CN110210342A - A kind of Humanface image matching method and its system, readable storage medium storing program for executing - Google Patents
A kind of Humanface image matching method and its system, readable storage medium storing program for executing Download PDFInfo
- Publication number
- CN110210342A CN110210342A CN201910417869.9A CN201910417869A CN110210342A CN 110210342 A CN110210342 A CN 110210342A CN 201910417869 A CN201910417869 A CN 201910417869A CN 110210342 A CN110210342 A CN 110210342A
- Authority
- CN
- China
- Prior art keywords
- image
- head zone
- matching
- registration
- subject
- 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
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000003709 image segmentation Methods 0.000 claims abstract description 35
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 19
- 230000011731 head segmentation Effects 0.000 claims abstract description 11
- 238000000605 extraction Methods 0.000 claims description 15
- 230000001815 facial effect Effects 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000005260 corrosion Methods 0.000 claims description 6
- 230000007797 corrosion Effects 0.000 claims description 6
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/08—Indexing scheme for image data processing or generation, in general involving all processing steps from image acquisition to 3D model generation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- Geometry (AREA)
- Computer Graphics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of Humanface image matching method and its system, readable storage medium storing program for executing, and method includes obtaining image and reference picture subject to registration;Image subject to registration is extracted using convolutional neural networks trained in advance and obtains the first characteristics of image, and is extracted reference picture using convolutional neural networks trained in advance and obtained the second characteristics of image;Progress subregion obtains the first head zone and the first torso area after determining image segmentation line subject to registration, and carries out subregion after determining reference picture Head segmentation line and obtain the second head zone and the second torso area;The characteristics of image of first head zone and the second head zone is matched to obtain the first matching result, the characteristics of image of the second head zone and the second head zone is matched to obtain the second matching result;The matching result of image and reference picture subject to registration is determined according to the first matching result and the second matching result.The system is for realizing the method.Feature error hiding rate and efficiency when the present invention can be improved recognition of face 3-dimensional reconstruction.
Description
Technical field
The present invention relates to technical field of face recognition, in particular to a kind of facial image based on study invariant features transformation
Matching process and its system, readable storage medium storing program for executing.
Background technique
With deepening continuously for computer vision research, three-dimensional reconstruction becomes the strong of recognition of face digital protection
Tool.The degree of precision digitlization recognition of face threedimensional model generated after recognition of face three-dimensional modeling can effectively be known for face
Indescribably supply important data and model supports.Most common recognition of face digitizing solution has, the three-dimensional reconstruction of multi-angle image.
This method only needs to obtain the photo of target, can recover the threedimensional model of the target;Universal with simple and fast, convenience
Feature.In reconstruction procedures, the recognition of face images match of different angle is the basis of three-dimensional reconstruction, and accuracy rate will directly affect
The precision of Model Reconstruction.
Summary of the invention
The problem that feature error hiding rate is higher, efficiency is lower when the present invention is for recognition of face 3-dimensional reconstruction proposes
A kind of Humanface image matching method and its system, readable storage medium storing program for executing based on study invariant features transformation.
In order to achieve the object of the present invention, according in a first aspect, the embodiment of the present invention provides a kind of Humanface image matching method,
Include:
Obtain image and reference picture subject to registration;
The image subject to registration is extracted using convolutional neural networks trained in advance and obtains the first characteristics of image, and using pre-
First trained convolutional neural networks extract the reference picture and obtain the second characteristics of image;
Progress subregion obtains the first head zone and the first torso area after determining the image segmentation line subject to registration, and really
Subregion, which is carried out, after the fixed reference picture Head segmentation line obtains the second head zone and the second torso area;
The characteristics of image of first head zone and the second head zone is matched to obtain the first matching result, it is right
The characteristics of image of second head zone and the second head zone is matched to obtain the second matching result;
The image subject to registration and described with reference to figure is determined according to first matching result and second matching result
The matching result of picture.
Preferably, the first image feature and the second characteristics of image include characteristic point and feature descriptor.
