CN107016414A - A kind of recognition methods of footprint - Google Patents
A kind of recognition methods of footprint Download PDFInfo
- Publication number
- CN107016414A CN107016414A CN201710228905.8A CN201710228905A CN107016414A CN 107016414 A CN107016414 A CN 107016414A CN 201710228905 A CN201710228905 A CN 201710228905A CN 107016414 A CN107016414 A CN 107016414A
- Authority
- CN
- China
- Prior art keywords
- footprint
- region
- frequency spectrum
- image
- size
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
-
- 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
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of recognition methods of footprint, it is divided into off-line training process SaWith ONLINE RECOGNITION process SbTwo parts, in off-line training, are pre-processed to the footprint region of identification, extract footprint provincial characteristics after slant correction and store, build footprint property data base；It is online real not during, by footprint characteristic to be identified and prestore the footprint data in property data base and calculate similarity score, the identification to footprint is completed by score rank.This method extracts the footprint shape informations such as footprint length, angle not from image, therefore is difficult to be influenceed by noise and the anglec of rotation, considers from the global feature of footprint, improves and automatically extract precision, reduce the difficulty of identification.And wavelet transformation is used, texture effects when wearing socks are eliminated, without to wearing socks and making a distinction barefoot, applicability is more extensive.
Description
Technical field
The present invention relates to a kind of recognition methods of footprint, belong to the identification field of footprint.
Background technology
Current footprint recognition methods mainly in footprint region, extract foot length, the palm it is wide, with wide, heel center to each
The center of toe and the morphological feature such as the angle of transverse axis and sole area, are identified with reference to various graders.But mesh
Preceding footprint recognition methods is typically to extract the morphological feature such as morphological feature such as length and width, angle from image to be identified.Should
The method of kind is easily affected by noise, and extraction accuracy is low, increases whole identification process difficulty and accuracy, and there is presently no dependence
The method that footprint provincial characteristics is identified.
The content of the invention
There is provided a kind of recognition methods of footprint for proposition of the invention for problem above, it is characterised in that including：Offline instruction
Practice process SaWith ONLINE RECOGNITION process Sb；
The off-line training process includes：
Sa1：The footprint tonogram picture of Acquisition Instrument collection removes noise and extracts footprint region；
Sa2：To the Sa1The footprint regional dip correction of middle acquisition；
Sa3:To the Sa2Image after the correction of middle acquisition, carries out size-normalized operation.
Sa4：To the Sa3Middle acquisition it is size-normalized after footprint region carry out before and after sufficient subregion, to the pin after subregion
Print Region Feature Extraction；
Sa5：Utilize the Sa4Obtained in feature, formed property data base,
The property data base is expressed as, D={ Fi, i=1,2 ..., N }, wherein, FiRepresent the i-th extracted footprint area
The feature in domain, N represents number of samples；
The ONLINE RECOGNITION process SbIncluding：
Sb1：Influence of noise and slant correction processing are removed to footprint region to be identified；
Sb2：Footprint region after correction is subjected to size-normalized operation.
Sb3：Sufficient subregion, then carries out wavelet transformation and footprint characteristic information respectively again before and after being carried out according to certain ratio
Extract；
Sb4：According to footprint characteristic information is extracted, by footprint data to be identified and offline instruction by the way of cosine is measured
Practice process SaFootprint data in the database of middle storage are calculated, by the similarity score calculated according to from big to small
Mode is ranked up, before finding out in the ranking in K footprint the individual most classification of occurrence number as footprint to be checked classification,
Complete the identification of footprint.
Further, Sa11：The threshold value in footprint region is obtained by Da-Jin algorithm, binaryzation is carried out to image according to the threshold value
Processing, obtains the bianry image of footprint；Then radius size is used to be carried out for 15 collar plate shape structural element to bianry image swollen
Swollen computing expanded after image f；Radius size is used to be carried out for 10 collar plate shape structural element to the image f after expansion again
Erosion operation；The connected domain of entire image is finally obtained using eight neighborhood labeling algorithm；
Sa12：Each connected region area in image is calculated, and by the arrangement of its size descending.Take and come the connected region of the 5th
The value of area removes the connected region that area is less than given threshold as noise processed as threshold value, obtains after noise processed
Footprint region.
