CN103077384B - A kind of method and system of vehicle-logo location identification - Google Patents
A kind of method and system of vehicle-logo location identification Download PDFInfo
- Publication number
- CN103077384B CN103077384B CN201310009960.XA CN201310009960A CN103077384B CN 103077384 B CN103077384 B CN 103077384B CN 201310009960 A CN201310009960 A CN 201310009960A CN 103077384 B CN103077384 B CN 103077384B
- Authority
- CN
- China
- Prior art keywords
- car mark
- vehicle
- mark
- car
- location
- 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
Abstract
The invention discloses a kind of vehicle-logo location and know method for distinguishing, including: utilize the headstock image of high-definition camera collection vehicle；Utilize described headstock image, carry out vehicle location and plate location recognition, and according to the prior information of car mark Yu car plate, generate car mark and estimate the band of position；Estimate in the band of position at car mark, the interference of car mark background area is filtered according to the texture information extracted, binaryzation is carried out on car mark texture gray-scale map after filtering the interference of car mark background area, the binary image generated carries out car mark by searching connected domain method in car mark estimates the band of position be accurately positioned, generate car mark precise position information；According to car mark precise position information, extract car mark accurate greyscale figure, and car mark accurate greyscale figure is carried out binaryzation, generate the binary map that car mark is corresponding；Binary map corresponding for car mark being mated with the car mark template prestored, be weighted according to positional information, car mark type corresponding to calculated mark soprano is vehicle-logo recognition result.
Description
Technical field
The present invention relates to vehicular traffic fixation and recognition field, particularly relate to a kind of vehicle-logo location recognition methods and system.
Background technology
Car mark is the significant image of vehicle, not only contains vehicle information, further comprises the information of manufacturer, heavier
If it is difficult to change.If car mark information combined with license board information, the reliability of vehicle identification can be improved greatly.?
In some criminal activity, although car plate is readily replaceable, but car mark is unique, and vehicle-logo recognition technology can be widely applied to traffic thing
Therefore process, vehicles peccancy monitors automatically, automatic charging, airport, harbour, parking lot, the automatic safe management of the vehicle such as community, beat
Hitting the fields such as vehicle crime, therefore vehicle recongnition technique in intelligent transportation is improved and develops and has important reason by vehicle-logo recognition
Opinion meaning and using value.
Can effectively detect car mark at present, the Main Means positioning and identifying is method based on video, but
Video method is utilized to carry out the problem of vehicle-logo location identification existence mainly:
1, self background of vehicle image and the impact of vehicle body background；
2, same vehicle exists multiple car mark, and car mark is affected seriously by weather and colour of sky change；
3, car target is different and size difference plot big, and image blurring, position is not quite similar；
4, the radiator-grid texture at car mark place is varied, and radiator-grid height is uneven；
5, there is a large amount of interference around car mark, such as car light, bumper etc..Car mark is on car body and some and body color
Similar.
The existence of the problems referred to above so that the accuracy that use video method carries out car target fixation and recognition is the highest, and mesh
Before also do not have a kind of detection equipment car mark can be accurately positioned and effectively identify.
Summary of the invention
The present invention provides the method and system of a kind of vehicle-logo location identification, accurate to solve the method for existing video location
Spend the highest problem.
In order to achieve the above object, the embodiment of the invention discloses a kind of vehicle-logo location and know method for distinguishing, including: utilize height
The headstock image of clear camera acquisition vehicle；Utilize described headstock image, carry out vehicle location and plate location recognition, and according to
Car mark and the prior information of car plate, generate car mark and estimate the band of position；Estimate in the band of position at described car mark, according to extract
Texture information filters the interference of car mark background area, and the car mark texture gray-scale map after filtering the interference of car mark background area carries out two
Value, carries out car mark essence by searching connected domain method on the binary image generated in described car mark estimates the band of position
Determine position, generate car mark precise position information；According to described car mark precise position information, extract car mark accurate greyscale figure, and will
Described car mark accurate greyscale figure carries out binaryzation, generates the binary map that described car mark is corresponding；By binary map corresponding for described car mark
Mating with the car mark template prestored, be weighted according to positional information, calculated mark soprano is corresponding
Car mark type is vehicle-logo recognition result.
In order to achieve the above object, the embodiment of the invention also discloses the system of a kind of vehicle-logo location identification, including: image
Harvester, for the headstock image of collection vehicle；Car mark estimates band of position generating means, is used for utilizing described headstock figure
Picture, carries out vehicle location and plate location recognition, and according to the prior information of car mark Yu car plate, generates car mark and estimate lane place
Territory；Car mark precise position information generating means, for estimating in the band of position at described car mark, according to the texture information filter extracted
Except car mark background area is disturbed, the car mark texture gray-scale map after filtering the interference of car mark background area carries out binaryzation, is giving birth to
In described car mark estimates the band of position, carry out car mark by searching connected domain method on the binary image become to be accurately positioned, raw
Become car mark precise position information；Car mark binary map generating means, for according to described car mark precise position information, extracts car mark essence
Really gray-scale map, and described car mark accurate greyscale figure is carried out binaryzation, generate the binary map that described car mark is corresponding；Vehicle-logo recognition is tied
Really generating means, for binary map corresponding for described car mark being mated with the car mark template prestored, enters according to positional information
Row weighted calculation, car mark type corresponding to calculated mark soprano is vehicle-logo recognition result.
The mark positioning identifying method of the present invention and system, can be accurately positioned car mark and effectively identify.
Accompanying drawing explanation
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
In having technology to describe, the required accompanying drawing used is briefly described, it should be apparent that, the accompanying drawing in describing below is only this
Some embodiments of invention, for those skilled in the art, on the premise of not paying creative work, it is also possible to root
Other accompanying drawing is obtained according to these accompanying drawings.
Fig. 1 is the flow chart of the vehicle-logo location recognition methods of the embodiment of the present invention；
Fig. 2 is the method flow diagram of the sub-step of step S102 in embodiment illustrated in fig. 1；
Fig. 3 is that certain two field picture gathered is carried out vehicle location, plate location recognition, generates of car mark discreet area
The result schematic diagram of specific embodiment；
Fig. 4 is the method flow diagram of the sub-step of step S103 in embodiment illustrated in fig. 1；
Fig. 5 is the schematic diagram of the cross grain in the car mark discreet area in embodiment illustrated in fig. 4；
Fig. 6 is the schematic diagram of the longitudinal texture in the car mark discreet area in embodiment illustrated in fig. 4；
Fig. 7 be embodiment illustrated in fig. 6 is pre-processed after generate binary map；
Fig. 8 is on the basis of the binary map of Fig. 7, carries out longitudinal projection's figure of longitudinal projection；
Fig. 9 is the final pinpoint result schematic diagram of car mark of embodiment illustrated in fig. 8；
Figure 10 is the method flow diagram of the sub-step of step S104 in embodiment illustrated in fig. 1；
Figure 11 is the method flow diagram of the sub-step of step S105 in embodiment illustrated in fig. 1；
Figure 12 is the vehicle-logo recognition result positioning out in embodiment illustrated in fig. 9；
Figure 13 be the vehicle-logo location recognition methods of the present invention the method flow diagram of another embodiment；
Figure 14 is the structural representation of the vehicle-logo location identification system of the embodiment of the present invention；
Figure 15 is the structural representation that the car mark in embodiment illustrated in fig. 14 estimates band of position generating means 101；
Figure 16 is the structural representation of the car mark precise position information generating means 102 in embodiment illustrated in fig. 14；
Figure 17 is the structural representation of the car mark binary map generating means 103 in embodiment illustrated in fig. 14；
Figure 18 is the structural representation of the vehicle-logo recognition result generating means 104 in embodiment illustrated in fig. 14；
Figure 19 is the structural representation of another embodiment of the vehicle-logo location identification system of the present invention；
Figure 20 a, b, c, d are vehicle-logo location recognition methods and the system utilizing the present invention, carry out the car mark of several vehicles
The handling process schematic diagram identified.