Preferably, subregion is carried out after the determination image segmentation line subject to registration obtain the first head zone and the first body
Dry region includes:
Binary conversion treatment is carried out to the image subject to registration and obtains the first binary image, and is first with the circular configuration generated
Element carries out the corrosion on Mathematical Morphology to first binary image;
First binary image is scanned from top to bottom using scanning function, and counts each horizontal scanning line
The number of upper characteristic point obtains fisrt feature point number distribution curve;
Image segmentation line subject to registration is determined according to the fisrt feature point number distribution curve.
Preferably, subregion is carried out after the determination image segmentation line subject to registration obtain the second head zone and the second body
Dry region includes:
Binary conversion treatment is carried out to the image subject to registration and obtains the second binary image, and is first with the circular configuration generated
Element carries out the corrosion on Mathematical Morphology to second binary image;
Second binary image is scanned from top to bottom using scanning function, and counts each horizontal scanning line
The number of upper characteristic point obtains second feature point number distribution curve;
Image segmentation line subject to registration is determined according to the second feature point number distribution curve.
Preferably, the characteristics of image of first head zone and the second head zone is matched to obtain the first matching
As a result it and to the characteristics of image of second head zone and the second head zone is matched to obtain the second matching result equal
It is matched in the following way:
For being located at two characteristic points of different images, according to the Euclidean distance and threshold between Feature Descriptor vector
The comparison result of value determines whether described two characteristic points are matching double points.
Preferably, if ViFor image, image subject to registration is Vi, reference picture V2, then V1And V2In the feature that detects retouch
State subclass are as follows:
Wherein,For description of the n-th characteristic point of i-th of image;
Wherein, the comparison result of the Euclidean distance and threshold value according between Feature Descriptor vector determines described two
Whether characteristic point is that matching double points include:
To eachIn F (V2) in search its arest neighbors l*(1)With secondary neighbour l*(2):
For anyArest neighbors l* is obtained using nearest neighbor search(1)With secondary neighbour l*(2),
When R is less than threshold value δ, then V1In k-th of characteristic point and V2In l*(1)A characteristic point is candidate matches, no
Then give up V1In k-th of characteristic point.
Humanface image matching method provided in an embodiment of the present invention party in recognition of face image characteristics extraction and matching
The correct matching rate of case can achieve 98%, improve 20% or so compared to the correct matching rate of SIFT and SURF method, feature
The repeatability of point improves 10% or so, while the iteration time of RANSAC is reduced 50%, and scale, illumination,
There is preferable robustness when angle converts.Therefore the scheme that the embodiment of the present invention proposes can realize characteristic point well
Correct matching, in the three-dimensional reconstruction of recognition of face have very high use value.
According to second aspect, the embodiment of the present invention provides a kind of facial image matching system, for realizing the face
Image matching method, comprising:
Image acquisition unit, for obtaining image and reference picture subject to registration;
Image characteristics extraction unit is obtained for extracting the image subject to registration using convolutional neural networks trained in advance
First characteristics of image simultaneously obtains the second characteristics of image using the convolutional neural networks extraction reference picture trained in advance;
Image segmentation unit, for determine carry out after the image segmentation line subject to registration subregion obtain the first head zone and
Subregion, which is carried out, after first torso area and the determining reference picture Head segmentation line obtains the second head zone and the second trunk
Region;
Image matching unit match for the characteristics of image to first head zone and the second head zone
It is matched to obtain second to the first matching result and to the characteristics of image of second head zone and the second head zone
With result;
Determine matching unit, it is described subject to registration for being determined according to first matching result and second matching result
The matching result of image and the reference picture.
Preferably, described image cutting unit includes the first image segmentation unit and the second image segmentation unit, and described the
One image segmentation unit obtains the first head zone and the first body for progress subregion after determining the image segmentation line subject to registration
Dry region;Second image segmentation unit obtains second for progress subregion after determining the reference picture Head segmentation line
Portion region and the second torso area.
Preferably, described image matching unit includes the first image matching unit and the second image matching unit, and described the
One image matching unit to the characteristics of image of first head zone and the second head zone for being matched to obtain first
Matching result, second image matching unit be used for the characteristics of image of second head zone and the second head zone into
Row matching obtains the second matching result.
According to the third aspect, the embodiment of the present invention provides a kind of computer readable storage medium, is stored thereon with computer
Program, when which is executed by processor, to realize the Humanface image matching method.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical 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 Humanface image matching method flow chart described in the embodiment of the present invention one.
Fig. 2 is a kind of facial image matching system schematic diagram described in the embodiment of the present invention two.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear and complete
Ground description, it is clear that described embodiment is only a part of the embodiments 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 without making creative work it is obtained it is all its
Its embodiment, shall fall within the protection scope of the present invention.
Here, it should also be noted that, in order to avoid having obscured the present invention because of unnecessary details, in the accompanying drawings only
Show with closely related structure and/or processing step according to the solution of the present invention, and be omitted little with relationship of the present invention
Other details.
As shown in Figure 1, the embodiment of the present invention one provides a kind of Humanface image matching method, comprising:
Step S1, image and reference picture subject to registration are obtained;
Step S2, the image subject to registration is extracted using convolutional neural networks trained in advance obtain the first characteristics of image,
And the reference picture is extracted using convolutional neural networks trained in advance and obtains the second characteristics of image;
Step S3, progress subregion obtains the first head zone and the first trunk area after determining the image segmentation line subject to registration
Domain, and carry out subregion after the determining reference picture Head segmentation line and obtain the second head zone and the second torso area;
Step S4, the characteristics of image of first head zone and the second head zone is matched to obtain the first matching
As a result, being matched to obtain the second matching result to the characteristics of image of second head zone and the second head zone;
Step S5, the image subject to registration and described is determined according to first matching result and second matching result
The matching result of reference picture.
Wherein, the first image feature and the second characteristics of image include characteristic point and feature descriptor.
The extraction of image characteristic point is preferably carried out in the present embodiment using LIFT neural network, wherein the people based on LIFT
Face identifies that image characteristic point extracts the recognition of face picture construction scale space needed to input, utilizes trained feature later
Detection convolutional neural networks map pyramid to calculate score, and filter out characteristic point using the method for non-maxima suppression,
Smaller piece is selected in Direction estimation as input, assesses spy also with trained Direction estimation convolutional neural networks
The direction for levying point, finally extracts the feature descriptor of 128 dimensions using trained Feature Descriptor convolutional neural networks.
Wherein, depth convolutional network is a kind of feedforward neural network, its main feature is that weight shares network structure and image
Local sensing.Learnt by the training to mass data, this method, which can be obtained, has invariance to translation, scaling and rotation
Observe the notable feature of data.LIFT is a kind of completely new depth convolutional network, it is sub by feature detection, Direction estimation, description
It calculates 3 important steps of feature extraction and is fused together and learnt, and give specific training method, maintain end
To the feature extracting method of end ga s safety degree.
In the present embodiment, after the determination image segmentation line subject to registration carry out subregion obtain the first head zone and
First torso area includes:
Binary conversion treatment is carried out to the image subject to registration and obtains the first binary image, and is first with the circular configuration generated
Element carries out the corrosion on Mathematical Morphology to first binary image;
First binary image is scanned from top to bottom using scanning function, and counts each horizontal scanning line
The number of upper characteristic point obtains fisrt feature point number distribution curve;
Image segmentation line subject to registration is determined according to the fisrt feature point number distribution curve.
In the present embodiment, after the determination image segmentation line subject to registration carry out subregion obtain the second head zone and
Second torso area includes:
Binary conversion treatment is carried out to the image subject to registration and obtains the second binary image, and is first with the circular configuration generated
Element carries out the corrosion on Mathematical Morphology to second binary image;
Second binary image is scanned from top to bottom using scanning function, and counts each horizontal scanning line
The number of upper characteristic point obtains second feature point number distribution curve;
Image segmentation line subject to registration is determined according to the second feature point number distribution curve.
Specifically, the embodiment of the present invention is to above-mentioned fisrt feature point in order to preferably find the position on segmentation head
Number distribution curve and second feature point number distribution curve are smoothed, by relevant human body priori knowledge is added,
It is known that first trough occurred is the position that recognition of face head chignon occurs, and second trough is that the present invention is real
Apply the position for the Head segmentation line that example to be looked for;Find out trough for convenience, the embodiment of the present invention is using calculating wave crest in MATLAB
Function calculates the wave trough position of requirement, so need to invert image.
Since the validity feature point of SIFT, SURF matching algorithm search is the characteristic point for being distributed in overlapping region, pass
The strategy that the matching algorithm of system carries out thorough search to image-region increases the runing time of algorithm, while but also matching is wrong
Accidentally rate increases.Therefore, the embodiment of the present invention, which proposes, determines recognition of face image and carries out subregion matching after Head segmentation line.
In the present embodiment, the characteristics of image of first head zone and the second head zone is matched to obtain
One matching result and the characteristics of image of second head zone and the second head zone is matched to obtain the second matching
As a result under type such as is all made of to be matched:
For being located at two characteristic points of different images, according to the Euclidean distance and threshold between Feature Descriptor vector
The comparison result of value determines whether described two characteristic points are matching double points.
Specifically, the recognition of face image that LIFT is extracted is special according to the position of recognition of face picture headers cut-off rule
Sign point is divided into head zone and torso area two parts, later carries out two parts image characteristic point after segmentation based on Europe respectively
2 view feature points of formula distance match.2 view feature point of recognition of face matching based on Euclidean distance is by calculating feature
Euclidean distance between vector is realized.If between the corresponding Feature Descriptor vector of two characteristic points being located on two width views
Euclidean distance be less than certain threshold value and then think that the two point features correspond to Same Scene point, i.e., as a pair of of matching double points.
Wherein, if ViFor image, image subject to registration is Vi, reference picture V2, then V1And V2In detect feature description
Subclass are as follows:
Wherein,For description of the n-th characteristic point of i-th of image;
Wherein, the comparison result of the Euclidean distance and threshold value according between Feature Descriptor vector determines described two
Whether characteristic point is that matching double points include:
To eachIn F (V2) in search its arest neighbors l*(1)With secondary neighbour l*(2):
The distance calculating method that two above formula uses all is Euclidean distance, and obtained each vector l* is 128
Dimension, since the matching point set that arest neighbors matching process obtains has the matching point set of mistake, therefore, it is necessary to further use distance
Go out candidate matches than test screen.
For anyArest neighbors l* is obtained using nearest neighbor search(1)With secondary neighbour l*(2),
When R is less than threshold value δ, then V1In k-th of characteristic point and V2In l*(1)A characteristic point is candidate matches, no
Then give up V1In k-th of characteristic point.
Humanface image matching method provided in an embodiment of the present invention party in recognition of face image characteristics extraction and matching
The correct matching rate of case can achieve 98%, improve 20% or so compared to the correct matching rate of SIFT and SURF method, feature
The repeatability of point improves 10% or so, while the iteration time of RANSAC is reduced 50%, and scale, illumination,
There is preferable robustness when angle converts.Therefore the scheme that the embodiment of the present invention proposes can realize characteristic point well
Correct matching, in the three-dimensional reconstruction of recognition of face have very high use value.
Such as Fig. 2, second embodiment of the present invention provides a kind of facial image matching systems, for realizing the facial image
Method of completing the square, comprising:
Image acquisition unit 1, for obtaining image and reference picture subject to registration;
Image characteristics extraction unit 2 is obtained for extracting the image subject to registration using convolutional neural networks trained in advance
The second characteristics of image is obtained to the first characteristics of image and using the convolutional neural networks extraction reference picture trained in advance;
Image segmentation unit 3 obtains the first head zone for progress subregion after determining the image segmentation line subject to registration
The second head zone and the second body are obtained with subregion is carried out after the first torso area and the determining reference picture Head segmentation line
Dry region;
Image matching unit 4 is matched for the characteristics of image to first head zone and the second head zone
It obtains the first matching result and the characteristics of image of second head zone and the second head zone is matched to obtain second
Matching result;
Determine matching unit 5, it is described wait match for being determined according to first matching result and second matching result
The matching result of quasi- image and the reference picture.
Preferably, described image feature extraction unit 2 includes the first image characteristics extraction unit 21 and the second characteristics of image
Extraction unit 22, the first image characteristics extraction unit 21 are used to extract using convolutional neural networks trained in advance described subject to registration
Image obtains the first characteristics of image, and the second image characteristics extraction unit 22 is used to extract using convolutional neural networks trained in advance
The reference picture obtains the second characteristics of image;
Preferably, described image cutting unit 3 includes the first image segmentation unit 31 and the second image segmentation unit 32, institute
State the first image segmentation unit 31 for determine carry out after the image segmentation line subject to registration subregion obtain the first head zone and
First torso area;Second image segmentation unit 32 is obtained for progress subregion after determining the reference picture Head segmentation line
To the second head zone and the second torso area.
Preferably, image matching unit 4 includes the first image matching unit 41 and the second image matching unit 42, the first figure
As matching unit 41 to the characteristics of image of first head zone and the second head zone for being matched to obtain first
With as a result, the second image matching unit 42 is used for the characteristics of image progress to second head zone and the second head zone
With obtaining the second matching result.
It should be noted that the system that embodiment two proposes is corresponding with the method for embodiment one, therefore, embodiment two is not detailed
The other parts stated see one the method part of embodiment and obtain, and details are not described herein again.
According to the third aspect, the embodiment of the present invention provides a kind of computer readable storage medium, is stored thereon with computer
Program, when which is executed by processor, to realize Humanface image matching method described in embodiment one.
The above is only the specific embodiment of the application, it is noted that for the ordinary skill people of the art
For member, under the premise of not departing from the application principle, several improvements and modifications can also be made, these improvements and modifications are also answered
It is considered as the protection scope of the application.
Claims (10)
1. a kind of Humanface image matching method characterized by comprising
Obtain image and reference picture subject to registration;
The image subject to registration is extracted using convolutional neural networks trained in advance and obtains the first characteristics of image, and using instruction in advance
Experienced convolutional neural networks extract the reference picture and obtain the second characteristics of image;
Progress subregion obtains the first head zone and the first torso area after determining the image segmentation line subject to registration, and determines institute
Progress subregion obtains the second head zone and the second torso area after stating reference picture Head segmentation line;
The characteristics of image of first head zone and the second head zone is matched to obtain the first matching result, to described
The characteristics of image of second head zone and the second head zone is matched to obtain the second matching result;
The image subject to registration and the reference picture are determined according to first matching result and second matching result
Matching result.
2. Humanface image matching method as described in claim 1, which is characterized in that the first image feature and the second image
Feature includes characteristic point and feature descriptor.
3. Humanface image matching method as claimed in claim 2, which is characterized in that the determination image segmentation subject to registration
Progress subregion obtains the first head zone after line and the first torso area includes:
Binary conversion treatment is carried out to the image subject to registration and obtains the first binary image, and with the circular configuration element pair of generation
First binary image carries out the corrosion on Mathematical Morphology;
First binary image is scanned from top to bottom using scanning function, and is counted special on each horizontal scanning line
The number of sign point obtains fisrt feature point number distribution curve;
Image segmentation line subject to registration is determined according to the fisrt feature point number distribution curve.
4. Humanface image matching method as claimed in claim 3, which is characterized in that the determination image segmentation subject to registration
Progress subregion obtains the second head zone after line and the second torso area includes:
Binary conversion treatment is carried out to the image subject to registration and obtains the second binary image, and with the circular configuration element pair of generation
Second binary image carries out the corrosion on Mathematical Morphology;
Second binary image is scanned from top to bottom using scanning function, and is counted special on each horizontal scanning line
The number of sign point obtains second feature point number distribution curve;
Image segmentation line subject to registration is determined according to the second feature point number distribution curve.
5. Humanface image matching method as claimed in claim 4, which is characterized in that first head zone and second
The characteristics of image in portion region is matched to obtain the first matching result and to second head zone and the second head zone
Characteristics of image matched to obtain the second matching result and be all made of under type such as and matched:
For being located at two characteristic points of different images, according to the Euclidean distance and threshold value between Feature Descriptor vector
Comparison result determines whether described two characteristic points are matching double points.
6. Humanface image matching method as claimed in claim 5, which is characterized in that set ViFor image, image subject to registration is Vi,
Reference picture is V2, then V1And V2In the Feature Descriptor set that detects are as follows:
Wherein,For description of the n-th characteristic point of i-th of image;
Wherein, the comparison result of the Euclidean distance and threshold value according between Feature Descriptor vector determines described two features
Whether point is that matching double points include:
To eachIn F (V2) in search its arest neighbors l*(1)With secondary neighbour l*(2):
For anyArest neighbors l* is obtained using nearest neighbor search(1)With secondary neighbour l*(2),
When R is less than threshold value δ, then V1In k-th of characteristic point and V2In l*(1)A characteristic point is candidate matches, is otherwise given up
Abandon V1In k-th of characteristic point.
7. a kind of facial image matching system, for realizing Humanface image matching method described in any one of claims 1-6,
It is characterised by comprising:
Image acquisition unit, for obtaining image and reference picture subject to registration;
Image characteristics extraction unit obtains first for extracting the image subject to registration using convolutional neural networks trained in advance
Characteristics of image simultaneously obtains the second characteristics of image using the convolutional neural networks extraction reference picture trained in advance;
Image segmentation unit obtains the first head zone and first for progress subregion after determining the image segmentation line subject to registration
Subregion, which is carried out, after torso area and the determining reference picture Head segmentation line obtains the second head zone and the second torso area;
Image matching unit is matched to obtain for the characteristics of image to first head zone and the second head zone
One matching result simultaneously matches the characteristics of image of second head zone and the second head zone to obtain the second matching knot
Fruit;
Matching unit is determined, for determining the image subject to registration according to first matching result and second matching result
With the matching result of the reference picture.
8. facial image matching system as claimed in claim 7, which is characterized in that described image cutting unit includes the first figure
As cutting unit and the second image segmentation unit, the first image cutting unit is for determining the image segmentation line subject to registration
Subregion is carried out afterwards obtains the first head zone and the first torso area;Second image segmentation unit is for determining the reference
Subregion is carried out after picture headers cut-off rule obtains the second head zone and the second torso area.