Further, Sa21：By Sa1The footprint region that is obtained is divided into forefoot region and hindfoot areas, according to 12:13
Height ratio is divided into forefoot region Img1 and hindfoot areas Img2 to footmark image；
Sa22：Footprint hindfoot areas Img2 is subjected to endpoint detections, the marginal point of footprint metapedes is obtained, takes hindfoot areas
The key point is sought to each marginal point of hindfoot areas as key point in the midpoint of most left edge point and most upper right edge point line
Distance, the direction where the maximum straight line of distance is set to the principal direction in footprint region.
Sa23：Line tilt correction is entered in inclination angle according to principal direction and horizontal direction to view picture footprint region.
Further, Sa31：Image after correction is carried out size-normalized.Concrete operations are：Build the empty square of w × h dimensions
Battle array T, using the coordinate difference of footprint regional center behind matrix T center and correction as offset (dx, dy), by the footprint area after correction
All pixels being not zero in domain move to matrix T center by offset (dx, dy), constitute new footprint region.
Sa32:Footprint region after size-normalized re-starts front foot, metapedes subregion, by judging the outer of footprint region
Connect the height and width of rectangle, obtain the forefoot region Imgt after subregion and with hindfoot areas Imgb；
Sa33：The texture information for wearing socks belongs to detailed information, i.e. high-frequency information, to eliminate the influence that texture information is brought, uses
Harr small echos are female wave function, to Sa32Middle image Imgt and image Imgb carry out n-layer wavelet transformation respectively, by the low frequency of n-th layer
Coefficient is designated as W respectivelyTAnd WB；
Sa34：To step Sa33Obtained WTAnd WBFourier transformation is carried out, sole region frequency spectrum is obtainedAnd heel
Region frequency spectrumBy building bandpass filter, obtaining filtered frequency spectrum isWithAnd to its carry out
Log-polar transform, obtains frequency spectrum rTAnd rB；To the frequency spectrum rTAnd rBFourier transformation is carried out, amplitude is extracted, M dimensions are obtained
Spectral vectors FTAnd FB, merged the vectorial F to form 2M dimensions as footprint feature.
The present invention is due to taking above technical scheme, and it has advantages below：1) this method does not extract footprint from image
The footprint shape information such as length, angle, therefore be difficult to be influenceed by noise and the anglec of rotation, consider from the global feature of footprint,
Improve and automatically extract precision, reduce the difficulty of identification.2) the footprint feature based on Fourier plum forests has rotation translation not
Denaturation, the recognition speed of this method is fast.3) use wavelet transformation, eliminate texture effects when wearing socks, without to wear socks and barefoot
Make a distinction, applicability is more extensive.
Brief description of the drawings
, below will be to embodiment or existing for clearer explanation embodiments of the invention or the technical scheme of prior art
There is the accompanying drawing used required in technology description to do one simply to introduce, it should be apparent that, drawings in the following description are only
Some embodiments of the present invention, for those of ordinary skill in the art, on the premise of not paying creative work, may be used also
To obtain other accompanying drawings according to these accompanying drawings.
Fig. 1 is a kind of flow chart of the recognition methods of footprint of the invention；
Fig. 2 is Fig. 1 off-line training process, the flow chart of ONLINE RECOGNITION process；
Corrections and extraction figure of the Fig. 3 for footprint of the present invention.
Embodiment
To make the purpose, technical scheme and advantage of embodiments of the invention clearer, with reference to the embodiment of the present invention
In accompanying drawing, clear complete description is carried out to the technical scheme in the embodiment of the present invention：
A kind of flow chart of footprint identifying system as shown in Figure 1, as preferred embodiment, its identification to footprint
Process is：The footprint region of identification is pre-processed, footprint provincial characteristics is extracted after slant correction and is stored, and builds footprint feature
Database.Footprint characteristic to be identified and the footprint data prestored in property data base calculate similarity score.Phase
Like occurrence number in K footprint before in property score rank most to classification of many classifications as footprint to be checked, footprint identification is completed.