Detailed description of the invention
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Describe, it is clear that described embodiment is only a part of embodiment of the present invention rather than whole embodiments wholely.Based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under not making creative work premise
Embodiment, broadly falls into the scope of protection of the invention.
The embodiment of the present invention provides a kind of vehicle-logo location recognition methods and system, on the premise of there is effective board in vehicle,
Can quickly orient car cursor position, and carry out single frames identification, many frame acknowledgments, determine vehicle and vehicle by car plate and car mark
Uniqueness, report background system, vehicle is carried out registration confirm.
Fig. 1 is the flow chart of the vehicle-logo location recognition methods of the embodiment of the present invention.As it can be seen, the car of the present embodiment is demarcated
Position recognition methods includes:
Step S101, utilizes the headstock image of high-definition camera collection vehicle；Step S102, utilizes described headstock image,
Carry out vehicle location and plate location recognition, and according to the prior information of car mark Yu car plate, generate car mark and estimate the band of position；Step
Rapid S103, estimates in the band of position at described car mark, filters the interference of car mark background area according to the texture information extracted, is filtering
Binaryzation is carried out, by the company of searching on the binary image generated on car mark texture gray-scale map after the interference of car mark background area
Logical territory method carries out car mark in described car mark estimates the band of position and is accurately positioned, and generates car mark precise position information；Step
S104, according to described car mark precise position information, extracts car mark accurate greyscale figure, and described car mark accurate greyscale figure is carried out two
Value, generates the binary map that described car mark is corresponding；Step S105, by binary map corresponding for described car mark and the car mark template prestored
Mating, be weighted according to positional information, car mark type corresponding to calculated mark soprano is car mark
Other result.
In step S101 of the present embodiment, rolling end away from expressway, install high-definition camera, capture enters the car in the visual field
Head image, 200W scene frame lower each second 15, image size is to be frame each second 7 under 1600*1200,500W scene, and size is
2432*2048, image resolution ratio can be adjusted according to the difference of high-definition camera.In the following step of the present embodiment, it is
The each two field picture taked in step S101 is processed, generates the vehicle-logo recognition result of corresponding each frame.
In step S102 of the present embodiment, as in figure 2 it is shown, utilize described headstock image, carry out vehicle location and car plate
Fixation and recognition, and according to the prior information of car mark Yu car plate, generates car mark and estimates the step of the band of position and include following sub-step:
Step S1021, is compressed in proportion to the headstock image of described collection, generates gray level image；Step S1022, at described ash
Set up background according to Gauss model on degree image, carry out vehicle location according to background subtraction and frame difference, generate vehicle location region；Step
Rapid S1023, in described vehicle location region, texture information and colouring information according to vehicle carry out plate location recognition, raw
Become the car plate band of position；Step S1024, in the described car plate band of position, enters according to the prior information of described car mark with car plate
Row vehicle-logo location, generates car mark discreet area.
In step S1021, after receiving the headstock image that high-definition camera gathers, in order to reduce complexity, to colour
Image is compressed in proportion, carries out dimensionality reduction simultaneously and extracts gray level image as the input followed by vehicle-logo location identification.
In step S1022, gray level image sets up background according to Gauss model, find connected region according to background subtraction
Carry out vehicle location, adjacent three frames do poor " take with " and the vehicle carrying out out is confirmed, thus generate vehicle location district
Territory.
In the present embodiment, present frame gray figure and background gray-scale map do difference, to background subtraction gray-scale map binaryzation, find
White point connected domain, area just it was initially believed that vehicle region more than certain threshold value.Because it is poor that adjacent two frames do the poor frame obtained
Figure can cause vehicle motion blur phenomenon, is elongated vehicle region, introduces interference, so using continuous three frames, adjacent two frames two
Two do frame difference then binaryzation, two the frame difference binary map obtained are taken with, on i.e. two width frame difference figures, same position has frame poor
Information be considered available point, obtain the final frame difference binary map after " take with ", find connected domain and be vehicle region, logical
Cross the vehicle that frame difference positions out to confirm with the vehicle positioning out by background subtraction, obtain final vehicle location.
In step S1023, in described vehicle location region, carry out corresponding vehicle according to texture information and colouring information
Plate location recognition, generate the car plate band of position.
In the present embodiment, in vehicle location region, cromogram carries out gray processing, then asks cross grain, common car
Board has 7 characters, and cross grain has 14 saltus steps, so saltus step is more than in texture maps, searching meets car plate width range
The UNICOM region of 10, is license plate candidate area.Carrying out license plate locating method by colouring information is, the mainly blue board of car plate and
Yellow card, in the range of the coloured picture in vehicle location region, extracts blue dot and yellow dots information, meets blue dot connected domain or Huang
What color dot connected domain size met car plate rule is candidate license plate, and candidate license plate gray-scale map is carried out binaryzation, character
Cutting, monocase stencil matching identification, a Mean match degree sought in seven characters of car plate, and Mean match degree is the highest and reaches to set
That determines recognition threshold is real car plate, and real car plate gray-scale map is carried out binaryzation, according to binary map projection and connected domain side
Method carries out the accurate cutting of character, obtains seven character lower edges positions accurately, according to the lower edges matching one of character
Bar straight line contrasts with horizontal line, thus draws license plate sloped angle.
In the present embodiment, also include by Character segmentation and slightly identify, finding optimum car plate in License Plate and then enter
Row accurately identifies, if recognition confidence is less than the recognition threshold preset, thinks that this vehicle is tied without effective board, vehicle location
Fruit likely has deviation, without effective car plate and car mark region in vehicle region, otherwise it is assumed that this vehicle location is effective.
In step S1024, in the described car plate band of position, carry out car according to the prior information of described car mark Yu car plate
Demarcate position, generate car mark discreet area.In the present embodiment, on the basis of plate location recognition, according to the position of car mark Yu car plate
The priori such as confidence breath, car mark shape, vehicle symmetry can estimate the region at car mark place, i.e. car mark discreet area.