9. facial image matching system as claimed in claim 8, which is characterized in that described image matching unit includes the first figure
As matching unit and the second image matching unit, the first image matching unit is used for first head zone and second
The characteristics of image of head zone is matched to obtain the first matching result, and second image matching unit is used for described second
The characteristics of image of head zone and the second head zone is matched to obtain the second matching result.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
When execution, to realize Humanface image matching method as claimed in any one of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910417869.9A CN110210342A (en) | 2019-05-20 | 2019-05-20 | A kind of Humanface image matching method and its system, readable storage medium storing program for executing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910417869.9A CN110210342A (en) | 2019-05-20 | 2019-05-20 | A kind of Humanface image matching method and its system, readable storage medium storing program for executing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110210342A true CN110210342A (en) | 2019-09-06 |
Family
ID=67787802
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910417869.9A Pending CN110210342A (en) | 2019-05-20 | 2019-05-20 | A kind of Humanface image matching method and its system, readable storage medium storing program for executing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110210342A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113221805A (en) * | 2021-05-25 | 2021-08-06 | 云南电网有限责任公司电力科学研究院 | Method and device for acquiring image position of power equipment |
CN115619756A (en) * | 2022-10-31 | 2023-01-17 | 北京鹰之眼智能健康科技有限公司 | Heart region identification method of human body infrared image |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102496158A (en) * | 2011-11-24 | 2012-06-13 | 中兴通讯股份有限公司 | Method and device for image information processing |
CN105205818A (en) * | 2015-09-18 | 2015-12-30 | 国网上海市电力公司 | Method for registering infrared image and visible light image of electrical equipment |
CN106778613A (en) * | 2016-12-16 | 2017-05-31 | 广东工业大学 | A kind of auth method and device based on the matching of face cut zone |
CN108304863A (en) * | 2018-01-12 | 2018-07-20 | 西北大学 | A kind of terra cotta warriors and horses image matching method using study invariant features transformation |
-
2019
- 2019-05-20 CN CN201910417869.9A patent/CN110210342A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102496158A (en) * | 2011-11-24 | 2012-06-13 | 中兴通讯股份有限公司 | Method and device for image information processing |
CN105205818A (en) * | 2015-09-18 | 2015-12-30 | 国网上海市电力公司 | Method for registering infrared image and visible light image of electrical equipment |
CN106778613A (en) * | 2016-12-16 | 2017-05-31 | 广东工业大学 | A kind of auth method and device based on the matching of face cut zone |
CN108304863A (en) * | 2018-01-12 | 2018-07-20 | 西北大学 | A kind of terra cotta warriors and horses image matching method using study invariant features transformation |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113221805A (en) * | 2021-05-25 | 2021-08-06 | 云南电网有限责任公司电力科学研究院 | Method and device for acquiring image position of power equipment |
CN113221805B (en) * | 2021-05-25 | 2022-08-02 | 云南电网有限责任公司电力科学研究院 | Method and device for acquiring image position of power equipment |
CN115619756A (en) * | 2022-10-31 | 2023-01-17 | 北京鹰之眼智能健康科技有限公司 | Heart region identification method of human body infrared image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | Robust visual tracking using oblique random forests | |
Wang et al. | Drpose3d: Depth ranking in 3d human pose estimation | |
Guo et al. | Efficient and deep person re-identification using multi-level similarity | |
CN105740780B (en) | Method and device for detecting living human face | |
Kanhangad et al. | Contactless and pose invariant biometric identification using hand surface | |
CN110490158B (en) | Robust face alignment method based on multistage model | |
CN110110694B (en) | Visual SLAM closed-loop detection method based on target detection | |
Held et al. | Deep learning for single-view instance recognition | |
CN106682700A (en) | Block quick matching algorithm based on key point description operator | |
Capellen et al. | ConvPoseCNN: Dense convolutional 6D object pose estimation | |
CN111506759B (en) | Image matching method and device based on depth features | |
CN104091145A (en) | Human palm vein feature image acquisition method | |
CN111126116A (en) | Unmanned ship river channel garbage identification method and system | |
CN110210342A (en) | A kind of Humanface image matching method and its system, readable storage medium storing program for executing | |
CN103324921B (en) | A kind of mobile identification method based on interior finger band and mobile identification equipment thereof | |
Li et al. | FC-SLAM: Federated learning enhanced distributed visual-LiDAR SLAM in cloud robotic system | |
CN113095385B (en) | Multimode image matching method based on global and local feature description | |
Li et al. | Solving a special type of jigsaw puzzles: Banknote reconstruction from a large number of fragments | |
CN104036494B (en) | A kind of rapid matching computation method for fruit image | |
Kao et al. | Gaze tracking with particle swarm optimization | |
Cadoni et al. | Iconic methods for multimodal face recognition: a comparative study | |
Bergström et al. | Integration of visual cues for robotic grasping | |
CN108932726A (en) | A kind of method for tracking target and device | |
Mei et al. | S2-net: Self-supervision guided feature representation learning for cross-modality images | |
Hongtao et al. | Object detection method based on improved one-stage detector |
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: 20190906 |