CN104881630B - Vehicle identification method based on vehicle window segmentation and fuzzy feature - Google Patents
Vehicle identification method based on vehicle window segmentation and fuzzy feature Download PDFInfo
- Publication number
- CN104881630B CN104881630B CN201510148453.3A CN201510148453A CN104881630B CN 104881630 B CN104881630 B CN 104881630B CN 201510148453 A CN201510148453 A CN 201510148453A CN 104881630 B CN104881630 B CN 104881630B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- horz
- image
- angle
- fuzzy
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses the vehicle identification methods based on vehicle window segmentation and fuzzy feature, by the edge for extracting a preceding face image；Then quadrangle fitting is carried out according to edge, so that Fast Segmentation goes out vehicle front window windshield profile；Then the HSV image histogram feature within the scope of windshield is extracted；Vehicle is retrieved finally by the region Fuzzy difference measurement comparison of HSV histogram.The present invention gets rid of the single unreliability by license number search, improves the effect of vehicle image retrieval.
Description
[technical field]
The invention belongs to traffic management technology fields, more particularly to a kind of vehicle based on vehicle window segmentation and fuzzy feature
Recognition methods.
[background technique]
It currently has stable political situation the fields such as public security in urban transportation, city, the application of the monitoring systems such as electronic eyes becomes increasingly popular.These are
The system acquisition data such as continuous collecting vehicle image and video extract useful information from the huge database collected to assist
Regulatory authorities further carry out scientific management and decision.For example at major urban arterial highway crossing and expressway key position, lead to
It crosses using license plate recognition technology and helps to carry out the Intelligent treatments such as vehicle tracking, flow analysis, have been achieved for very ten-strike.
But to social public security require it is higher and higher instantly, because suspected vehicles often deliberately block or more change trains
Board only identifies that license plate can no longer meet the Search Requirement for specific suspicion of crime vehicle, it is necessary to pass through other in image
The characteristic information for being difficult to replace to retrieve specific objective vehicle in huge database.And on the front window windshield of vehicle, one
As be pasted with annual test, compulsory insurance for traffic accident of motor-drivenvehicle and environmental protection tests etc. label, it is also possible to be placed with the relatively-stationary object in some positions such as mascot
The color characteristic and relative positional relationship of part, these labels and object just constitute the characteristics of image of windscreen, facilitate big
Measure the fast and reliable retrieval in front face image data base to target vehicle.
[summary of the invention]
It is an object of the invention to solve the problems, such as the above-mentioned of the existing field, a kind of reliable target vehicle image is provided
Search method, this method pass through the windshield profile of identification front part of vehicle, and according to image HSV histogram feature in profile
Region Fuzzy diversity factor is retrieved to realize, comprising the following steps:
S01 obtains the edge graph of image to be detected；
S02 carries out straight line fitting close to horizontal ledgement and subvertical vertical moulding to edge graph, and different is straight
Line is combined into different quadrangles；
S03 using with target vehicle vehicle window sample shape similarity and the highest quadrangle of the edge goodness of fit as to be detected
The elementary contour of image vehicle window；
S04 determines the HSV histogram of image to be detected vehicle window, specifically includes:
Using the highest quadrangle of weight as image to be detected vehicle window elementary contour, the image in profile is subjected to affine change
Regular rectangular shape is changed to as characteristic area, the subregion for being M*N by this feature region even partition, and calculates each subregion
Hsv color histogram, steps are as follows:
1) by the image in the quadrangle for representing image to be detected vehicle window elementary contour by affine transformation to uniform sizes
Regular rectangular shape；
2) the uniform subregion R for being M*N by this feature region divisionm,n, wherein m≤M, n≤N；
3) image is transformed into hsv color space from RGB color；
4) all subregion R is calculatedm,nThe histogram Hist in the middle channel H, S, Vm,n H, Histm,n S, Histm,n VInstitute in histogram
There is bin to normalize to [0,1] section；