In the present embodiment, step S1021-1024 is the method flow estimating car mark region.With reference to Fig. 3, it is right
Certain two field picture gathered carries out vehicle location, plate location recognition, generates a specific embodiment of car mark discreet area.Its
In, 1. dashed box is the vehicle region that step S1022 positions out according to background subtraction and frame difference, and 2. dashed box is that step S1023 is at car
Band of position dashed box 1. on the basis of the car plate of fixation and recognition estimate the band of position, and 3. dashed box is to find in step S1023
The car plate band of position that effectively bridge queen finally positions, then according to car mark and the positional information of car plate, car mark shape, vehicle symmetry
Property etc. priori, searched out dashed box dashed box 3. above 4. for car mark discreet area, this dashed box 4. in carry out car demarcation
Position.
In step S103 of the present embodiment, as shown in Figure 4, estimate in the band of position at described car mark, according to extract
Texture information filters the interference of car mark background area, and the car mark texture gray-scale map after filtering the interference of car mark background area carries out two
Value, carries out car mark essence by searching connected domain method on the binary image generated in described car mark estimates the band of position
Determine position, generate car mark precise position information, realized by following sub-step: step S1031, estimate position at described car mark
Sobel cross grain and longitudinal texture is extracted in region；Step S1032, compares the gray-scale map after extracting cross grain and extraction
The average gray of the gray-scale map after longitudinal texture, determines, by both differences, the texture side that described car mark background area is disturbed
To；Step S1033, extracts the texture image mutually anti-phase with the grain direction of described car mark background area interference, generates car mark pre-
Estimate the car mark texture gray-scale map in the band of position；Step S1034, carries out binary conversion treatment to described car mark veining gray-scale map,
And carry out morphologic filtering and corrosion expansion, generate the binary map that described car mark texture gray-scale map is corresponding；Step S1035, in institute
Project on the basis of stating the binary map that car mark veining gray-scale map is corresponding, find the right boundary that car mark is corresponding, generate car mark
Precise position information.
In step S1031, preceding step 102 have found and comprised car target rough position region, i.e. as in Fig. 3
Dashed box 4. inner region, 4. shown in inner region, is extracted sobel cross grain and longitudinal direction on the basis of this area grayscale figure by dashed box
Texture, shown in below figure 5 and Fig. 6.Wherein, Fig. 5 is the cross grain in car mark discreet area, in Fig. 6 is car mark discreet area
Longitudinal texture.
In step S1032 and step S1033, extracted the gray scale of the gray-scale map after texture respectively by comparison diagram 5 and Fig. 6
Mean value, if both differences are more than the texture threshold set, (texture threshold typically takes 15, certainly, can be according to actual conditions
Other numerical value), then it is assumed that car mark background texture is cross grain, so the longitudinal texture that should extract in car mark discreet area,
Filter heatsink transverse window interference etc., the most remaining car mark texture information.If after transverse and longitudinal background texture is extracted in car mark discreet area,
If longitudinal texture mean value is much larger than horizontal mean value, then it is assumed that background is longitudinal texture, it is necessary to extract cross grain, filter
Car mark information only it is left, if but cross grain is close, then with longitudinal texture average except longitudinal interference of texture of background
Background is considered reticular texture, now will extract Texture by sobel filter operator.That is: extract and described car mark background
The mutually anti-phase texture image of grain direction of region interference, to generate the car mark texture gray-scale map that car mark is estimated in the band of position.
Texture blending herein, compare and technology that the computational methods of sobel filter operator are known to the skilled person, so place
Repeat no more.
In step S1034, the gray level image of the rough filtering interfering of Fig. 6 finds the accurate position of car mark through pretreatment
Putting, seek the binary-state threshold of the gray-scale map shown in Fig. 6 by the method for OTSU, the light and shade caused in order to avoid light, shade etc. is not
Equal phenomenon, and the impact of car mark background area reflector lamp, the binaryzation threshold obtained with the mean value correction OTSU of gray-scale map
Value, carries out binary conversion treatment to Fig. 6, carries out morphologic filtering, corrosion expansion the most again, eliminates horizontal long line etc. and process, and Fig. 6 is pre-
Effect after process is as shown in Figure 7.It will be seen that gray level image is the most binarized for binary map.
In step S1035, project on the basis of the binary map that described car mark veining gray-scale map is corresponding, find car
The right boundary that mark is corresponding, generates car mark precise position information.In the embodiment shown in Fig. 2-Fig. 7, in the binary map of Fig. 7
On the basis of, carry out longitudinal projection, and ask for longitudinal projection mean value nAvg, the certain proportion k averaged as find about
Threshold value nTh=k*nAvg on border, on the basis of car plate axis, on transverse projection figure, limit, car mark left and right is found in left and right expansion
Boundary, when continuous one pinpoints ratio less than the threshold value finding border, then it is assumed that border terminates, and otherwise the node from new connection is clear
Zero breaks, left and right is found the most always, and Fig. 7 longitudinal projection schemes as shown in Figure 8, after determining car target right boundary, root
Carrying out transverse projection according to same method, to find car target up-and-down boundary, the final pinpoint result of car mark is as in the middle of Fig. 9
Dashed box is 5. shown.
In step S104, as shown in Figure 10, according to described car mark precise position information, extract car mark accurate greyscale figure,
And described car mark accurate greyscale figure is carried out binaryzation, generate the step of binary map corresponding to described car mark, including following sub-step
Rapid: step S1041, according to the license plate sloped angle generated in described License Plate, described car mark accurate greyscale figure is inclined
Tiltedly correction；Step S1042, is carried out at binaryzation according to the method for bilinearity OTSU the car mark accurate greyscale figure after described correction
Reason, generates the binary map that described car mark is corresponding.
In the present embodiment, the car mark precise position information be given according to vehicle-logo location, extracts car mark gray-scale map, by above
The license plate sloped angle carrying out calculating in plate location recognition, car mark region is carried out slant correction, then normalizes
For the figure of 40*40, in order to prevent the light and shade caused because of interference effects such as light inequality, the shadow of the trees uneven, car mark gray-scale map is pressed
Method according to bilinearity OTSU carries out binaryzation.
In step S105, as shown in figure 11, binary map corresponding for described car mark and the car mark template prestored are carried out
The step joined, including: step S1051, carry out outline by the car mark in the binary map that hough conversion is corresponding to described car mark
Matching；Step S1052, classifies to described car mark according to fitting result；Step S1053, selects and the described corresponding class of car mark
The car mark template of type is mated.
In the present embodiment, on the basis of binary map, by hough conversion, car mark is carried out outline matching, simultaneously basis
The prioris such as car mark depth-width ratio shape, are the class in oval, circle and irregular figure by this car mark rough segmentation, then select
The car mark masterplate of the type mates.Hough conversion is the method for fitting circle in binary map, it may be assumed that according to car plate and car mark
Relation, it can be deduced that car mark least radius and maximum radius, the radius minimum of a value of fitting circle and maximum, according to symmetry
Property, with central point as central coordinate of circle in the binary map of car mark region, scan circumference according to radius, calculate white point on circumference
Number, when white point number and girth ratio are during higher than certain threshold value, it is believed that effectively, this car is designated as rounded outer profiles car mark to this circle, otherwise
Point based on car mark regional center point, with two times of least radius and maximum radius for the long scope of transverse or square
Side size range run-down, it is judged that meet square or oval point and account for the ratio of girth, just recognize during higher than certain threshold value
For being oval or square-shaped outer profile car mark, if matching circular, oval, square was all lost efficacy, think and belong to
Irregular type car mark.