As shown in Fig. 2 in the present embodiment, footprint identification process includes：Off-line training process and ONLINE RECOGNITION process；
When off-line training process, the footprint tonogram picture of Acquisition Instrument collection removes influence of noise, extracts footprint area
Domain.In the present embodiment, the threshold value in footprint region is obtained using Da-Jin algorithm, image is carried out at binaryzation according to the threshold value
Reason, obtains the bianry image of footprint.It can be understood as in other embodiments, footprint can be handled using other modes
Image, as long as disclosure satisfy that the threshold value that can effectively judge image has reached the effect judged image.Using radius
Size carries out the image f after dilation operation is expanded to bianry image for 15 collar plate shape structural element, then using radius
Size carries out erosion operation for 10 collar plate shape structural element to the image f after expansion, is finally obtained using eight neighborhood labeling algorithm
Obtain the connected domain of entire image；Each connected region area in image is calculated, and by the arrangement of its size descending.Take and come the 5th
The value of connected region area as threshold value, area is less than given threshold connected region is as noise processed and removes noise,
Obtain the footprint region after noise processed.
Further, in the present embodiment, it is reduction anglec of rotation influence, centering obtains footprint region and enters line tilt school
Just.As preferred embodiment, by the footprint region of acquisition roughly be divided into forefoot region and hindfoot areas.According to certain
Height ratio is divided into forefoot region Img1 and hindfoot areas Img2 to footmark image.In the present embodiment, highly than for 12:
13.Footprint hindfoot areas Img2 is subjected to endpoint detections, the marginal point of footprint metapedes is obtained, takes hindfoot areas most left edge
The midpoint of point and most upper right edge point line calculates the key point to the distance of each marginal point of hindfoot areas as key point, will
Direction where the maximum straight line of distance is set to the principal direction in footprint region.Inclination angle pair according to principal direction and horizontal direction again
Line tilt correction is entered in view picture footprint region.
Further, in the present embodiment, the image after correction is carried out size-normalized.Build the empty square of w × h dimensions
Battle array T, using the coordinate difference of footprint regional center behind matrix T center and correction as offset (dx, dy), by the footprint area after correction
All pixels being not zero in domain move to matrix T center by offset (dx, dy), constitute new footprint region.
Further, in the present embodiment, to acquisition it is size-normalized after image carry out feature extraction, correction
Footprint region afterwards re-starts front foot, metapedes subregion.In the present embodiment, the zoning ordinance used for：First determine whether pin
The height and width of the boundary rectangle in region are printed, if height is more than width, according to 3：2 height ratio is entered to footprint region
The rough subregion of row, is divided into forefoot region Imgt and hindfoot areas Imgb.Otherwise according to 3:2 width ratio is carried out to footprint region
Subregion.
Footprint when the footprint of required identification is to wear socks, its texture information gathered belongs to detailed information, i.e. high-frequency information,
To eliminate the influence that texture information is brought, Harr small echos are used for female wave function, to forefoot region Img1 in previous step with after
Sufficient region Img2 carries out n-layer wavelet transformation respectively, the low frequency coefficient of their n-th layers is designated as W respectivelyTAnd WB, in this embodiment party
In formula, n=3.
In the present embodiment, the W to obtainingTAnd WBCarry out Fourier transformation,
Wherein M representing matrixs vector WTLine number, N representing matrix vectors WTColumns, M-1 representing matrix vectors WTLine number subtracts one,
N-1 representing matrix vectors WTColumns subtracts one, A representing matrix vectors WBLine number, B representing matrix vectors WBColumns, A-1 representing matrixs
Vector WBLine number subtracts one, B-1 representing matrix vectors WBColumns subtracts one, takes its amplitude to obtain forefoot region Img1 frequency spectrums
With hindfoot areas Img2 frequency spectrums
Then bandpass filter is built, passes through formula
N=20 in the present embodiment, wherein LP represent low pass filter, and HP represents high-pass filter.