The HSV histogram of S05 and target vehicle vehicle window sample carries out region Fuzzy difference angle value and calculates；
The HSV histogram of image in vehicle database and target vehicle in the same way, is carried out region one by one by S06
Fuzzy difference angle value calculates；
S07 is worth search result according to region Fuzzy diversity factor；
The preceding face image feature of image and target vehicle to be checked in vehicle database is carried out region one by one by the method
Fuzzy difference angle value calculates, and retrieves most like several images as search result, meter according to region Fuzzy difference angle value
Calculation method is as follows:
1) to each of vehicle characteristics region to be checked subregion Rm,n T, calculate pair of itself and target vehicle characteristic area
Answer subregion Rm,n SFuzzy diversity factor, the channel difference angle value of each corresponding sub-region are as follows:
Wherein c indicates Histm,n H, Histm,n S, Histm,n VThree channel histograms, bmaxIndicate the maximum under the histogram
Bin number, Δb,kIndicate that b, the difference of two bin of k, formula ensure that when two histograms are in the difference of the same bin or adjacent bin
When being worth smaller, each channel difference angle value Fm,n CAlso smaller；
According to each channel difference degree, the Fuzzy difference angle value of the corresponding sub-region is calculated are as follows:
Fm,n=αH*Fm,n H+αS*Fm,n S+αV*Fm,n V
Wherein αcFor the constant factor in each channel；
2) according to the diversity factor F of corresponding sub-regionm,n, zoning diversity factor F；
It is identified as target carriage with the region lesser several images of Fuzzy difference angle value of target vehicle in vehicle to be measured
?.
Further, edge graph described in step S01 is that the Canny operator of using area adaptive threshold is found out.
Further, step S02 is specifically included: is searched in edge graph close to horizontal ledgement and subvertical
Vertical moulding, and lines are polymerize and screened；Straight line fitting is carried out to horizontal line, filters out the high horizontal line of degree of fitting, and calculate
Main level tilt angle；Screening and straight line fitting are carried out to vertical line；According to Straight Line Fitting Parameters, from horizontal line Resource selection two
Horizontal line, with from left and right vertical line concentrate it is each select a line group to be combined into a quadrangle, different combinations constitutes quadrangle collection.
Further, to horizontal line straight line fitting, the high horizontal line of degree of fitting is filtered out, and calculates main level tilt angle
During, main horizontal tilt angle Horz_Angle is determined by the way of ballot, voting process is as follows:
1) least square method is used, according to the point sequence fitting a straight line y=kx+b of each horizontal line, and digital simulation degree is
Sigma is as follows, wherein xs, ysFor the point in horizontal line point sequence, len is horizontal line length；
Y=kxs+b
Sigma=∑ (y-ys)2/len
2) according to the parameter k of fitting a straight line, all fitting a straight lines of the level angle within the scope of positive and negative 10 ° are found out, and really
Determine angular range [Ang_Min, Ang_Max], the parameter k is straight slope；
3) [Ang_Min, Ang_Max] is divided into 20 bin, each bin angular span is Angbin, due to filtering out
Horizontal line between ± 10 °, therefore AngbinLess than or equal to 1 °；
4) to each horizontal line, its angle A ng is found out according to its k value, according to the angle and apart from i-th nearest of bin
Centric angle AngiDistance, vote i-th of bin and two bin adjacent with i, ballot value vote calculate it is as follows:
Wherein, b=i-1, i, i+1
5) Horz_Angle is determined after the ballot value of all horizontal lines adds up to each bin are as follows:
Further, two horizontal lines are horztAnd horzb, gradient is respectively less than 10 °, the angle of wherein at least one
The difference of degree and main level angle Horz_Angle are less than 5%；And horztAnd horzbDegree of overlapping cannot be small on the horizontal direction x
In wherein compared with the half of hyphen line length, it may be assumed that
Overlap(horzt,horzb)>0.5*min(lenhorzt,lenhorzb)
Left vertical line left and right vertical line right, the weight of the two in y-direction are selected from left and right vertical line concentration again
It is folded to spend the half for having to be larger than shorter vertical line length；Then calculate left, right respectively with horztAnd horzbFour intersection points
pti(x, y), i=1,2,3,4；Seek the solution of following equations group:
Y=kh·x+bh, h=horzt, horzb
X=kv·y+bv, v=left, right
By four intersection points since the upper left corner, it is named as pt counterclockwise1, pt2, pt3, pt4, form four sides
Shape Q；Above step is repeated, the quadrangle collection of various combination is generated.