Car mark masterplate is the representative car mark two extracted according to the principle that between class, difference is maximum, class interpolation is maximum
Value figure, size is 40*40, in order to save space complexity and time complexity, binary map eight is synthesized a byte, 40*
The masterplate boil down to 40*5 of 40 sizes, the car of current each type indicates three templates, and number can be carried out according to actual conditions
Adjust.
When carrying out car mark template matches, it is contemplated that location is difficult to be accurate to the location of pixels on car mark border, and
Be likely to certain border interference information when choosing masterplate, thus use shiding matching method, car mark binary map upper,
Under, the slip of 1 to 2 pixels can be carried out on left and right four direction, blank parts mends 0 alignment, then by binary map eight conjunction
Become a byte to mate with template, it is contemplated that same type of car mark, such as Ford and BYD, Kia, popular, Buick with
Benz etc., outline is identical is all circular or oval, and only inside center information is different, so when stencil matching, root
Being weighted according to positional information, in the car mark binary map of 40*40,8 layers of weight of outermost are 1, and middle 16 layers of weight are 2, center 16
Layer weight is 3, if not matching as zero, is set to 1 point, 2 points, 3 points respectively according to position weight difference after the match is successful,
After be 0 100 points by score normalization, by all sort results after shiding matching, choose mark the highest for final single frames
Vehicle-logo recognition result, arranges single frames vehicle-logo recognition confidence level, if high confidence level is higher than this value, thinks that identification is effective, otherwise
Rejection is other.
As shown in figure 12, wherein, in Figure 12, the first from left figure is original location gray scale to the vehicle-logo recognition result positioning out in Fig. 9
Figure, the second from left figure is through the pretreated figure such as slant correction, normalization, after right two figures are bilinearity OTSU binaryzation
Binary map, a right figure be right two figures with ATL mates after, the template that similarity is the highest, coupling mark is the Audi of 89 points
Template, so this car mark single frames recognition result is Audi.
In embodiments of the present invention, as shown in figure 13, in addition to comprising the method step of Fig. 1-embodiment illustrated in fig. 11, also
Including step S106, set up vehicle tracking chain, obtain multiple vehicle-logo recognition results of described vehicle, select the plurality of car to identify
Car mark type corresponding to mark soprano in other result is final vehicle-logo recognition result.
Its method particularly includes: setting up the vehicle entered in the visual field and follow the tracks of chain, each node followed the tracks of on chain has this car
Vehicle-logo recognition result, when vehicle meets condition of publishing picture or when rolling the visual field away from, add up single frames vehicle-logo recognition on this tracking chain
Highest score and the car mark type of correspondence, car mark that occurrence number is most and corresponding confidence level, if two types is identical,
For final vehicle-logo recognition result, otherwise in the car mark that occurrence number is most, ask for an average confidence level, if the first is
High confidence level be higher than this certain mark of average confidence level, then the highest by mark the occurs car final recognition result of mark Sort positioning,
The car mark type otherwise selecting occurrence number most is final recognition result.
Set up tracking chain and carry out Optimum Matching, matching precision can be improved, use raising vehicle-logo recognition precision.
Figure 14 is the structural representation of the vehicle-logo location identification system of the embodiment of the present invention.As it can be seen, the present embodiment
Vehicle-logo location identification system includes:
Image collecting device 101, for the headstock image of collection vehicle；Car mark estimates band of position generating means 102, uses
In utilizing described headstock image, carry out vehicle location and plate location recognition, and according to the prior information of car mark Yu car plate, generate
Car mark estimates the band of position；Car mark precise position information generating means 103, for estimating in the band of position at described car mark, root
Filter the interference of car mark background area according to texture information, the car mark texture gray-scale map after filtering the interference of car mark background area is carried out
Binaryzation, carries out car mark by searching connected domain method on the binary image generated in described car mark estimates the band of position
It is accurately positioned, generates car mark precise position information；Car mark binary map generating means 104, for according to described car mark exact position
Information, extracts car mark accurate greyscale figure, and described car mark accurate greyscale figure is carried out binaryzation, generate that described car mark is corresponding two
Value figure；Vehicle-logo recognition result generating means 105, for carrying out binary map corresponding for described car mark and the car mark template prestored
Joining, be weighted according to positional information, car mark type corresponding to calculated mark soprano is vehicle-logo recognition result.
In the present embodiment, image collecting device can be high-definition camera, and capture enters the headstock image in the visual field, 200W field
Scape frame lower each second 15, image size is to be frame each second 7 under 1600*1200,500W scene, and size is 2432*2048, image
Resolution ratio can be adjusted according to the difference of high-definition camera.In the following step of the present embodiment, it is to adopt in step S101
The each two field picture got processes, and generates the vehicle-logo recognition result of corresponding each frame.
In the present embodiment, as shown in figure 15, car mark is estimated band of position generating means 101 and is included:
Image compression unit 1011, for being compressed the headstock image of described collection in proportion, generates gray level image.
After receiving the headstock image that high-definition camera gathers, in order to reduce complexity, coloured image is compressed in proportion, simultaneously
Carry out dimensionality reduction and extract gray level image as the input followed by vehicle-logo location identification.
Vehicle region signal generating unit 1012, for setting up background according to Gauss model, according to the back of the body on described gray level image
Scape difference and frame difference carry out vehicle location, are i.e. confirmed, by adjacent three frames " do difference take with ", the vehicle carrying out out, thus give birth to
Become vehicle location region.
License plate area signal generating unit 1013, in described vehicle location region, according to texture information and the face of vehicle
Look information carries out License Plate, generates the car plate band of position.In the present embodiment, also include by Character segmentation and slightly identify,
Find optimum car plate in License Plate the most accurately to identify, if recognition confidence is less than the recognition threshold preset, think
This vehicle does not contains effective board, and vehicle location result likely has deviation, without effective car plate and car mark region in vehicle region, no
Then think that this vehicle location is effective.
Car mark discreet area signal generating unit 1014, in the described car plate band of position, according to described car mark and car plate
Prior information carry out vehicle-logo location, generate car mark discreet area, wherein, described car mark includes car mark with the prior information of car plate
Positional information, car mark shape and vehicle symmetry with car plate.
In the present embodiment, as shown in figure 16, car mark precise position information generating means 102 includes:
Texture blending unit 1021, for extracting sobel cross grain and longitudinal direction in described car mark estimates the band of position
Texture.
Ambient interferences texture determines unit 1022, for comparing the gray-scale map after extracting cross grain and extracting longitudinal texture
After the average gray of gray-scale map, determine, by both differences, the grain direction that described car mark background area is disturbed.