HP (ξ, η)=(1-X (ξ, η)) (2-X (ξ, η)) (35)
Wherein：D1=ξ, η | ξ2+η2≤r2,X (ξ, η)=cos
ξπcosηπ,ξ≥-0.5,η≤0.5。
D value is higher value in the frequency spectrum line number or columns to be filtered, after filtering the filtered frequency spectrum difference of device
ForWithIn the present embodiment, set, ρ=0.35m, θ=0.35l, by log-polar transform to be carried out
The ranks number of frequency spectrum is determined.To filtered frequency spectrumWithLog-polar transform is carried out, frequency spectrum r is obtainedTAnd rB。
As preferred embodiment, to rTAnd rBFourier transformation is carried out, amplitude is extracted, the spectral vectors F of M dimensions is obtainedTAnd FB, will
It merges the vectorial F for forming 2M dimensions as footprint feature.Wherein,It is forefoot region Img1 by after bandpass filter
The frequency spectrum arrived,Pass through the frequency spectrum obtained after bandpass filter, r for hindfoot areas Img2TIt is right to pass through for forefoot region Img1
The frequency spectrum obtained after number polar coordinate transform, rBPass through the frequency spectrum that is obtained after log-polar transform for hindfoot areas Img2.
In the present embodiment, footprint database is built, by extracting the feature of every width footprint, footprint property data base D is constituted
={ Fi, i=1,2 ..., N }, FiThe feature in the i-th extracted footprint region is represented, N represents number of samples.Building database
When, do not differentiate between and wear socks footmark and barefoot footmark and left and right pin.
When ONLINE RECOGNITION process, influence of noise is removed to footprint region to be identified first and slant correction is pre-
After processing, the footprint region for removing influence of noise is carried out to the operation of yardstick standardization, then carried out according to a certain percentage forward and backward
Sufficient subregion, then carries out wavelet transformation and feature extraction respectively to forward and backward sufficient area.
In the present embodiment, according to footprint characteristic information is extracted, by footprint data to be identified and the number prestored
According to the footprint data in storehouse, the method measured by cosine carries out calculating similarity score, by similarity score according to from greatly to
Small mode is ranked up, before finding out in similitude ranking in K footprint the individual most classification of occurrence number as footprint to be checked
Classification, complete to footprint recognize.
The foregoing is only a preferred embodiment of the present invention, but protection scope of the present invention be not limited thereto,
Any one skilled in the art the invention discloses technical scope in, technique according to the invention scheme and its
Inventive concept is subject to equivalent substitution or change, should all be included within the scope of the present invention.
Claims (8)
1. a kind of recognition methods of footprint, it is characterised in that comprise the following steps：
Off-line training process Sa：Noise and slant correction pretreatment are removed to footprint region in training set, footprint area is extracted
Domain, by the footprint region of the extraction be put in predetermined null images center realize it is size-normalized；To the figure after size-normalized
Sufficient subregion before and after as carrying out, and feature is extracted, store and build footprint property data base D；The predetermined null images T refers to structure
Build the full null matrix that size is w × h, wherein w=max (wi, i=1 ..., N), h=max (hi, i=1 ..., N), wiAnd hiPoint
The width and height of every width training image are not represented；It is described size-normalized to refer to all pictures being not zero in footprint region
Vegetarian refreshments, matrix T center is moved to by offset, constitutes new footprint region；The offset (dx, dy) is equal to matrix T
Center and correction after footprint regional center coordinate difference；
ONLINE RECOGNITION process Sb：Footprint region to be identified pre-processes, standardizes and extracted feature as stated above, and with it is advance
The footprint data stored in property data base D calculate characteristic similarity, according to similarity score, provide result of determination, completion pair
The identification of footprint.
2. a kind of recognition methods of footprint according to claim 1, is further characterized in that：The off-line training process SaBag
Include following steps：
Sa1：The footprint tonogram picture of collection, is removed after noise, extracts footprint region；
Sa2：Correct the Sa1The footprint region of middle acquisition；
Sa3：To the picture size standardized operation after correction；
Sa4：To the Sa3Middle acquisition it is size-normalized after footprint region carry out before and after sufficient subregion, to the footprint area after subregion
Domain carries out feature extraction；
Sa5：By the Sa4Middle extracted feature, construction feature database；
The property data base is expressed as, D={ Fi, i=1,2 ..., N }, wherein, FiRepresent i-th extracted of footprint region
Feature, N represents number of samples.
3. a kind of recognition methods of footprint according to claim 1, is further characterized in that：The ONLINE RECOGNITION process SbBag
Include following steps：
Sb1：Influence of noise and slant correction processing are removed to footprint region to be identified；
Sb2：Footprint region after correction is carried out size-normalized；
Sb3：To the footprint region after size-normalized according to 12:13 carry out front and rear sufficient subregion, to the front foot after subregion, metapedes point
Carry out not wavelet transformation and footprint feature information extraction；It is described it is size-normalized be build w × h dimension empty matrix T after, with square
The coordinate difference of footprint regional center is offset (dx, dy) behind battle array T center and correction, and the footprint region after correction is all not
The pixel for being zero moves to matrix T center by offset (dx, dy)；
Sb4：According to footprint characteristic information is extracted, by footprint data to be identified and off-line training mistake by the way of cosine is measured
Footprint data in the database stored in journey are calculated, and the similarity score calculated is entered in the way of from big to small
Row sequence, a most classification of occurrence number, as the classification of footprint to be checked, completes pin in K footprint before finding out in the ranking
The identification of print.
4. a kind of recognition methods of footprint according to claim 2, is further characterized in that：Described step Sa1Specifically include
Following steps：
Sa11：The threshold value in footprint region is obtained by Da-Jin algorithm, binary conversion treatment is carried out to image according to the threshold value, footprint is obtained
Bianry image, use radius size for 15 pixels collar plate shape structural element to bianry image carry out dilation operation expanded
Image f afterwards；Radius size is used to carry out erosion operation to the image f after expansion for the collar plate shape structural element of 10 pixels again；
The connected domain of entire image is obtained using eight neighborhood labeling algorithm；
Sa12：Each connected region area in image is calculated, and by the arrangement of its size descending.Take and come the connected region area of the 5th
Value as threshold value, the connected region that area is less than given threshold is removed as noise processed, the pin after noise processed is obtained
Print region.
5. a kind of recognition methods of footprint according to claim 2, is further characterized in that：Described step Sa2Specifically include
Following steps：
Sa21：The footprint region obtained is divided into forefoot region Img1 and hindfoot areas Img2, according to 12:13 height ratio
Example is divided into forefoot region Img1 and hindfoot areas Img2 to footmark image；
Sa22：Footprint hindfoot areas Img2 is subjected to endpoint detections, the marginal point of footprint metapedes is obtained, takes metapedes image-region
On the midpoint of most left edge point and most upper right edge point line be used as key point, calculate the key point to each side of hindfoot areas
The distance of edge point, the direction where the maximum straight line of distance is set to the principal direction in footprint region；
Sa23：Line tilt correction is entered in angle of inclination according to principal direction and horizontal direction to view picture footprint region.
6. a kind of recognition methods of footprint according to claim 2, is further characterized in that：
Sa31：Image after correction is carried out size-normalized；
Sa32：By the height and width of the boundary rectangle that judges footprint region, before being re-started to the footprint region after correction
Foot, metapedes subregion, obtain the forefoot region Img1 after subregion and hindfoot areas Img2；
Sa33：Harr small echos are used for female wave function, to Sa32It is small that middle forefoot region Img1 and hindfoot areas Img2 carry out n-layer respectively
Wave conversion, W is designated as by the low frequency coefficient of n-th layer respectivelyTAnd WB；Wherein WTRepresent the low frequency for the n-th layer that forefoot region Img1 is obtained
Coefficient, WBRepresent the low frequency coefficient for the n-th layer that hindfoot areas Img2 is obtained；
Sa34：To step Sa33Obtained WTAnd WBFourier transformation is carried out, forefoot region Img1 frequency spectrum is obtainedWith it is rear
Sufficient region Img2 frequency spectrumBy building bandpass filter, obtaining filtered frequency spectrum isWithAnd
Log-polar transform is carried out to it, frequency spectrum r is obtainedTAnd rB；
Wherein,It is forefoot region Img1 by the frequency spectrum that is obtained after bandpass filter,It is logical for hindfoot areas Img2
Cross the frequency spectrum obtained after bandpass filter, rTPass through the frequency spectrum obtained after log-polar transform, r for forefoot region Img1BTo be rear
Sufficient region Img2 passes through the frequency spectrum that is obtained after log-polar transform；
To the frequency spectrum rTAnd rBFourier transformation is carried out, amplitude is extracted, the spectral vectors F of M dimensions is obtainedTAnd FB, merged shape
Footprint feature is used as into the 2M vectorial F tieed up.