Preferably, step S03 is specifically included:
Shape similarity Similarity calculating process is as follows:
1) different types of vehicle sample, four point pt at artificial selection vehicle windshield are selected1, pt2, pt3, pt4,
The ratio between following t2b and standard deviation sigma on averagely are calculated accordinglyt2b；The ratio between high and bottom edge h2b and standard deviation sigmah2b；Base angle mean value θ and mark
Quasi- difference σθ；
2) to each of quadrangle collection quadrangle Qi, calculating parameter lent、lenb、h1、h2, α, β, wherein lentFor
pt1And pt4The distance between, lenbFor pt2And pt3The distance between, h1For pt1To straight line pt2pt3Distance, h2For pt4To straight
Line pt2pt3The distance between, α is straight line pt1pt2With straight line pt2pt3Angle, β be straight line pt4pt3With straight line pt2pt3Folder
Angle, and calculate four characteristic value f1, f2, f3, f4；
3) according to four characteristic value f1, f2, f3, f4Calculate Similarity:
Similarity=e-(f1+f2+f3+f4)
Edge goodness of fit Fitness calculating process is as follows:
1) to horizontal line horztIt is above all to meet condition pt1.(x)≤x≤pt4.(x) point set { p (xi,yi) | i=1,2 ...
.Nt, according to horztStraight Line Fitting Parameters ktAnd bt, calculating parameter fitt:
In formula,
2) meet condition pt to all on left vertical line left1.(y)≤y≤pt2.(y) point set { p (xi,yi) | i=1,
2...Nl, according to the Straight Line Fitting Parameters k of leftlAnd bl, calculating parameter fitl:
3) horz is calculated by similar approachbWith the parameter fit of rightbAnd fitr；
4) according to fitt、fitl、fitbAnd fitrCalculate Fitness:
Fitness=(fit+fitl+fitb+fitr)/4
Finally, calculating Priority according to Similarity and Fitness:
Priority=α1*Similarity+α2* Fitness, wherein α1Take 0.7, α20.3 is taken,
Select the highest quadrangle of weight as image to be detected vehicle window elementary contour.
Preferably, the HSV histogram of target vehicle described in step S06 determines by the following method: one target of selection
The front image of vehicle, therefrom four points of selection determine its front window home position by hand；By the image in range through affine transformation
To the rectangular characteristic region of specified size, and the image in this feature region is divided into the subregion of M*N, while calculating each son
The HSV histogram in region.
The invention has the following advantages: the vehicle image search method utilizes image within the scope of vehicle windscreen
The region Fuzzy difference angle value of HSV histogram feature identifies similar image, gets rid of single by the unreliable of license number search
Property, improve the effect of vehicle image retrieval.
[Detailed description of the invention]
The present invention will be further explained below with reference to the attached drawings:
Fig. 1 is the flow chart based on vehicle window segmentation and the vehicle identification method of fuzzy feature；
The parameter schematic diagram of Fig. 2 quadrangle Q；
Fig. 3 calculates vehicle window region HSV histogram.
[specific embodiment]
Combined with specific embodiments below, and in conjunction with attached drawing, further description of the technical solution of the present invention:
Embodiment 1: in step S01, when the Canny operator of using area adaptive threshold seeks edge graph, Canny is calculated
Son needs to set two threshold values of height.Conventional method determines that this method uses region using artificial setting or the overall situation are adaptive
Adaptive approach threshold value simultaneously runs Canny operator, and steps are as follows:
1) 4*4 totally 16 pieces of subregions are divided an image into；
2) in each sub-regions, the accumulation grey level histogram in the region is calculated；
Wherein i is i-th of gray level, and range 0~255, I (g) is the number of pixels that gray scale is g in image, according to accumulation
Grey level histogram chooses high threshold ThhWith Low threshold Thl。
Thl=0.4*Thh
3) Canny operator is run according to region threshold.When certain subregion is arrived in the processing of Canny operator, using in the region
High-low threshold value is handled, and is handled in trans-regional boundary neighborhood using the threshold average in adjacent subarea domain.