Car mark texture gray-scale map signal generating unit 1023, for extracting and the grain direction phase of described car mark background area interference
Anti-phase texture image, generates car mark and estimates the car mark texture gray-scale map in the band of position.In the present embodiment, if both differences
More than the texture threshold (texture threshold typically takes 15, certainly, can be other numerical value according to actual conditions) set, then it is assumed that Che Biao
Background texture is cross grain, so the longitudinal texture that should extract in car mark discreet area, filters heatsink transverse window interference etc.,
The most remaining car mark texture information.If after transverse and longitudinal background texture is extracted in car mark discreet area, if longitudinal texture mean value is remote
More than horizontal mean value, then it is assumed that background is longitudinal texture, it is necessary to extract cross grain, longitudinal interference of texture of wiping out background
And only it is left car mark information, if but cross grain is close with longitudinal texture average, then and background is considered reticular texture, this
Time will by sobel filter operator extract Texture.That is: extract contrary with the grain direction of described car mark background area interference
The texture image of phase, to generate the car mark texture gray-scale map that car mark is estimated in the band of position.
Car mark texture binary map signal generating unit 1024, for described car mark veining gray-scale map is carried out binary conversion treatment,
And carry out morphologic filtering and corrosion expansion, generate the binary map that described car mark texture gray-scale map is corresponding.
Car mark boundary alignment unit 1025, for carrying out on the basis of the binary map that described car mark veining gray-scale map is corresponding
Projection, finds the right boundary that car mark is corresponding, generates car mark precise position information.
In the present embodiment, as shown in figure 17, car mark binary map generating means 103 includes:
Tilt corrector unit 1031, for according to the license plate sloped angle generated in described plate location recognition, to described
Car mark accurate greyscale figure carries out slant correction；Car mark binary map signal generating unit 1032, for accurate to the car mark after described correction
Gray-scale map carries out binary conversion treatment according to the method for bilinearity OTSU, generates the binary map that described car mark is corresponding.At the present embodiment
In, the car mark precise position information be given according to vehicle-logo location, extract car mark gray-scale map, by above carry out plate location recognition
In the license plate sloped angle calculated, car mark region is carried out slant correction, is then normalized to the figure of 40*40, in order to
Prevent the light and shade caused because of interference effects such as light inequality, the shadow of the trees uneven, to car mark gray-scale map according to the method for bilinearity OTSU
Carry out binaryzation.
In the present embodiment, as shown in figure 18, vehicle-logo recognition result generating means 104 includes: fitting unit 1041, is used for
Outline matching is carried out by the car mark in the binary map that hough conversion is corresponding to described car mark；Taxon 1042, for root
According to fitting result, described car mark is classified；Matching unit 1043, for by described car mark and described car mark corresponding types
Car mark template is mated.
In the present embodiment, on the basis of binary map, by hough conversion, car mark is carried out outline matching, simultaneously basis
The prioris such as car mark depth-width ratio shape, are the class in oval, circle and irregular figure by this car mark rough segmentation, then select
The car mark masterplate of the type mates.Car mark masterplate is to extract according to the principle that between class, difference is maximum, class interpolation is maximum
Representative car mark binary map, size is 40*40, in order to save space complexity and time complexity, by binary map eight
One byte of position synthesis, the masterplate boil down to 40*5 of 40*40 size, the car of current each type indicates three templates, and number can
To be adjusted according to actual conditions.
When carrying out car mark template matches, it is contemplated that location is difficult to be accurate to the location of pixels on car mark border, and
Be likely to certain border interference information when choosing masterplate, thus use shiding matching method, car mark binary map upper,
Under, the slip of 1 to 2 pixels can be carried out on left and right four direction, blank parts mends 0 alignment, then by binary map eight conjunction
Become a byte to mate with template, it is contemplated that same type of car mark, such as Ford and BYD, Kia, popular, Buick with
Benz etc., outline is identical is all circular or oval, and only inside center information is different, so when stencil matching, root
Being weighted according to positional information, in the car mark binary map of 40*40,8 layers of weight of outermost are 1, and middle 16 layers of weight are 2, center 16
Layer weight is 3, if not matching as zero, is set to 1 point, 2 points, 3 points respectively according to position weight difference after the match is successful,
After be 0 100 points by score normalization, by all sort results after shiding matching, choose mark the highest for final single frames
Vehicle-logo recognition result, arranges single frames vehicle-logo recognition confidence level, if high confidence level is higher than this value, thinks that identification is effective, otherwise
Rejection is other.
In the present invention, as shown in figure 19, the vehicle-logo location identification system of the present invention is real except comprising shown in Figure 14-Figure 18
Execute outside the structure of example, also include multiframe car mark preferred embodiment 106, be used for setting up vehicle tracking chain, obtain the multiple of described vehicle
Vehicle-logo recognition result, the car mark type selecting the mark soprano in the plurality of vehicle-logo recognition result corresponding is final car mark
Recognition result.
This device vehicle to entering in the visual field is set up and is followed the tracks of chain, and following the tracks of each node on chain has the car mark of this vehicle
Recognition result, when vehicle meets condition of publishing picture or rolls the visual field away from, adds up the best result of single frames vehicle-logo recognition on this tracking chain
Several and corresponding car mark type, car mark that occurrence number is most and corresponding confidence level, if two types is identical, be final car
Mark recognition result, otherwise in the car mark that occurrence number is most, asks for an average confidence level, if the first high confidence level
Higher than this certain mark of average confidence level, then the highest by mark the occurs car final recognition result of mark Sort positioning, otherwise select
The most car mark type of occurrence number is final recognition result.Set up tracking chain and carry out Optimum Matching, matching precision, mat can be improved
To improve vehicle-logo recognition precision.
Figure 20 a, b, c, d are vehicle-logo location recognition methods and the system utilizing the present invention, carry out the car mark of several vehicles
The process chart identified.Its display process is image → bilinearity OTSU bis-after vehicle-logo location gray-scale map → slant correction
Matching template in binary map → ATL after value.Wherein, Figure 20 (a) is to Audi's car car target identification process, its
Partition number is 85 points；Figure 20 (b) is to popular car car target recognition result, and its coupling mark is 82 points；Figure 20 (c) is to BMW
Car car target recognition result, its coupling mark is 77 points；Figure 20 (d) is to Suzuki car car target recognition result, and it mates mark
It it is 78 points.And to several frequently seen car mark is identified on certain expressway test, the index of test is as follows:
Audi: 99%；Popular: 99%；Suzuki: 90%；BMW: 86%；Lucky: 85%；Benz: 83%；Modern: 89%；Toyota:
97%；Honda: 98%.Therefore, car mark can be accurately positioned and effectively know by mark positioning identifying method and the system of the present invention
Not.
The vehicle-logo location recognition methods of the embodiment of the present invention and system, on the premise of there is effective board in vehicle, it is possible to fast
Speed orient car cursor position, and carry out single frames identification, many frame acknowledgments, determine the unique of vehicle and vehicle by car plate and car mark
Property.