7. a kind of recognition methods of footprint according to claim 6, is further characterized in that：The conversion of the Fourier transformation
Process is：
Wherein M representing matrixs vector WTLine number, N representing matrix vectors WTColumns, M-1 representing matrix vectors WTLine number subtracts one, N-1
Representing matrix vector WTColumns subtracts one, A representing matrix vectors WBLine number, B representing matrix vectors WBColumns, A-1 representing matrix vectors
WBLine number subtracts one, B-1 representing matrix vectors WBColumns subtracts one, takes its amplitude to obtain forefoot region Img1 frequency spectrumsWith it is rear
Sufficient region Img2 frequency spectrums
8. a kind of recognition methods of footprint according to claim 6, is further characterized in that：The bandpass filtering process is：
Wherein n=20, LP represent low pass filter；HP represents high-pass filter；
HP (ξ, η)=(1-X (ξ, η)) (2-X (ξ, η)) (35)
Wherein：D1=ξ, η | ξ2+η2≤r2,X (ξ, η)=cos ξ π
cosηπ,ξ≥-0.5,η≤0.5；
It is D to compare between the frequency spectrum line number of filtering and the columns of filtering higher value, after filtering after frequency spectrum be respectivelyWith
Set, ρ=0.35m, θ=0.35l are determined by the ranks number of the frequency spectrum of log-polar transform to be carried out.After filtering
Frequency spectrumWithLog-polar transform is carried out, frequency spectrum r is obtainedTAnd rB；
Again to rTAnd rBFourier transformation is carried out, amplitude is extracted, the spectral vectors F of M dimensions is obtainedTAnd FB, merged to form 2M dimensions
Vectorial F be used as footprint feature.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710228905.8A CN107016414A (en) | 2017-04-10 | 2017-04-10 | A kind of recognition methods of footprint |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710228905.8A CN107016414A (en) | 2017-04-10 | 2017-04-10 | A kind of recognition methods of footprint |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107016414A true CN107016414A (en) | 2017-08-04 |
Family
ID=59446637
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710228905.8A Pending CN107016414A (en) | 2017-04-10 | 2017-04-10 | A kind of recognition methods of footprint |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107016414A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325546A (en) * | 2018-10-19 | 2019-02-12 | 大连海事大学 | A kind of combination footwork feature at time footprint recognition method |
| CN110163173A (en) * | 2019-05-27 | 2019-08-23 | 大连海事大学 | A kind of footprint expression based on weighting partial structurtes |
| CN112257662A (en) * | 2020-11-12 | 2021-01-22 | 安徽大学 | Pressure footprint image retrieval system based on deep learning |
| CN112699783A (en) * | 2020-12-29 | 2021-04-23 | 深圳力维智联技术有限公司 | Footprint identification method, footprint identification device and computer-readable storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5768420A (en) * | 1994-01-21 | 1998-06-16 | Lucent Technologies Inc. | Method and apparatus for handwriting recognition using invariant features |
| CN1482571A (en) * | 2003-04-11 | 2004-03-17 | 清华大学 | Statistic handwriting identification and verification method based on separate character |
| CN104392205A (en) * | 2014-10-24 | 2015-03-04 | 浙江力石科技股份有限公司 | Abnormal vehicle license plate recognition method and system |
-
2017
- 2017-04-10 CN CN201710228905.8A patent/CN107016414A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5768420A (en) * | 1994-01-21 | 1998-06-16 | Lucent Technologies Inc. | Method and apparatus for handwriting recognition using invariant features |
| CN1482571A (en) * | 2003-04-11 | 2004-03-17 | 清华大学 | Statistic handwriting identification and verification method based on separate character |
| CN104392205A (en) * | 2014-10-24 | 2015-03-04 | 浙江力石科技股份有限公司 | Abnormal vehicle license plate recognition method and system |