Embodiment 2: to each of quadrangle collection quadrangle Q, quadrangle and car bumper wind glass sample parameter are calculated
Shape similarity Similarity and edge goodness of fit Fitness, calculates the weight Priority of the quadrangle accordingly；
Shape similarity Similarity reflects found quadrangle Q and the similar journey of windshield shape in sample
Degree.Specific calculating process is as follows:
1) selecting 100 includes the different types of vehicle samples such as car, minibus, lorry, SUV, artificial selection vehicle
Four point pt at windshield1, pt2, pt3, pt4, averagely above the ratio between following t2b=0.7716, standard deviation sigma are calculated accordinglyt2b
=0.0245；The ratio between high and bottom edge h2b=0.3570, standard deviation sigmah2b=0.0346；Base angle mean value θ=72.12, standard deviation sigmaθ
=2.5373.
2) to each of quadrangle collection quadrangle Qi, its parameter as shown in Figure 2 is calculated, and calculate four characteristic values
f1, f2, f3, f4:
3) according to four characteristic value f1, f2, f3, f4Calculate Similarity:
Similarity=e-(f1+f2+f3+f4)
Edge goodness of fit Fitness reflects the four edges of Q and participates in four line horz of fitting Qt, horzb, left and
Laminating degree of the right on image space positions, calculating process are as follows:
1) to horizontal line horztIt is above all to meet condition pt1. (x)≤x≤pt4. (x) point set { p (xi,yi) | i=1,
2….Nt, according to horztStraight Line Fitting Parameters ktAnd bt, calculating parameter fitt:
In formula,
2) meet condition pt to all on left vertical line left1.(y)≤y≤pt2.(y) point set { p (xi,yi) | i=1,
2...Nl, according to the Straight Line Fitting Parameters k of leftlAnd bl, calculating parameter fitl:
3) horz is calculated by similar approachbWith the parameter fit of rightbAnd fitr；
4) according to fitt、fitl、fitbAnd fitrCalculate Fitness:
Fitness=(fit+fitl+fitb+fitr)/4
Finally, calculating Priority according to Similarity and Fitness:
Priority=α1*Similarity+α2* Fitness, wherein α1Take 0.7, α20.3 is taken, because according to statistics, shape
The contribution that similarity is fitted window edge is bigger.
It selects the highest quadrangle Q of weight Priority as windshield elementary contour, which is existed
Characteristic area of the part through affine transformation to uniform sizes in elementary contour, and it is divided into the subregion of M*N (12*8).Then
Image is transformed into HSV space by rgb space, is divided into 36 bin according to the channel H, channel S is divided into 10 bin, and the channel V is divided into
16 bin, and all bin mode for normalizing to [0,1] section is calculated to the hsv color histogram of each subregion, as
The preceding face image feature of the vehicle.As shown in Figure 3.
Above-mentioned specific embodiment is used to illustrate the present invention, rather than limits the invention, of the invention
In spirit and scope of protection of the claims, to any modifications and changes that the present invention makes, protection model of the invention is both fallen within
It encloses.