Particular embodiments described above, has been carried out the purpose of the present invention, technical scheme and beneficial effect the most in detail
Describe in detail bright, be it should be understood that the specific embodiment that the foregoing is only the present invention, the guarantor being not intended to limit the present invention
Protect scope, all within the spirit and principles in the present invention, any modification, equivalent substitution and improvement etc. done, should be included in this
Within the protection domain of invention.
Claims (12)
1. a vehicle-logo location knows method for distinguishing, it is characterised in that described method includes:
Utilize the headstock image of high-definition camera collection vehicle；
Utilize described headstock image, carry out vehicle location and plate location recognition, and according to the prior information of car mark Yu car plate, raw
Car mark is become to estimate the band of position；
Estimate in the band of position at described car mark, filter the interference of car mark background area according to the texture information extracted, filtering car
Carry out binaryzation on car mark texture gray-scale map after the interference of mark background area, the binary image generated is connected by searching
Territory method carries out car mark in described car mark estimates the band of position and is accurately positioned, and generates car mark precise position information；
According to described car mark precise position information, extract car mark accurate greyscale figure, and described car mark accurate greyscale figure is carried out two
Value, generates the binary map that described car mark is corresponding；
Binary map corresponding for described car mark is mated, according to position by the method using shiding matching with the car mark template prestored
Information is weighted, and car mark type corresponding to calculated mark soprano is vehicle-logo recognition result, wherein uses cunning
The method of dynamic coupling is particularly as follows: carry out 1 to 2 pictures by binary map corresponding for described car mark on the four direction of upper and lower, left and right
The slip of element, blank parts is mended 0 alignment, then one byte of binary map eight synthesis corresponding for described car mark is prestored with described
Car mark template mate；
Described utilize described headstock image, carry out vehicle location and License Plate, and according to the prior information of car mark Yu car plate, raw
Car mark is become to estimate the step of the band of position, comprising:
The headstock image of described collection is compressed in proportion, generates gray level image；
Described gray level image sets up background according to Gauss model, does poor taking according to background subtraction and adjacent three frames and carry out with operation
Vehicle location, generates vehicle location region, and wherein, adjacent three frames do difference and take and operate particularly as follows: use continuous three frames, and adjacent two
Frame does frame difference then binaryzation two-by-two, two the frame difference binary map obtained are taken with, i.e. same in said two frame difference binary map
What there was frame difference information position is considered available point, obtain taking with after final frame difference binary map, find connected domain and be vehicle
Region, the vehicle positioning out by frame difference confirms with the vehicle positioning out by background subtraction, obtains final car
Position；
In described vehicle location region, texture information and colouring information according to vehicle carry out plate location recognition, generate car
The board band of position；
In the described car plate band of position, carry out vehicle-logo location according to the prior information of described car mark Yu car plate, generate car mark pre-
Estimate region.
Vehicle-logo location the most according to claim 1 knows method for distinguishing, it is characterised in that the priori letter of described car mark and car plate
Breath includes car mark and the positional information of car plate, car mark shape and vehicle symmetry.
Vehicle-logo location the most according to claim 1 knows method for distinguishing, it is characterised in that described estimate position at described car mark
In region, filter the interference of car mark background area according to the texture information extracted, the car mark after filtering the interference of car mark background area
Carry out binaryzation on texture gray-scale map, the binary image generated estimates lane place by connected domain method at described car mark
Carry out car mark in territory to be accurately positioned, generate the step of car mark precise position information, including:
Sobel cross grain and longitudinal texture is extracted in described car mark estimates the band of position；
The average gray of the gray-scale map after relatively extracting the gray-scale map after cross grain and extracting longitudinal texture, poor by both
Value determines the grain direction that described car mark background area is disturbed；
Extract the texture image mutually anti-phase with the grain direction of described car mark background area interference, generate car mark and estimate the band of position
Interior car mark texture gray-scale map；
Described car mark veining gray-scale map is carried out binary conversion treatment, and carries out morphologic filtering and corrosion expansion, generate described
The binary map that car mark texture gray-scale map is corresponding；
Project on the basis of the binary map that described car mark veining gray-scale map is corresponding, find the right boundary that car mark is corresponding,
Generate car mark precise position information.
Vehicle-logo location the most according to claim 1 knows method for distinguishing, it is characterised in that described according to the described accurate position of car mark
Confidence ceases, and extracts car mark accurate greyscale figure, and described car mark accurate greyscale figure is carried out binaryzation, generate described car mark corresponding
The step of binary map, including:
According to the license plate sloped angle generated in described plate location recognition, carry out described car mark accurate greyscale figure tilting school
Just；
Car mark accurate greyscale figure after described correction is carried out binary conversion treatment according to the method for bilinearity OTSU, generates described car
The binary map that mark is corresponding.
Vehicle-logo location the most according to claim 4 knows method for distinguishing, it is characterised in that described by corresponding for described car mark two
Value figure carries out the step mated with the car mark template prestored, including:
Outline matching is carried out by the car mark in the binary map that hough conversion is corresponding to described car mark；
According to fitting result, described car mark is classified；
Select to mate with the car mark template of described car mark corresponding types.
Vehicle-logo location the most according to claim 1 knows method for distinguishing, it is characterised in that after obtaining vehicle-logo recognition result,
Also include:
Set up vehicle tracking chain, obtain multiple vehicle-logo recognition results of described vehicle, select in the plurality of vehicle-logo recognition result
Car mark type corresponding to mark soprano be final vehicle-logo recognition result.
7. the system of a vehicle-logo location identification, it is characterised in that described system includes:
Image collecting device, for the headstock image of collection vehicle；
Car mark estimates band of position generating means, is used for utilizing described headstock image, carries out vehicle location and plate location recognition,
And according to the prior information of car mark Yu car plate, generate car mark and estimate the band of position；
Car mark precise position information generating means, for estimating in the band of position at described car mark, according to the texture information extracted
Filter the interference of car mark background area, the car mark texture gray-scale map after filtering the interference of car mark background area carries out binaryzation,
In described car mark estimates the band of position, carry out car mark by searching connected domain method on the binary image generated to be accurately positioned,
Generate car mark precise position information；
Car mark binary map generating means, for according to described car mark precise position information, extracts car mark accurate greyscale figure, and by institute
State car mark accurate greyscale figure and carry out binaryzation, generate the binary map that described car mark is corresponding；
Vehicle-logo recognition result generating means, for use the method for shiding matching by binary map corresponding for described car mark with prestore
Car mark template is mated, and is weighted according to positional information, the car mark type that calculated mark soprano is corresponding
For vehicle-logo recognition result, wherein use the method for shiding matching particularly as follows: by binary map corresponding for described car mark upper and lower, left,
Carrying out the slip of 1 to 2 pixels on right four direction, blank parts mends 0 alignment, then by binary map eight corresponding for described car mark
Position one byte of synthesis is mated with the car mark template prestored；
Described car mark is estimated band of position generating means and is included:
Image compression unit, for being compressed the headstock image of described collection in proportion, generates gray level image；
Vehicle region signal generating unit, for setting up background according to Gauss model, according to background subtraction and phase on described gray level image
Adjacent three frames do poor taking and carry out vehicle location with operation, generate vehicle location region, and wherein, adjacent three frames do difference and take and operate concrete
For: use continuous three frames, adjacent two frames to do frame difference then binaryzation two-by-two, two the frame difference binary map obtained are taken and, i.e. institute
That states that same position in two frame difference binary map has frame difference information is considered available point, obtain taking with after final frame difference two-value
Figure, finds connected domain and is vehicle region, the vehicle positioning out by frame difference and the car positioning out by background subtraction
Confirm, obtain final vehicle location；
License plate area signal generating unit, in described vehicle location region, texture information and colouring information according to vehicle enter
Row plate location recognition, generates the car plate band of position；
Car mark discreet area signal generating unit, in the described car plate band of position, according to the priori letter of described car mark with car plate
Breath carries out vehicle-logo location, generates car mark discreet area.
The system of vehicle-logo location identification the most according to claim 7, it is characterised in that the priori letter of described car mark and car plate
Breath includes car mark and the positional information of car plate, car mark shape and vehicle symmetry.
The system of vehicle-logo location identification the most according to claim 7, it is characterised in that described car mark precise position information is raw
Device is become to include:
Texture blending unit, for extracting sobel cross grain and longitudinal texture in described car mark estimates the band of position；
Ambient interferences texture determines unit, the gray scale after comparing the gray-scale map after extracting cross grain and extracting longitudinal texture
The average gray of figure, determines, by both differences, the grain direction that described car mark background area is disturbed；
Car mark texture gray-scale map signal generating unit, for extracting the line mutually anti-phase with the grain direction of described car mark background area interference
Reason image, generates car mark and estimates the car mark texture gray-scale map in the band of position；
Car mark texture binary map signal generating unit, for described car mark veining gray-scale map is carried out binary conversion treatment, and carries out shape
State filtering and corrosion expand, and generate the binary map that described car mark texture gray-scale map is corresponding；
Car mark boundary alignment unit, for projecting on the basis of the binary map that described car mark veining gray-scale map is corresponding, seeks
Look for the right boundary that car mark is corresponding, generate car mark precise position information.
The system of vehicle-logo location identification the most according to claim 7, it is characterised in that described car mark binary map generates dress
Put according to described car mark precise position information, extract car mark accurate greyscale figure, and described car mark accurate greyscale figure is carried out two-value
Change, generate the binary map that described car mark is corresponding, including:
Tilt corrector unit, for according to the license plate sloped angle generated in described plate location recognition, accurate to described car mark
Gray-scale map carries out slant correction；
Car mark binary map signal generating unit, for entering according to the method for bilinearity OTSU the car mark accurate greyscale figure after described correction
Row binary conversion treatment, generates the binary map that described car mark is corresponding.
The system of 11. vehicle-logo location identifications according to claim 7, it is characterised in that described vehicle-logo recognition result generates
Device includes:
Fitting unit, the car mark in the binary map corresponding to described car mark by hough conversion carries out outline matching；
Taxon, for classifying to described car mark according to fitting result；
Matching unit, for mating the car mark template of described car mark with described car mark corresponding types.
The system of 12. vehicle-logo location identifications according to claim 7, it is characterised in that described vehicle-logo location identification device
Also include:
Multiframe car mark preferred embodiment, is used for setting up vehicle tracking chain, obtains multiple vehicle-logo recognition results of described vehicle, selects institute
Stating car mark type corresponding to the mark soprano in multiple vehicle-logo recognition result is final vehicle-logo recognition result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310009960.XA CN103077384B (en) | 2013-01-10 | 2013-01-10 | A kind of method and system of vehicle-logo location identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310009960.XA CN103077384B (en) | 2013-01-10 | 2013-01-10 | A kind of method and system of vehicle-logo location identification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103077384A CN103077384A (en) | 2013-05-01 |
| CN103077384B true CN103077384B (en) | 2016-08-31 |
Family
ID=48153908
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310009960.XA Active CN103077384B (en) | 2013-01-10 | 2013-01-10 | A kind of method and system of vehicle-logo location identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103077384B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108154122A (en) * | 2017-12-26 | 2018-06-12 | 天津天地伟业投资管理有限公司 | A kind of automobile logo identification method based on image |
Families Citing this family (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103473774B (en) * | 2013-09-09 | 2017-04-05 | 长安大学 | A kind of vehicle positioning method based on pavement image characteristic matching |
| CN103530622B (en) * | 2013-11-07 | 2017-03-29 | 青岛海信网络科技股份有限公司 | A kind of car logo identification method |
| CN103577829B (en) * | 2013-11-08 | 2016-04-20 | 中安消技术有限公司 | A kind of vehicle-logo location method and apparatus |
| CN103559492A (en) * | 2013-11-12 | 2014-02-05 | 公安部第三研究所 | Car logo recognition device and method |
| CN103544489A (en) * | 2013-11-12 | 2014-01-29 | 公安部第三研究所 | Device and method for locating automobile logo |
| CN104331691A (en) * | 2014-11-28 | 2015-02-04 | 深圳市捷顺科技实业股份有限公司 | Vehicle logo classifier training method, vehicle logo recognition method and device |
| CN104391966B (en) * | 2014-12-03 | 2017-09-29 | 中国人民解放军国防科学技术大学 | Typical logo searching method based on deep learning |
| CN104504398B (en) * | 2014-12-31 | 2017-12-01 | 深圳市捷顺科技实业股份有限公司 | Automobile logo identification method and vehicle-logo recognition device |
| CN105205486B (en) * | 2015-09-15 | 2018-12-07 | 浙江宇视科技有限公司 | A kind of automobile logo identification method and device |
| CN106650550A (en) * | 2015-10-28 | 2017-05-10 | 中通服公众信息产业股份有限公司 | Vehicle model identification method and vehicle model identification system through image characteristics of vehicle mark and vehicle head |
| CN105262174B (en) * | 2015-11-02 | 2017-12-12 | 国网山东省电力公司电力科学研究院 | The orderly charging system of Itellectualized uptown electric automobile and method based on Intelligent Recognition |
| CN105426896A (en) * | 2015-11-16 | 2016-03-23 | 成都神州数码索贝科技有限公司 | Car logo automatic identification method and system |
| WO2017088478A1 (en) * | 2015-11-24 | 2017-06-01 | 乐视控股（北京）有限公司 | Number separating method and device |
| CN105868683A (en) * | 2015-11-24 | 2016-08-17 | 乐视致新电子科技（天津）有限公司 | Channel logo identification method and apparatus |
| CN105868755A (en) * | 2015-11-24 | 2016-08-17 | 乐视致新电子科技（天津）有限公司 | Number separation method and apparatus |
| CN105550684B (en) * | 2015-12-14 | 2019-10-29 | 广东安居宝数码科技股份有限公司 | License plate locating method and its system based on image |
| CN105760876B (en) * | 2016-03-18 | 2019-03-08 | 中山大学 | A kind of vehicle-logo location method based on the ablation of radiator grid background |
| CN106250555B (en) * | 2016-08-15 | 2019-07-09 | 东方网力科技股份有限公司 | Vehicle retrieval method and device based on big data |
| CN108614985A (en) * | 2016-12-12 | 2018-10-02 | 广西师范大学 | A kind of vehicle-logo location method |
| CN108154127A (en) * | 2017-12-27 | 2018-06-12 | 天津智芯视界科技有限公司 | A kind of vehicle identification method based on video and radar |
| CN108268811B (en) * | 2018-01-15 | 2021-02-02 | 深圳市易成自动驾驶技术有限公司 | Image processing method, image processing apparatus, and computer-readable storage medium |
| CN108196729A (en) * | 2018-01-16 | 2018-06-22 | 安徽慧视金瞳科技有限公司 | A kind of finger tip point rapid detection method based on infrared video |
| CN108154510A (en) * | 2018-01-17 | 2018-06-12 | 深圳市亿图视觉自动化技术有限公司 | Method for detecting surface defects of products, device and computer readable storage medium |
| CN109141829B (en) * | 2018-07-23 | 2020-10-09 | 北京大恒图像视觉有限公司 | Target line definition calculation and industrial camera rear intercept detection method |
| CN109241973B (en) * | 2018-08-21 | 2022-02-08 | 南京工程学院 | Full-automatic soft segmentation method for characters under texture background |
| CN109214386B (en) * | 2018-09-14 | 2020-11-24 | 京东数字科技控股有限公司 | Method and apparatus for generating image recognition model |
| CN109360236B (en) * | 2018-10-26 | 2022-05-17 | 郑州轻工业学院 | Sunflower disc size measurement method |
| KR102604821B1 (en) * | 2018-11-09 | 2023-11-20 | 에스케이텔레콤 주식회사 | Apparatus and method for estimating location of vehicle |
| CN109658584B (en) * | 2018-12-14 | 2021-01-12 | 泰康保险集团股份有限公司 | Bill information identification method and device |
| CN110969135B (en) * | 2019-12-05 | 2023-04-18 | 中南大学 | Vehicle logo recognition method in natural scene |
| CN111695565A (en) * | 2020-06-14 | 2020-09-22 | 荆门汇易佳信息科技有限公司 | Automobile mark accurate positioning method based on road barrier fuzzy image |
| CN111815545A (en) * | 2020-07-14 | 2020-10-23 | 南京信息工程大学 | Intelligent terminal-based color image rapid processing method |
| CN113111883B (en) * | 2021-03-23 | 2023-06-06 | 浙江大华技术股份有限公司 | License plate detection method, electronic device and storage medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7912288B2 (en) * | 2006-09-21 | 2011-03-22 | Microsoft Corporation | Object detection and recognition system |
| CN101246551A (en) * | 2008-03-07 | 2008-08-20 | 北京航空航天大学 | Fast license plate locating method |
| CN101630361A (en) * | 2008-12-30 | 2010-01-20 | 北京邮电大学 | Plate number, body color and mark identification-based equipment and plate number, body color and mark identification-based method for identifying fake plate vehicles |
| CN101719217A (en) * | 2009-12-15 | 2010-06-02 | 中国地质大学（武汉） | Elastic relaxation algorithm-based vehicle type identification system and elastic relaxation algorithm-based vehicle type identification method |
| CN101789080B (en) * | 2010-01-21 | 2012-07-04 | 上海交通大学 | Detection method for vehicle license plate real-time positioning character segmentation |
| CN102521891B (en) * | 2011-12-06 | 2014-09-10 | 北京万集科技股份有限公司 | Method and system for detecting vehicle congestion at toll station |
-
2013
- 2013-01-10 CN CN201310009960.XA patent/CN103077384B/en active Active
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108154122A (en) * | 2017-12-26 | 2018-06-12 | 天津天地伟业投资管理有限公司 | A kind of automobile logo identification method based on image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103077384A (en) | 2013-05-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103077384B (en) | A kind of method and system of vehicle-logo location identification | |
| CN105426864B (en) | One kind being based on the matched Multi-lane Lines Detection method of equidistant marginal point | |
| CN101030256B (en) | Method and apparatus for cutting vehicle image | |
| CN104778444B (en) | The appearance features analysis method of vehicle image under road scene | |
| CN104036262B (en) | A kind of method and system of LPR car plates screening identification | |
| CN109657632B (en) | Lane line detection and identification method | |
| CN104809433B (en) | A kind of zebra line detecting method based on maximum stable region and stochastical sampling | |
| CN109446917B (en) | Vanishing point detection method based on cascading Hough transform | |
| CN111563412B (en) | Rapid lane line detection method based on parameter space voting and Bessel fitting | |
| Li et al. | Nighttime lane markings recognition based on Canny detection and Hough transform | |
| CN109635656A (en) | Vehicle attribute recognition methods, device, equipment and medium neural network based | |
| CN106650553A (en) | License plate recognition method and system | |
| CN102842037A (en) | Method for removing vehicle shadow based on multi-feature fusion | |
| CN101634705B (en) | Method for detecting target changes of SAR images based on direction information measure | |
| CN102750708A (en) | Affine motion target tracing algorithm based on fast robust feature matching | |
| CN106887004A (en) | A kind of method for detecting lane lines based on Block- matching | |
| CN103049788B (en) | Based on space number for the treatment of object detection system and the method for computer vision | |
| CN103679704A (en) | Video motion shadow detecting method based on lighting compensation | |
| CN103198319A (en) | Method of extraction of corner of blurred image in mine shaft environment | |
| CN112949484B (en) | High-speed railway rockfall real-time detection method | |
| CN108304749A (en) | The recognition methods of road speed line, device and vehicle | |
| CN109858438A (en) | A kind of method for detecting lane lines based on models fitting | |
| Satzoda et al. | Robust extraction of lane markings using gradient angle histograms and directional signed edges | |
| Maya et al. | Performance analysis of lane detection algorithm using partial Hough transform | |
| Wang et al. | Lane detection algorithm based on density clustering and RANSAC |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C53 | Correction of patent of invention or patent application | ||
| CB03 | Change of inventor or designer information |
Inventor after: Ge Leiming Inventor after: Fang Yanming Inventor after: Cheng Yujuan Inventor after: Wang Qin Inventor before: Ge Leiming Inventor before: Fang Yanming Inventor before: Cheng Yujuan |
|
| COR | Change of bibliographic data |
Free format text: CORRECT: INVENTOR; FROM: GE LEIMING FANG YANMING CHENG YUJUAN TO: GE LEIMING FANG YANMING CHENG YUJUAN WANG QIN |
|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20221103 Address after: Room 601, 6th Floor, Building B, Building 9, Start Zone, Yangtze River Delta International R&D Community, No. 286, Qinglonggang Road, High-speed Rail New Town, Xiangcheng District, Suzhou City, Jiangsu Province, 215131 Patentee after: Suzhou Wanji IoV Technology Co.,Ltd. Address before: 100085 No. 5, building 1, East Road, Haidian District, Beijing, 601 Patentee before: BEIJING WANJI TECHNOLOGY Co.,Ltd. |
|
| TR01 | Transfer of patent right |