Non-Patent Citations (1)
| Title |
|---|
| XINNIAN WANG, CHI ZHANG, YANJUN WU, YINGYING SHU: "A manifold ranking based method using hybrid features for crime scene shoeprint retrieval", 《MULTIMED TOOLS APPL》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325546A (en) * | 2018-10-19 | 2019-02-12 | 大连海事大学 | A kind of combination footwork feature at time footprint recognition method |
| CN109325546B (en) * | 2018-10-19 | 2022-04-08 | 大连海事大学 | Step-by-step footprint identification method combining features of step method |
| CN110163173A (en) * | 2019-05-27 | 2019-08-23 | 大连海事大学 | A kind of footprint expression based on weighting partial structurtes |
| CN110163173B (en) * | 2019-05-27 | 2022-10-25 | 大连海事大学 | Footprint expression method based on weighted local structure |
| CN112257662A (en) * | 2020-11-12 | 2021-01-22 | 安徽大学 | Pressure footprint image retrieval system based on deep learning |
| CN112699783A (en) * | 2020-12-29 | 2021-04-23 | 深圳力维智联技术有限公司 | Footprint identification method, footprint identification device and computer-readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107316031B (en) | Image feature extraction method for pedestrian re-identification | |
| CN104751465A (en) | ORB (oriented brief) image feature registration method based on LK (Lucas-Kanade) optical flow constraint | |
| CN111340701B (en) | Circuit board image splicing method for screening matching points based on clustering method | |
| Matkan et al. | Road extraction from lidar data using support vector machine classification | |
| CN107016414A (en) | A kind of recognition methods of footprint | |
| CN107092871B (en) | Remote sensing image building detection method based on multiple dimensioned multiple features fusion | |
| CN109508663A (en) | A kind of pedestrian's recognition methods again based on multi-level supervision network | |
| CN103593832A (en) | Method for image mosaic based on feature detection operator of second order difference of Gaussian | |
| CN103136525B (en) | High-precision positioning method for special-shaped extended target by utilizing generalized Hough transformation | |
| CN110569861B (en) | Image matching positioning method based on point feature and contour feature fusion | |
| CN103473551A (en) | Station logo recognition method and system based on SIFT operators | |
| CN104200461A (en) | Mutual information image selected block and sift (scale-invariant feature transform) characteristic based remote sensing image registration method | |
| CN107066969A (en) | A kind of face identification method | |
| CN107862319B (en) | Heterogeneous high-light optical image matching error eliminating method based on neighborhood voting | |
| CN106294705A (en) | A kind of batch remote sensing image preprocess method | |
| CN110197113B (en) | Face detection method of high-precision anchor point matching strategy | |
| CN114708208B (en) | Machine vision-based famous tea tender bud identification and picking point positioning method | |
| CN111160192A (en) | Remote sensing image positioning method based on coastal linear features | |
| CN109086350B (en) | Mixed image retrieval method based on WiFi | |
| CN103913166A (en) | Star extraction method based on energy distribution | |
| CN108319961B (en) | Image ROI rapid detection method based on local feature points | |
| CN115731257A (en) | Leaf form information extraction method based on image | |
| CN112614167A (en) | Rock slice image alignment method combining single-polarization and orthogonal-polarization images | |
| CN110675442A (en) | Local stereo matching method and system combined with target identification technology | |
| CN105701473B (en) | A kind of matched method of palmprint image minutiae feature |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170804 |