Claims (7)
1. the vehicle identification method based on vehicle window segmentation and fuzzy feature, which comprises the following steps:
S01 obtains the edge graph of image to be detected；
S02 carries out straight line fitting, different straight line groups close to horizontal ledgement and subvertical vertical moulding to edge graph
Synthesize different quadrangles；
S03 using with target vehicle vehicle window sample shape similarity and the highest quadrangle of the edge goodness of fit as image to be detected
The elementary contour of vehicle window；
S04 determines the HSV histogram of image to be detected vehicle window, specifically includes:
Using the highest quadrangle of weight as image to be detected vehicle window elementary contour, the image in profile, which is carried out affine transformation, is
Regular rectangular shape calculates the HSV of each subregion as characteristic area, the subregion for being M*N by this feature region even partition
Color histogram, steps are as follows:
1) rule by the image in the quadrangle for representing image to be detected vehicle window elementary contour by affine transformation to uniform sizes
Then rectangle；
2) the uniform subregion R for being M*N by this feature region divisionm,n, wherein m≤M, n≤N；
3) image is transformed into hsv color space from RGB color；
4) all subregion R is calculatedm,nThe histogram Hist in the middle channel H, S, Vm,n H, Histm,n S, Histm,n VAll bin in histogram
Normalize to [0,1] section；
The HSV histogram of S05 and target vehicle vehicle window sample carries out region Fuzzy difference angle value and calculates；
The HSV histogram of image in vehicle database and target vehicle in the same way, is carried out region Fuzzy one by one by S06
Difference angle value calculates；
S07 is worth search result according to region Fuzzy diversity factor；
The preceding face image feature of image and target vehicle to be checked in vehicle database is carried out region Fuzzy one by one by the method
Difference angle value calculates, and retrieves most like several images as search result, calculation method according to region Fuzzy difference angle value
It is as follows:
1) to each of vehicle characteristics region to be checked subregion Rm,n T, calculate its son corresponding with target vehicle characteristic area
Region Rm,n SFuzzy diversity factor, the channel difference angle value of each corresponding sub-region are as follows:
Wherein c indicates Histm,n H, Histm,n S, Histm,n VThree channel histograms, bmaxIndicate the maximum bin under the histogram
Number, Δb,kIndicate that b, the difference of two bin of k, formula ensure that when two histograms are in the difference of the same bin or adjacent bin
When smaller, each channel difference angle value Fm,n CAlso smaller；
According to each channel difference degree, the Fuzzy difference angle value of the corresponding sub-region is calculated are as follows:
Fm,n=αH*Fm,n H+αS*Fm,n S+αV*Fm,n V
Wherein αcFor the constant factor in each channel；
2) according to the diversity factor F of corresponding sub-regionm,n, zoning diversity factor F；
It is identified as target vehicle with the region lesser several images of Fuzzy difference angle value of target vehicle in vehicle to be measured.
2. the vehicle identification method according to claim 1 based on vehicle window segmentation and fuzzy feature, it is characterised in that: step
Edge graph described in rapid S01 is that the Canny operator of using area adaptive threshold is found out.
3. the vehicle identification method according to claim 1 based on vehicle window segmentation and fuzzy feature, which is characterized in that step
Rapid S02 is specifically included: search is close to horizontal ledgement and subvertical vertical moulding in edge graph, and gathers to lines
It closes and screens；Straight line fitting is carried out to horizontal line, filters out the high horizontal line of degree of fitting, and calculate main level tilt angle；To perpendicular
Line carries out screening and straight line fitting；According to Straight Line Fitting Parameters, from two horizontal lines of horizontal line Resource selection, and from left and right vertical line collection
In respectively select a line group to be combined into a quadrangle, different combinations constitutes quadrangle collection.
4. the vehicle identification method according to claim 3 based on vehicle window segmentation and fuzzy feature, it is characterised in that: right
Horizontal line straight line fitting filters out the high horizontal line of degree of fitting, and during calculating main level tilt angle, using the side of ballot
Formula determines main horizontal tilt angle Horz_Angle, and voting process is as follows:
1) least square method is used, according to the point sequence fitting a straight line y=kx+b of each horizontal line, and digital simulation degree is sigma
It is as follows, wherein xs, ysFor the point in horizontal line point sequence, len is horizontal line length；
Y=kxs+b
Sigma=∑ (y-ys)2/len
2) according to the parameter k of fitting a straight line, all fitting a straight lines of the level angle within the scope of positive and negative 10 ° are found out, and determine angle
It spends range [Ang_Min, Ang_Max], the parameter k is straight slope；
3) [Ang_Min, Ang_Max] is divided into 20 bin, each bin angular span is Angbin, due to the cross filtered out
Line is between ± 10 °, therefore AngbinLess than or equal to 1 °；
4) to each horizontal line, its angle A ng is found out according to its k value, according to the angle and apart from i-th nearest of center bin
Angle A ngiDistance, vote i-th of bin and two bin adjacent with i, ballot value vote calculate it is as follows:
Wherein, b=i-1, i, i+1
5) Horz_Angle is determined after the ballot value of all horizontal lines adds up to each bin are as follows:
5. the vehicle identification method according to claim 4 based on vehicle window segmentation and fuzzy feature, it is characterised in that: institute
Stating two horizontal lines is horztAnd horzb, gradient is respectively less than 10 °, the angle of wherein at least one and main level angle Horz_
The difference of Angle is less than 5%；And horztAnd horzbDegree of overlapping cannot be less than wherein compared with hyphen line length on the horizontal direction x
Half, it may be assumed that
Overlap(horzt,horzb)>0.5*min(lenhorzt,lenhorzb)
Left vertical line left and right vertical line right, the degree of overlapping of the two in y-direction are selected from left and right vertical line concentration again
Have to be larger than the half of shorter vertical line length；Then calculate left, right respectively with horztAnd horzbFour intersection point pti
(x, y), i=1,2,3,4；Seek the solution of following equations group:
Y=kh·x+bh, h=horzt, horzb
X=kv·y+bv, v=left, right
By four intersection points since the upper left corner, it is named as pt counterclockwise1, pt2, pt3, pt4, form a quadrangle Q；Weight
Multiple above step, generates the quadrangle collection of various combination.
6. the vehicle identification method according to claim 5 based on vehicle window segmentation and fuzzy feature, which is characterized in that step
Rapid S03 is specifically included:
Shape similarity Similarity calculating process is as follows:
1) different types of vehicle sample, four point pt at artificial selection vehicle windshield are selected1, pt2, pt3, pt4, calculate accordingly
The ratio between following t2b and standard deviation sigma are averagely gone up outt2b；The ratio between high and bottom edge h2b and standard deviation sigmah2b；Base angle mean value θ and standard deviation
σθ；
2) to each of quadrangle collection quadrangle Qi, calculating parameter lent、lenb、h1、h2, α, β, wherein lentFor pt1With
pt4The distance between, lenbFor pt2And pt3The distance between, h1For pt1To straight line pt2pt3Distance, h2For pt4To straight line
pt2pt3The distance between, α is straight line pt1pt2With straight line pt2pt3Angle, β be straight line pt4pt3With straight line pt2pt3Folder
Angle, and calculate four characteristic value f1, f2, f3, f4；
3) according to four characteristic value f1, f2, f3, f4Calculate Similarity:
Similarity=e-(f1+f2+f3+f4)
Edge goodness of fit Fitness calculating process is as follows:
1) to horizontal line horztIt is above all to meet condition pt1·(x)≤x≤pt4·(x) point set { p (xi,yi) | i=1,2 ... .Nt,
According to horztStraight Line Fitting Parameters ktAnd bt, calculating parameter fitt:
In formula,
2) meet condition pt1. (y)≤y≤pt2 (y) point set { p (x to all on left vertical line lefti,yi) | i=1,
2...Nl, according to the Straight Line Fitting Parameters k of leftlAnd bl, calculating parameter fitl:
3) horz is calculated by similar approachbWith the parameter fit of rightbAnd fitr；
4) according to fitt、fitl、fitbAnd fitrCalculate Fitness:
Fitness=(fit+fitl+fitb+fitr)/4
Finally, calculating Priority according to Similarity and Fitness:
Priority=α1*Similarity+α2* Fitness, wherein α1Take 0.7, α20.3 is taken,
Select the highest quadrangle of weight as image to be detected vehicle window elementary contour.
