorff distances, with determine multizone combination of multiple features vector
Hausdorff distances.
Wherein, the sub-district between each corresponding sub-region in above-mentioned two width images to be recognized M sub-regions
The Hausdorff distances of domain combination of multiple features vector, to obtain the Hausdorff distances of multizone combination of multiple features vector, bag
It includes:
Pass through second function DH(Pm,Qm) connect the Hausdorff distances of above-mentioned subregion combination of multiple features vector
Come, to determine the Hausdorff distances of above-mentioned multizone combination of multiple features vector;
Wherein, above-mentioned second function DH(Pm,Qm) be expressed as:
DH(Pm,Qm)=f (dH(P1,Q1),...,dH(PM,QM))
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, above-mentioned second function DH(Pm,Qm) it is to every sub-regions combination of multiple features vector in M sub-regions
The linear weighted function summing function of Hausdorff distances.
Wherein, the Hausdorff distances of above-mentioned subregion combination of multiple features vector are connected above by second function
Come, with determine above-mentioned multizone combination of multiple features vector Hausdorff distances may particularly denote for:
DH(Pm,Qm)=f (dH(P1,Q1),...,dH(PM,QM))=a1dH(P1,Q1)+a2dH(P2,Q2)+,...,+aMdH
(PM,QM)
Wherein, m is more than or equal to 1 and less than or equal to M.
Wherein, a1, a2... aMFor weighting coefficient, every sub-regions feature in above-mentioned M sub-regions is in above-mentioned figure
The size of weighting coefficient is set as the weight in identification process.If a certain sub-regions feature in above-mentioned M sub-regions exists
Weight is big in above-mentioned image recognition processes, then increases corresponding weighting coefficient;It is on the contrary then reduce corresponding weighting coefficient.
Optionally, above-mentioned second function DH(Pm,Qm) can also be every sub-regions spy in above-mentioned M sub-regions
The weight in above-mentioned image recognition processes is levied to set the summing function of corresponding weighting coefficient.
As can be seen that in the scheme of the embodiment of the present invention, first, by the way that the two width images to be recognized is divided into M
Sub-regions simultaneously extract N number of feature vector respectively to the M sub-regions, how special to obtain the multizone of two width images to be recognized
Levy mix vector;Secondly, the subregion combination of multiple features vector between each corresponding region in M sub-regions is determined
Hausdorff distances;The finally subregion between each corresponding region in the two width images to be recognized M sub-regions
The Hausdorff distances of combination of multiple features vector, to determine the Hausdorff distances of multizone combination of multiple features vector;If institute
It is bigger to state the Hausdorff distance values of multizone combination of multiple features vector, then shows that the two width images to be recognized is more different.It can
See, compared to the prior art, this programme is then obtained by dividing region in the same manner respectively to two width images to be recognized
The Hausdorff distances of corresponding sub-region combination of multiple features vector are taken, the group of the final multizone multiple features for obtaining described image
The Hausdorff distances of resultant vector, the similarity of this two images is judged according to distance value size, and this method can make figure
Character pair is used both independently of each as in, the multizone that is merged without using kernel function, multi-feature recognition method, while can
To reduce operand, the recognition efficiency during image similarity degree is identified is improved.
One of ordinary skill in the art will appreciate that realizing all or part of flow in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a computer read/write memory medium
In, the program is upon execution, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
The above disclosed power for being only a kind of preferred embodiment of the present invention, the present invention cannot being limited with this certainly
Sharp scope one of ordinary skill in the art will appreciate that realizing all or part of flow of above-described embodiment, and is weighed according to the present invention
Profit requires made equivalent variations, still falls within and invents covered scope.
Claims (8)
1. a kind of image-recognizing method, which is characterized in that including:
Obtain the multizone combination of multiple features vector of two width images to be recognized;The multizone combination of multiple features vector is by by institute
It states images to be recognized and is divided into M sub-regions and N number of feature vector is extracted respectively to the M sub-regions and obtain, it is described to wait to know
Other image is coloured image and is image of clothing, and the corresponding feature in each region is used both independently of each in the images to be recognized,
The feature of N number of type is included in the M sub-regions per sub-regions;
Determine the subregion combination of multiple features between each corresponding sub-region in the M sub-regions of the two width images to be recognized
The Hausdorff distances of vector;
The subregion combination of multiple features between each corresponding sub-region in the two width images to be recognized M sub-regions to
The Hausdorff distances of amount, to determine the Hausdorff distances of multizone combination of multiple features vector;
The similar of the two width images to be recognized is determined according to the Hausdorff distance values of the multizone combination of multiple features vector
Property.