7. the vehicle identification method according to claim 1 based on vehicle window segmentation and fuzzy feature, which is characterized in that step
The HSV histogram of target vehicle described in rapid S06 determines by the following method: the front image of one target vehicle of selection, therefrom
Four points of selection determine its front window home position by hand；Rectangular characteristic by the image in range through affine transformation to specified size
Region, and the image in this feature region is divided into the subregion of M*N, while calculating the HSV histogram of all subregion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510148453.3A CN104881630B (en) | 2015-03-31 | 2015-03-31 | Vehicle identification method based on vehicle window segmentation and fuzzy feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510148453.3A CN104881630B (en) | 2015-03-31 | 2015-03-31 | Vehicle identification method based on vehicle window segmentation and fuzzy feature |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104881630A CN104881630A (en) | 2015-09-02 |
| CN104881630B true CN104881630B (en) | 2018-12-04 |
Family
ID=53949118
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510148453.3A Active CN104881630B (en) | 2015-03-31 | 2015-03-31 | Vehicle identification method based on vehicle window segmentation and fuzzy feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104881630B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106340198A (en) * | 2016-08-31 | 2017-01-18 | 广州地理研究所 | Parking lot parking stall guiding system and guiding method thereof |
| CN111063016A (en) * | 2019-12-31 | 2020-04-24 | 螳螂慧视科技有限公司 | Multi-depth lens face modeling method and system, storage medium and terminal |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101609452A (en) * | 2009-07-10 | 2009-12-23 | 南方医科大学 | The fuzzy SVM feedback that is used for the medical image target identification is estimated method |
| CN103854290A (en) * | 2014-03-25 | 2014-06-11 | 中国科学院光电技术研究所 | Extended target tracking method combining skeleton characteristic points and distribution field descriptors |
| CN103927512A (en) * | 2014-03-11 | 2014-07-16 | 浙江工商大学 | Vehicle identification method |
-
2015
- 2015-03-31 CN CN201510148453.3A patent/CN104881630B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101609452A (en) * | 2009-07-10 | 2009-12-23 | 南方医科大学 | The fuzzy SVM feedback that is used for the medical image target identification is estimated method |
| CN103927512A (en) * | 2014-03-11 | 2014-07-16 | 浙江工商大学 | Vehicle identification method |
| CN103854290A (en) * | 2014-03-25 | 2014-06-11 | 中国科学院光电技术研究所 | Extended target tracking method combining skeleton characteristic points and distribution field descriptors |
Non-Patent Citations (1)
| Title |
|---|
| 医学图像的特征自动提取及基于模糊特征的图像检索研究;江少锋;《中国博士学位论文全文数据库》;20090615(第6期);第48-80页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104881630A (en) | 2015-09-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zaklouta et al. | Real-time traffic sign recognition in three stages | |
| CN111126224A (en) | Vehicle detection method and classification recognition model training method | |
| CN105335702B (en) | A kind of bayonet model recognizing method based on statistical learning | |
| CN103942560B (en) | A kind of high-resolution video vehicle checking method in intelligent traffic monitoring system | |
| US7561721B2 (en) | System and method for range measurement of a preceding vehicle | |
| CN109635733B (en) | Parking lot and vehicle target detection method based on visual saliency and queue correction | |
| CN105740886B (en) | A kind of automobile logo identification method based on machine learning | |
| CN102663357A (en) | Color characteristic-based detection algorithm for stall at parking lot | |
| CN103258432A (en) | Traffic accident automatic identification processing method and system based on videos | |
| CN107194393B (en) | Method and device for detecting temporary license plate | |
| CN106257490A (en) | The method and system of detection driving vehicle information | |
| CN109214345A (en) | The method for searching the driving trace for changing board vehicle is compared based on similarity | |
| CN105005989A (en) | Vehicle target segmentation method under weak contrast | |
| CN102902983B (en) | A kind of taxi identification method based on support vector machine | |
| Prates et al. | Brazilian license plate detection using histogram of oriented gradients and sliding windows | |
| CN106845444A (en) | A kind of vehicle well cover detection method combined based on acnode | |
| Islam et al. | An efficient method for extraction and recognition of bangla characters from vehicle license plates | |
| CN103927512B (en) | vehicle identification method | |
| CN105160292A (en) | Vehicle identification recognition method and system | |
| CA2815591A1 (en) | Method for detecting and recognising an object in an image, and an apparatus and a computer program therefor | |
| CN110321855A (en) | A kind of greasy weather detection prior-warning device | |
| CN104881630B (en) | Vehicle identification method based on vehicle window segmentation and fuzzy feature | |
| CN105975949A (en) | Visual-information-based automobile identification method | |
| CN110619255B (en) | Target detection method and device | |
| Hossen et al. | Vehicle license plate detection and tilt correction based on HSI color model and SUSAN corner detector |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20191227 Address after: 314413 No.2, Fengshou Avenue, Haining Economic and knitting industrial park, Jiaxing City, Zhejiang Province Patentee after: Zhejiang Haining Warp Knitting Industrial Park Development Co., Ltd Address before: Hangzhou City, Zhejiang province 310018 Xiasha Higher Education Park is 18 street. Patentee before: Zhejiang Gongshang University |
|
| TR01 | Transfer of patent right |