2. the according to the method described in claim 1, it is characterized in that, multizone combination of multiple features of the two width images to be recognized
Vector can be expressed as:
P={ P1,...,PM, Q={ Q1,...,QM}
It is described to be represented by per sub-regions combination of multiple features vector:
Pm={ pm,1,...,pm,N, Qm={ qm,1,...,qm,N}
Wherein m is more than or equal to 1 and less than or equal to M.
3. according to the method described in claim 2, it is characterized in that, described determine the two width images to be recognized M sub-regions
Each corresponding sub-region between subregion combination of multiple features vector Hausdorff distances, including:
Pass through first function dH(Pm,Qm) determine the Hausdorff of subregion feature vector between each corresponding sub-region away from
From the dH(Pm,Qm) function is as follows:
dH(Pm,Qm)=max { hk(Pm,Qm),hk(Qm,Pm),
Wherein hk(Pm,Qm) the Hausdorff distances that represent k ranks be all element distances in P and Q | | p-q | | middle kth is small
Value, wherein m are more than or equal to 1 and less than or equal to M.
It is 4. according to the method described in claim 3, it is characterized in that, described according to the two width images to be recognized M sub-regions
In each corresponding sub-region between subregion combination of multiple features vector Hausdorff distances, it is how special to obtain multizone
The Hausdorff distances of mix vector are levied, including:
Pass through second function DH(Pm,Qm) connect the Hausdorff distances of the subregion combination of multiple features vector, with
Obtain the Hausdorff distances of the multizone combination of multiple features vector;
Wherein, DH(Pm,Qm)=f (dH(P1,Q1),...,dH(PM,QM)), the second function DH(Pm,Qm) it is to described M son
The linear weighted function summing function of the Hausdorff distances of every sub-regions combination of multiple features vector in region.
5. a kind of pattern recognition device, which is characterized in that including:
Acquisition module, for obtaining the multizone combination of multiple features of two width images to be recognized vector;The multizone multiple features group
Resultant vector extracts by the way that the images to be recognized is divided into M sub-regions and respectively N number of feature vector to the M sub-regions
It obtains, the images to be recognized is coloured image and is image of clothing, the corresponding feature in each region in the images to be recognized
It is used both independently of each, includes the feature of N number of type in the M sub-regions per sub-regions;
First determining module, for determining between each corresponding sub-region in the M sub-regions of the two width images to be recognized
Subregion combination of multiple features vector Hausdorff distances;
Second determining module, between each corresponding sub-region in the two width images to be recognized M sub-regions
The Hausdorff distances of subregion combination of multiple features vector, with determine multizone combination of multiple features vector Hausdorff away from
From;
3rd determining module, for the Hausdorff distance values according to the multizone combination of multiple features vector come determine described in
The similitude of two width images to be recognized.
6. device according to claim 5, which is characterized in that the multizone combination of multiple features of the two width images to be recognized
Vector is expressed as:
P={ P1,...,PM, Q={ Q1,...,QM}
It is described to be represented by per sub-regions combination of multiple features vector:
Pm={ pm,1,…,pm,N, Qm={ qm,1,...,qm,N}
Wherein m is more than or equal to 1 and less than or equal to M.
7. device according to claim 6, which is characterized in that the first determining module, it is every in M sub-regions for determining
The Hausdorff distances of subregion combination of multiple features vector between a corresponding sub-region, including:
Pass through first function dH(Pm,Qm) determine the Hausdorff of subregion feature vector between each corresponding sub-region away from
From the first function dH(Pm,Qm) formula is expressed as:
dH(Pm,Qm)=max { hk(Pm,Qm),hk(Qm,Pm),
Wherein, hk(Pm,Qm) the Hausdorff distances that represent k ranks be all element distances in P and Q | | p-q | | middle kth is small
Value;
Wherein, m is more than or equal to 1 and less than or equal to M.
8. device according to claim 7, which is characterized in that the second determining module, for be identified according to described two
The Hausdorff distances of subregion combination of multiple features vector between each corresponding sub-region in the M sub-regions of image, with
Determine the Hausdorff distances of multizone combination of multiple features vector, including:
Pass through second function DH(Pm,Qm) connect the Hausdorff distances of the subregion combination of multiple features vector, with
Obtain the Hausdorff distances of the multizone combination of multiple features vector;
Wherein, second function DH(Pm,Qm) be expressed as:
DH(Pm,Qm)=f (dH(P1,Q1),...,dH(PM,QM)),
The second function DH(Pm,Qm) it is Hausdorff to every sub-regions combination of multiple features vector in M sub-regions
The linear weighted function summing function of distance.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610865339 | 2016-09-29 | ||
CN2016108653397 | 2016-09-29 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106529546A CN106529546A (en) | 2017-03-22 |
CN106529546B true CN106529546B (en) | 2018-05-29 |
Family
ID=58331202
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610881419.1A Active CN106529546B (en) | 2016-09-29 | 2016-10-09 | A kind of image-recognizing method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106529546B (en) |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101763507A (en) * | 2010-01-20 | 2010-06-30 | 北京智慧眼科技发展有限公司 | Face recognition method and face recognition system |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5668932B2 (en) * | 2011-05-23 | 2015-02-12 | 株式会社モルフォ | Image identification device, image identification method, image identification program, and recording medium |
-
2016
- 2016-10-09 CN CN201610881419.1A patent/CN106529546B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101763507A (en) * | 2010-01-20 | 2010-06-30 | 北京智慧眼科技发展有限公司 | Face recognition method and face recognition system |
Non-Patent Citations (1)
Title |
---|
基于改进的加权Hausdorff距离的图像匹配;蒋新士 等;《计算机应用研究》;20070430;第24卷(第4期);第182页-183页 * |
Also Published As
Publication number | Publication date |
---|---|
CN106529546A (en) | 2017-03-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zeiler et al. | Visualizing and understanding convolutional networks | |
Pires et al. | Local descriptors for soybean disease recognition | |
US9053367B2 (en) | Detector evolution with multi-order contextual co-occurrence | |
WO2013053320A1 (en) | Image retrieval method and device | |
Xiong et al. | AI-NET: Attention inception neural networks for hyperspectral image classification | |
CN111738090A (en) | Pedestrian re-recognition model training method and device and pedestrian re-recognition method and device | |
CN108664986B (en) | Based on lpNorm regularized multi-task learning image classification method and system | |
CN108537732A (en) | Fast image splicing method based on PCA-SIFT | |
CN109063776A (en) | Image identifies network training method, device and image recognition methods and device again again | |
WO2023173599A1 (en) | Method and apparatus for classifying fine-granularity images based on image block scoring | |
CN109886267A (en) | A kind of soft image conspicuousness detection method based on optimal feature selection | |
CN107977948B (en) | Salient map fusion method facing community image | |
CN113536978B (en) | Camouflage target detection method based on saliency | |
Zhang et al. | Detection of regions of interest in a high-spatial-resolution remote sensing image based on an adaptive spatial subsampling visual attention model | |
CN109858494A (en) | Conspicuousness object detection method and device in a kind of soft image | |
Yang et al. | Research into a feature selection method for hyperspectral imagery using PSO and SVM | |
Demir et al. | Clustering-based extraction of border training patterns for accurate SVM classification of hyperspectral images | |
CN106952232B (en) | A kind of picture and text fragment restoration methods based on ant group algorithm | |
Fan et al. | Color-SURF: A surf descriptor with local kernel color histograms | |
Tokarczyk et al. | Beyond hand-crafted features in remote sensing | |
Hu et al. | MINet: Multilevel inheritance network-based aerial scene classification | |
CN106529546B (en) | A kind of image-recognizing method and device | |
Bhagavathy et al. | Modeling object classes in aerial images using texture motifs | |
Tang et al. | An improved local feature descriptor via soft binning | |
Estrada et al. | Appearance-based keypoint clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |