CN106250824A - Vehicle window localization method and system - Google Patents
Vehicle window localization method and system Download PDFInfo
- Publication number
- CN106250824A CN106250824A CN201610579549.XA CN201610579549A CN106250824A CN 106250824 A CN106250824 A CN 106250824A CN 201610579549 A CN201610579549 A CN 201610579549A CN 106250824 A CN106250824 A CN 106250824A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- mark
- vehicle window
- region
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/59—Context or environment of the image inside of a vehicle, e.g. relating to seat occupancy, driver state or inner lighting conditions
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of vehicle window localization method and system, relates to technical field of computer vision.Wherein, vehicle window localization method includes: based on the pretreatment to vehicle forward detection image, at least obtains the marginal information in vehicle image region and region；Going out the mark in vehicle image region at least based on vehicle image area judging, mark includes the mark in vehicle image region independent of vehicle window；Determining each border of vehicle window in marginal information to determine vehicle window region at least based on mark and default distance model, default distance model includes mark and vehicle window and the distance relation on each border of vehicle window.The vehicle window targeting scheme of the present invention is by differentiating vehicle window region according to mark and predeterminable range model, vehicle window location more accurately can be realized, further, owing to mark has the good robustness to light and color, various different occasion therefore can be preferably applicable to.
Description
Technical field
The present invention relates to technical field of computer vision, particularly relate to a kind of vehicle window localization method and system.
Background technology
In the big data platform of urban public security bayonet vehicle, a critically important function is according to special in vehicle drive window
Mark find out rapidly target carriage, can be very helpful work for analyzing the event such as fake-licensed car, unlicensed car, hit-and-run
With, it addition, the detection of a series of vehicles " fingerprint " such as annual test Mark Detection, co-driver personnel detection, Herba Plantaginis goods of furniture for display rather than for use detection is all
Need first to navigate to corresponding vehicle window region.Therefore, vehicle window location is effectively to realize other functions multiple and detection accurately
Basis.
Having the research that some relevant vehicle windows position at present, major part is all based on Image Edge-Detection, and inventor is in reality
Find that these methods of the prior art position for fuzzy scene, night scenes, specular scene etc. during the existing present invention
Ability, it is impossible to effective satisfied reality application.
Summary of the invention
The present invention provides a kind of vehicle window localization method and system, in order to solve one or more present in prior art asking
Topic.
First aspect, the embodiment of the present invention provides a kind of vehicle window localization method, including: detect image based on to vehicle forward
Pretreatment, at least obtain the marginal information in vehicle image region and region, wherein, pretreatment includes rim detection；At least
Going out the mark in vehicle image region based on vehicle image area judging, mark includes in vehicle image region independent of car
The mark of window, it determines include differentiating based on support vector machine；Believe at edge at least based on mark and default distance model
Determining each border of vehicle window in breath to determine vehicle window region, default distance model includes mark and vehicle window and vehicle window
The distance relation on each border.
Second aspect, the embodiment of the present invention provides a kind of vehicle window alignment system, including: pretreatment module, it is configured to base
In the pretreatment to vehicle forward detection image, at least obtain the marginal information in vehicle image region and region, wherein, locate in advance
Reason includes rim detection；Mark discrimination module, is configured to go out vehicle image region at least based on vehicle image area judging
In mark, mark includes the mark in vehicle image region independent of vehicle window, it determines include based on support vector machine
Differentiate；Vehicle window area judging module, is configured to differentiate in marginal information at least based on mark and default distance model
Going out each border of vehicle window to determine vehicle window region, default distance model includes mark and vehicle window and each border of vehicle window
Distance relation.
The third aspect, the embodiment of the present application additionally provides a kind of nonvolatile computer storage media, and storage has computer
Executable instruction, described computer executable instructions is used for performing any of the above-described vehicle window localization method of the application.
Fourth aspect, the embodiment of the present application additionally provides a kind of electronic equipment, including: one or more processors；And
Memorizer；Wherein, described memorizer storage has the program that can be performed by the one or more processor, and described instruction is described
One or more processors perform, so that the one or more processor is able to carry out any of the above-described vehicle window location of the present invention
Method.
The vehicle window localization method of embodiment of the present invention offer and system, by obtaining the mark being easier to determine on vehicle
Will thing, and determine vehicle window region further based on mark and predeterminable range model, it is possible to achieve more accurately vehicle window is sentenced
Not.Further, due to the mark on vehicle, there is the good robustness to light, color etc., it is possible to achieve multiple
Vehicle window identification under complex scene.
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 this
Some bright embodiments, for those of ordinary skill 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 embodiment flow chart of vehicle window localization method of the present invention；
Fig. 2 is the idiographic flow schematic diagram of the embodiment of vehicle window targeting scheme of the present invention；
Fig. 3 a is an embodiment application scenarios schematic diagram of vehicle window targeting scheme of the present invention；
Fig. 3 b is another embodiment application scenarios schematic diagram of vehicle window targeting scheme of the present invention；
Fig. 3 c is the further embodiment application scenarios schematic diagram of vehicle window targeting scheme of the present invention；
Fig. 3 d is a still further embodiment application scenarios schematic diagram of vehicle window targeting scheme of the present invention；
Fig. 4 is the example structure schematic diagram of vehicle window alignment system of the present invention；
Electronic equipment hard performing multimedia file sharing method based on preview thumbnail that Fig. 5 provides for the application
Part structural representation.
Detailed description of the invention
For making the purpose of the embodiment of the present invention, technical scheme and advantage clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is
The a part of embodiment of the present invention rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art
The every other embodiment obtained under not making creative work premise, broadly falls into the scope of protection of the invention.
In an embodiment of the present invention, relate to multiple can be used for and reflect characteristics of image, image is judged and carries out
The algorithm of rim detection, is the most first explained and illustrated the algorithm that may use, so that those skilled in the art can be more
The good understanding present invention.
Haar eigenvalue reflects the grey scale change situation of image, is divided three classes: edge feature, linear character, center spy
Seek peace diagonal feature, be combined into feature templates.There are white and two kinds of rectangles of black in feature templates, and define the spy of this template
Value indicative be white rectangle pixel and deduct black rectangle pixel and.Such as: some features of face can be simple by rectangular characteristic
Describing, such as: eyes are deeper than cheek color, and bridge of the nose both sides are deeper than bridge of the nose color, face is deeper etc. than ambient color.But square
Shape feature is only to some simple graphic structures, as more sensitive in edge, line segment, (level, hangs down so particular orientation can only be described
Directly, diagonal angle) structure.
LBP (Local Binary Patterns, local binary patterns), basic thought is the pixel to image and its office
Result after portion's surrounding pixel contrasts is sued for peace.Using this pixel as center, neighbor is carried out threshold ratio relatively.
If the brightness of center pixel is more than or equal to his neighbor, he is labeled as 1, is otherwise labeled as 0.This description method can
Well to capture the details in image.It practice, researchers can obtain state-of-the-art water with it in Texture classification
Flat.LBP has been successfully applied to Face datection, lip reading identification, expression detection, dynamic texture etc. field.Its algorithm complex
Low, consume internal memory little, principle is simple, but is not necessarily suitable for all of feature description.
Adaboost is a kind of iterative algorithm, and its core concept is the grader different for the training of same training set
(Weak Classifier), then gets up these weak classifier set, constitutes a higher final grader (strong classifier).It is calculated
Method itself realizes by changing data distribution, and it is the most correct according to the classification of sample each among each training set,
And the accuracy rate of the general classification of last time, determine the weights of each sample.The new data set revising weights is given down
Layer grader is trained, and finally the grader obtained will be trained finally to merge, as last Decision Classfication device every time.
Use adaboost grader can get rid of some unnecessary training data features, and be placed on above the training data of key.
SVM (Support Vector Machine, support vector machine) is the study having supervision of one, machine learning field
Model, is commonly used to carry out pattern recognition, classification and regression analysis.The main thought of SVM may be summarized to be 2 points: (1) it
It is that linear can a point situation be analyzed, during for linearly inseparable, by using non-linear map by low-dimensional
The sample of input space linearly inseparable is converted into high-dimensional feature space and makes its linear separability, so that high-dimensional feature space is adopted
With linear algorithm, the nonlinear characteristic of sample is carried out linear analysis to be possibly realized.(2) it is based on structural risk minimization theory
On in feature space construction optimum segmentation hyperplane so that learner obtains global optimization, and at whole sample space
Expected risk meet certain upper bound with certain probability.
Sobel operator (Sobel operator) is mainly used as rim detection, and technically, it is that a discreteness difference is calculated
Son, for the approximation of the gray scale of arithmograph image brightness function.This operator is used, it will produce correspondence in any point of image
Gray scale vector or its law vector.Sobel operator is upper and lower according to pixel, left and right adjoint point intensity-weighted is poor, reaches in edge
Edge is detected to this phenomenon of extreme value.Noise had smoothing effect, it is provided that more accurate edge directional information, edge positions
Precision is not high enough.When not being the highest to required precision, it it is a kind of more conventional edge detection method.
HOG (Histogram of Oriented Gradient, histograms of oriented gradients) feature is a kind of at computer
Vision and image procossing are used for carrying out the Feature Descriptor of object detection.The method employs the gradient direction of of image itself
Feature, is similar to edge orientation histogram method, but it is characterized in that it is at the unified grid list of the size that grid is intensive
Calculate in unit, and employ the normalized method of local contrast of overlap in order to improve degree of accuracy.Histograms of oriented gradients
The core concept of feature is that the presentation of the object in piece image and shape can by the direction Density Distribution at gradient or edge very
Describe well.Its implementation is first to divide the image into little to be called pane location connected region；Then gather in pane location
The gradient direction of each pixel or edge orientation histogram；Finally these set of histograms be can be formed by feature description altogether
Son.In order to improve degree of accuracy, it is also possible to these local histograms are carried out contrast in the bigger interval (block) of image
Normalization, the method is by first calculating each rectangular histogram density in this interval (block), then according to this density value pair
Each pane location in interval does normalization.After this normalization, can illumination variation and shade be obtained the most steady
Qualitative.Compared with other character description method, HOG has many good qualities.Firstly, since HOG is the local grid list at image
Operate in unit, thus it to image geometry and the deformation of optics good invariance, both deformation can be kept only to go out
On the biggest space field.Secondly, at thick spatial domain sampling, the sampling of fine direction and stronger indicative of local optical normalizing
Change under the conditions of waiting, as long as the posture that pedestrian generally can be kept upright, can allow that pedestrian has some trickle limb actions,
These trickle actions can be left in the basket and not affect Detection results.Therefore HOG feature is particularly suitable the people doing in image
Health check-up is surveyed.
Corrosion and expansion are all the basic operations during morphological images processes, and corrosion is mainly used in morphology removing image
Some part, Matlab imerode function realizes Image erosion.
It should be noted that in the case of not conflicting, the embodiment in the present invention and the feature in embodiment can phases
Combination mutually.Describe the present invention below with reference to the accompanying drawings and in conjunction with the embodiments in detail.
With reference to Fig. 1, it illustrates embodiment flow process Figure 100 of vehicle window localization method of the present invention.
As it is shown in figure 1, in a step 101, based on the pretreatment to vehicle forward detection image, vehicle image is at least obtained
Marginal information in region and region.
In the present embodiment, vehicle window alignment system carries out pretreatment, to obtain vehicle image to vehicle forward detection image
Marginal information in region and vehicle image region.Wherein, vehicle forward detection image can be that in prior art, some are important
The bayonet socket image of traffic block port, such as expressway tol lcollection mouth, public security inspection post etc., it will usually utilize the shooting dress laid
Put, to each vehicle of process shoot, and the bayonet vehicle image photographed is saved in data base, in order to follow-up
Carrying out retrieving and checking, bayonet vehicle image mostly is vehicle frontal image, and certainly, vehicle forward detection image can also is that not
The direct picture of the new detection mode detection developed, the present invention does not the most limit.
In the present embodiment, vehicle window alignment system can be one individually for detecting the system of vehicle window, it is also possible to be
One ingredient of the subsystem of other system or other system such as traffic checking system, the present invention does not the most limit
System.The pretreatment of vehicle forward detection image can be included vehicle forward detection image is carried out textural characteristics sign by system,
As utilized Haar eigenvalue or LBP to carry out binaryzation, or can also is that gradient information sign etc., by corresponding pre-place
Reason, can be analyzed it by relevant Adaboost training aids, it can be deduced that the region at vehicle image place further,
Can downscaled images process range further.
In the present embodiment, the pretreatment to vehicle forward detection image can also include before by some pretreatment
The vehicle image region obtained carries out rim detection, optionally, it is possible to use Sobel operator carries out rim detection, to detect
Marginal information.
Then, in a step 102, the mark in vehicle image region is gone out at least based on vehicle image area judging.
In the present embodiment, system determines in vehicle image region further according to the vehicle image region determined before
Mark.Wherein, mark is the vehicle mark independent of vehicle window, can be such as rearview mirror, rain brush, arc angle point,
Car plate etc..By determining mark from vehicle image region, can get rid of what other objects on vehicle travel caused
Interference.Mark generally has the object of good robustness to light, color, it is not easy to become by external world's light intensity and color
The object changed and change.
Afterwards, in step 103, in marginal information, vehicle window is determined at least based on mark and default distance model
Each border to determine vehicle window region.
In the present embodiment, system determines in marginal information according to the mark detected and default distance model
Each border of vehicle window is to determine vehicle window region.Wherein, the distance model preset includes each of mark and vehicle window and vehicle window
The distance relation on border.Such as, can record vehicle window height in the distance model preset is 1.5～5.5 times of car plate width, rain brush
Position essentially identical with the position of the lower boundary of vehicle window, or the coboundary of the position of rearview mirror and vehicle window or lower boundary or a left side
The relation etc. of the position of right margin, the present invention does not the most limit.In further alternative embodiment, system can be first
Witness marker thing, wherein mark can be rearview mirror, then according to the level of Sobel horizontal edge detection template detection vehicle
Edge；Afterwards according to the positional information of rearview mirror, differentiate from the horizontal edge detected based on SVM and draw the position with rearview mirror
Put the final position of immediate lower boundary；Further according to final position and the default distance model of lower boundary, such as, can be
Vehicle window height and the relation of car plate width, differentiate the final position drawing coboundary from the horizontal edge detected based on SVM.
In some optional embodiments, based on the pretreatment to vehicle forward detection image, at least obtain vehicle image
Marginal information in region and region farther includes: detects image according to vehicle forward, goes out based on vehicle discriminating Model checking
Vehicle image region；Based on the rim detection to vehicle image region, obtain the marginal information in region.Wherein, vehicle discriminating
Model can be first to process vehicle forward detection image with Haar eigenvalue, then determine vehicle by Adaboost algorithm further
Image-region, carries out Sobel edge edge detection afterwards, detects the marginal information in vehicle image region vehicle image region.
It should be noted that and other models of other algorithms known in the art or following exploitation can also be used to carry out the data being correlated with
Processing or image procossing, the present invention does not the most limit.
In other optional embodiments, go out the mark in vehicle image region at least based on vehicle image area judging
Thing farther includes: according to vehicle image region, determining mark based on mark discrimination model, wherein, mark includes
Light and color had good robustness.Wherein, Marker Identity model can be first with LBP, image to be carried out binaryzation
Process, then differentiate the mark in vehicle image region by Adaboost algorithm.It should be noted that and can also use the most
The data that other algorithms known or other models of following exploitation carry out being correlated with process or image procossing, and the present invention is in this regard
Do not limit.
In some optional embodiments, vehicle discriminating model and mark discrimination model can include that one or more are calculated
The discrimination model that method combination is formed.Can be such as HAAR-ADABOOST algorithm or LBP-ADABOOST algorithm etc., this
Bright the most do not limit.
In other optional embodiments, mark can include rearview mirror, rain brush, arc angle point, car light and car plate
In one or more.Differentiate for example, it is possible to utilize mark rain brush to combine arc angle point and default distance model based on SVM
Go out vehicle window border, it is also possible to utilizing rearview mirror to combine rain brush, rearview mirror combines car plate and car light etc., and the present invention does not has in this regard
Limit.
The vehicle window localization method of the present embodiment is by utilizing mark and and pre-based between mark and vehicle window border
If distance model, can make border differentiate more accurate.Further, typically have due to mark light and color
Good robustness, therefore this vehicle window localization method may be used under complex scene such as the vehicle window under fuzzy, Gao Guang, night scenes
Location.
With further reference to Fig. 2, the idiographic flow of the embodiment that it illustrates vehicle window targeting scheme of the present invention illustrates Figure 200.
In conjunction with Fig. 2, whole identification work can be divided into three big portions for the practical situation in reality scene by the present invention
Point.First carry out vehicle detection, edge analysis positions with vehicle mirrors, then utilizes support vector machine method to wait according to edge
Favored area carries out up-and-down boundary and tentatively confirms；Finally combine the initial results of position of rear view mirror information and up-and-down boundary to position and obtain
Drive window region.
As in figure 2 it is shown, in step 201, vehicle detection, obtain vehicle subimage.
In the present embodiment, system can utilize existing HAAR-ADABOOST method to carry out vehicle image region determining
Position detection, owing to needs carry out vehicle window location, therefore we only carry out vehicle forward detection.After orienting vehicle image region,
Can set up coordinate system with the upper left corner in vehicle image region as initial point, the mark etc. detected afterwards further all may be used
With with this coordinate system quantization means, it is simple to differentiation afterwards and calculating.The reason taking this method is, the edge contour of vehicle
Information is all apparent under various scenes, and HAAR-ADABOOST method is higher for clear-cut object detection rate.
Then, in step 202, vehicle mirrors detects, and horizontal edge detects.
In the present embodiment, we can design the Sobel horizontal edge detection template of a kind of 15*15, with the most
The multiple horizontal edges detected in vehicle image region, vehicle mirrors, rain brush etc. are carried out based on LBP-simultaneously
The detection and localization of ADABOOST, due to vehicle mirrors, the texture information relatively horn of plenty of rain brush, therefore selection the method can be more
Well the mark that the texture information such as rearview mirror, rain brush is abundant is detected.
Afterwards, in step 203, edge length statistical analysis, obtain drive window up-and-down boundary candidate region.
In the present embodiment, system is entered after the multiple edges detected before can first carry out appropriate dilation erosion again
The connective length statistical analysis of row, will meet length (can be such as the length between the abscissa meeting two rearview mirrors in left and right
Degree) edge record, and each acquiring size edge subimage expanding regulation up and down；When implementing, we are with edge
Five pixels are respectively expanded centered by vertical coordinate.
Then, in step 204, utilize SVM that candidate region is differentiated.
In the present embodiment, system can utilize SVM statistical analysis current edge subimage to be vehicle window top edge, lower limb
Or without relative edges, during statistical analysis, we select histogram of gradients feature description candidate subimage, make full use of on drive window
The arc angle point on border, the arc angle point of lower boundary and rain brush information.
In step 205, differentiate that result carries out lower boundary confirmation according to position of rear view mirror with SVM.
In the present embodiment, through above step, system can obtain several candidate's lower limbs, and now we combine detection
Position of rear view mirror carry out the optimization of vehicle window lower boundary and confirm, obtain the distance nearest marginal position of rearview mirror vertical coordinate, ask for
Meansigma methods is in this, as the final position of lower boundary.Generally, the recognition result only one of which of lower boundary, occasionally there are two
Individual above situation.
Afterwards, in step 206, combine SVM according to drive window lower boundary and differentiate that result carries out coboundary confirmation.
In the present embodiment, after confirmation to lower boundary in above-mentioned steps, according to lower boundary information, integrating step
The some top edge results obtained in 204, according to vehicle window height priori (can be such as vehicle window height be 1.5～5.5
Car plate width again), it is judged that final position, coboundary also exports accordingly result.Generally the recognition result of coboundary is only
One, occasionally there are the situation of two or three.
Finally, in step 207, according to the abscissa validation of information right boundary of rearview mirror.
In the present embodiment, owing to the rearview mirror testing result obtained is the most stable, we are directly by the horizontal stroke of rearview mirror
Coordinate information is as the right boundary of drive window.Experimental verification this have no effect on follow-up use.Confirm up-and-down boundary and left and right
Vehicle window region can be obtained after border.
Concrete application scenarios schematic diagram may refer to Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d.Wherein, Fig. 3 a shows this
The original vehicle forward detection image schematic diagram of one embodiment application scenarios of bright vehicle window targeting scheme；Fig. 3 b shows this
The schematic diagram of the location vehicle image-region of another embodiment application scenarios of bright vehicle window targeting scheme；Fig. 3 c shows this
The schematic diagram of the location vehicle mark (being rearview mirror in figure) of the further embodiment application scenarios of bright vehicle window targeting scheme；With
And Fig. 3 d show vehicle window targeting scheme of the present invention a still further embodiment application scenarios location up-and-down boundary schematic diagram.
As shown in Figure 3 a, owing to needs carry out vehicle window location, so original image needs to be that vehicle forward detects image,
The bayonet socket image of such as each checkpoint.With further reference to Fig. 3 b, it illustrates location vehicle image-region (in figure shown in 310)
Application scenarios figure, can be the interference that subsequent detection gets rid of a lot of roads, environment etc. by first orienting vehicle image region
Factor, improves the accuracy of detection further, such as, can set up coordinate system with the upper left corner in vehicle image region for initial point, with
It is easy to follow-up calculating.With reference to Fig. 3 c, it illustrates and detect further in vehicle image region 310 in mark such as figure
The application scenarios figure of rearview mirror 321 and 322, figure also show each edge (non-reference number) detected, such as rain brush
Edge, car light, the edge etc. of car plate etc..With further reference to Fig. 3 d, it illustrates and enter from each edge that Fig. 3 c detects
One step filters out the application scenarios figure of up-and-down boundary by the distance model preset and distinguished number.Further can also be according to mark
The position of will thing such as rearview mirror determines that right boundary is to orient vehicle window region.
Pass through above example, it is possible to achieve effective location to vehicle window region under complex scene, due to mark with
There is fixing distance between vehicle window, utilize mark and the distance model just energy preferably exclusive PCR factor preset.Such as
Rim detection may detect a lot of medicated clothing etc. without relative edges such as sunshading board, color and texture distinctness, by mark with
The distance model preset can be got rid of without relative edges effectively, it is achieved vehicle window location more accurately.
Refer to Fig. 4, it illustrates the example structure schematic diagram of vehicle window alignment system of the present invention.
As shown in Figure 4, vehicle window alignment system 400 includes pretreatment module 401, mark discrimination module 402 and vehicle window district
Territory discrimination module 403.Wherein, pretreatment module 401, it is configured to based on the pretreatment to vehicle forward detection image, at least
Obtaining the marginal information in vehicle image region and region, wherein, described pretreatment includes rim detection；Mark discrimination module
402, it is configured to the vehicle image area judging at least based on pretreatment module 401 obtains and goes out the mark in vehicle image region
Thing, mark includes the mark in vehicle image region independent of vehicle window, it determines include differentiating based on support vector machine；And
Vehicle window area judging module 403, be configured to the mark that determines at least based on mark discrimination module 402 and default away from
Determining each border of vehicle window in marginal information to determine vehicle window region from model, default distance model includes mark
Distance relation with each border of vehicle window and vehicle window.
In some optional embodiments, pretreatment module 401 includes being configured to further: detect according to vehicle forward
Image, goes out vehicle image region based on vehicle discriminating Model checking；And based on the rim detection to vehicle image region, obtain
Marginal information in region.
In other optional embodiments, mark discrimination module 402 includes being configured to further: according to vehicle figure
As region, determining mark based on mark discrimination model, wherein, mark includes that light and color are had good Shandong
Rod.
In some optional embodiments, vehicle discriminating model and mark discrimination model can include that one or more are calculated
The discrimination model that method combination is formed.Can be such as HAAR-ADABOOST algorithm or LBP-ADABOOST algorithm etc., this
Bright the most do not limit.
In other optional embodiments, mark can include rearview mirror, rain brush, arc angle point, car light and car plate
In one or more.
In the present embodiment, vehicle window alignment system 400, by vehicle window region and mark are detected and positioned, is gone forward side by side
One step determines each border of vehicle window to obtain vehicle window region according to mark with the distance model preset, it is possible to achieve based on
Independent of the mark on the vehicle of vehicle window, vehicle window is positioned.Further, typically it is chosen to be light due to mark
With the mark that color has good robustness, therefore the detection to mark is not easily susceptible to external environment condition such as light color
The impact of change, has good stability, therefore goes in the scene of complexity.
Should be appreciated that all modules described in Fig. 4 are corresponding with each step in the method with reference to described in Fig. 1.By
This, all modules that the operation described above with respect to method and feature and corresponding technique effect are equally applicable in Fig. 4, at this
Repeat no more.
The system related in each above system and method can be all a server or server cluster, its
In each module above-mentioned can also be single server or server cluster, now, the interaction table between above-mentioned module
Now mutual between the server corresponding to each module.
This application provides a kind of nonvolatile computer storage media, described computer-readable storage medium storage has computer
Executable instruction, this computer executable instructions can perform based on preview thumbnail the many matchmakers in above-mentioned any means embodiment
Body sharing files method.
Fig. 5 is the hardware architecture diagram of the electronic equipment performing vehicle window localization method that the application provides, such as Fig. 5 institute
Showing, this equipment includes:
One or more processors 510 and memorizer 520, in Fig. 5 as a example by a processor 510.
The equipment performing vehicle window localization method can also include: input equipment 530 and output device 540.
Processor 510, memorizer 520, input equipment 530 and output device 540 can be by bus or other modes
Connect, in Fig. 5 as a example by being connected by bus.
Memorizer 520, as a kind of non-volatile computer readable storage medium storing program for executing, can be used for storing non-volatile software journey
Sequence, non-volatile computer executable program and module, program as corresponding in the vehicle window localization method in the embodiment of the present application
Instruction/module (such as, the pretreatment module 401 shown in accompanying drawing 4, mark discrimination module 402 and vehicle window area judging module
403).Non-volatile software program, instruction and the module that processor 510 is stored in memorizer 520 by operation, thus hold
The various functions application of row server and data process, and i.e. realize said method embodiment vehicle window localization method.
Memorizer 520 can include storing program area and storage data field, and wherein, storage program area can store operation system
Application program required for system, at least one function；The use that storage data field can store according to vehicle window alignment system is created
Data etc..Additionally, memorizer 520 can include high-speed random access memory, it is also possible to include nonvolatile memory, example
Such as at least one disk memory, flush memory device or other non-volatile solid state memory parts.In certain embodiments, deposit
Reservoir 520 is optional includes the memorizer remotely located relative to processor 510, and these remote memories can be connected by network
To vehicle window alignment system.The example of above-mentioned network includes but not limited to the Internet, intranet, LAN, mobile radio communication
And combinations thereof.
Input equipment 530 can receive numeral or the character information of input, and produce the user with vehicle window alignment system and set
Put and function controls relevant key signals input.Output device 540 can include the display devices such as display screen.
One or more module stores is in described memorizer 520, when by one or more processor
During 510 execution, perform the vehicle window localization method in above-mentioned any means embodiment.
The said goods can perform the method that the embodiment of the present application is provided, and possesses the corresponding functional module of execution method and has
Benefit effect.The ins and outs of the most detailed description, can be found in the method that the embodiment of the present application is provided.
The electronic equipment of the embodiment of the present application exists in a variety of forms, includes but not limited to:
(1) mobile communication equipment: the feature of this kind equipment is to possess mobile communication function, and to provide speech, data
Communication is main target.This Terminal Type includes: smart mobile phone (such as iPhone), multimedia handset, functional mobile phone, and low
End mobile phone etc..
(2) super mobile personal computer equipment: this kind equipment belongs to the category of personal computer, has calculating and processes merit
Can, the most also possess mobile Internet access characteristic.This Terminal Type includes: PDA, MID and UMPC equipment etc., such as iPad.
(3) portable entertainment device: this kind equipment can show and play content of multimedia.This kind equipment includes: audio frequency,
Video player (such as iPod), handheld device, e-book, and intelligent toy and portable car-mounted navigator.
(4) server: providing the equipment of the service of calculating, the composition of server includes that processor, hard disk, internal memory, system are total
Lines etc., server is similar with general computer architecture, but owing to needing to provide highly reliable service, is therefore processing energy
The aspects such as power, stability, reliability, safety, extensibility, manageability require higher.
(5) other have the electronic installation of data interaction function.
Device embodiment described above is only schematically, and the wherein said unit illustrated as separating component can
To be or to may not be physically separate, the parts shown as unit can be or may not be physics list
Unit, i.e. may be located at a place, or can also be distributed on multiple NE.Can be selected it according to the actual needs
In some or all of module realize the purpose of the present embodiment scheme.
Through the above description of the embodiments, those skilled in the art it can be understood that to each embodiment can
The mode adding general hardware platform by software realizes, naturally it is also possible to pass through hardware.Based on such understanding, above-mentioned technology
The part that correlation technique is contributed by scheme the most in other words can embody with the form of software product, this computer
Software product can store in a computer-readable storage medium, such as ROM/RAM, magnetic disc, CD etc., including some instructions in order to
Make a computer equipment (can be personal computer, server, or the network equipment etc.) perform each embodiment or
The method described in some part of embodiment.
Last it is noted that above example is only in order to illustrate the technical scheme of the application, it is not intended to limit；Although
With reference to previous embodiment, the application is described in detail, it will be understood by those within the art that: it still may be used
So that the technical scheme described in foregoing embodiments to be modified, or wherein portion of techniques feature is carried out equivalent；
And these amendment or replace, do not make appropriate technical solution essence depart from the application each embodiment technical scheme spirit and
Scope.
Claims (10)
1. a vehicle window localization method, including:
Based on the pretreatment to vehicle forward detection image, at least obtain the edge letter in vehicle image region and described region
Breath, wherein, described pretreatment includes rim detection；
Going out the mark in described vehicle image region at least based on described vehicle image area judging, described mark includes institute
Stating the mark independent of vehicle window in vehicle image region, described differentiation includes differentiating based on support vector machine；
In described marginal information, each limit of described vehicle window is determined at least based on described mark and default distance model
Boundary is to determine vehicle window region, and described default distance model includes described mark and described vehicle window and each limit of described vehicle window
The distance relation on boundary.
Method the most according to claim 1, described based on the pretreatment to vehicle forward detection image, at least obtain vehicle
Marginal information in image-region and described region includes:
Detect image according to vehicle forward, go out described vehicle image region based on vehicle discriminating Model checking；
Based on the rim detection to described vehicle image region, obtain the marginal information in described region.
Method the most according to claim 1, described goes out described vehicle image at least based on described vehicle image area judging
Mark in region includes:
According to described vehicle image region, determine described mark, wherein, described mark pair based on mark discrimination model
Light and color have good robustness.
The most according to the method in claim 2 or 3, described vehicle discriminating model and described mark discrimination model include one
Or the discrimination model that many algorithms combination is formed.
5., according to the method according to any one of claim 1-3, described mark includes rearview mirror, rain brush, arc angle point, car
One or more in lamp and car plate.
6. a vehicle window alignment system, including:
Pretreatment module, be configured to based on to vehicle forward detection image pretreatment, at least obtain vehicle image region and
Marginal information in described region, wherein, described pretreatment includes rim detection；
Mark discrimination module, is configured to go out in described vehicle image region at least based on described vehicle image area judging
Mark, described mark includes the mark in described vehicle image region independent of vehicle window, and described differentiation includes based on propping up
Hold vector machine to differentiate；
Vehicle window area judging module, is configured at least based on described mark and default distance model in described marginal information
In determine each border of described vehicle window to determine vehicle window region, described default distance model includes described mark and institute
State the distance relation on each border of vehicle window and described vehicle window.
System the most according to claim 6, described pretreatment module includes being configured to further:
Detect image according to vehicle forward, go out described vehicle image region based on vehicle discriminating Model checking；
Based on the rim detection to described vehicle image region, obtain the marginal information in described region.
System the most according to claim 6, described mark discrimination module includes being configured to further:
According to described vehicle image region, determine described mark, wherein, described mark pair based on mark discrimination model
Light and color have good robustness.
9., according to the system described in claim 7 or 8, described vehicle discriminating model and described mark discrimination model include one
Or the discrimination model that many algorithms combination is formed.
10. according to the system according to any one of claim 6-8, described mark include rearview mirror, rain brush, arc angle point,
One or more in car light and car plate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610579549.XA CN106250824A (en) | 2016-07-21 | 2016-07-21 | Vehicle window localization method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610579549.XA CN106250824A (en) | 2016-07-21 | 2016-07-21 | Vehicle window localization method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106250824A true CN106250824A (en) | 2016-12-21 |
Family
ID=57603578
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610579549.XA Pending CN106250824A (en) | 2016-07-21 | 2016-07-21 | Vehicle window localization method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106250824A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108108656A (en) * | 2017-11-15 | 2018-06-01 | 浙江工业大学 | Vehicle window corner detection and multidirectional projection-based vehicle window accurate positioning method |
| CN108133170A (en) * | 2017-10-31 | 2018-06-08 | 浙江浩腾电子科技股份有限公司 | A kind of vehicle window localization method of multi-direction vehicle |
| CN108182376A (en) * | 2017-11-15 | 2018-06-19 | 浙江工业大学 | Vehicle window corner detection-based vehicle window positioning method |
| WO2018182538A1 (en) * | 2017-03-31 | 2018-10-04 | Agency For Science, Technology And Research | Systems and methods that improve alignment of a robotic arm to an object |
| CN108960228A (en) * | 2017-05-18 | 2018-12-07 | 富士通株式会社 | Detection device, method and the image processing equipment of vehicle |
| CN109949331A (en) * | 2019-04-17 | 2019-06-28 | 合肥泰禾光电科技股份有限公司 | Container edge detection method and device |
| CN110182023A (en) * | 2019-05-29 | 2019-08-30 | 京东方科技集团股份有限公司 | Control method, system, device, medium and the equipment of vehicle window transparent region range |
| CN110610519A (en) * | 2019-09-25 | 2019-12-24 | 江苏鸿信系统集成有限公司 | Vehicle window positioning method based on deep learning |
| CN111949307A (en) * | 2020-07-06 | 2020-11-17 | 北京大学 | Optimization method and system of open source project knowledge graph |
| CN115359650A (en) * | 2022-07-06 | 2022-11-18 | 浙江大华技术股份有限公司 | Parking position detection method and device, computer equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105005759A (en) * | 2015-05-04 | 2015-10-28 | 南京理工大学 | Multi-characteristic fused monitoring image front vehicle window positioning and extracting method |
| CN105404874A (en) * | 2015-11-27 | 2016-03-16 | 成都神州数码索贝科技有限公司 | Vehicle window recognition system based on projection and hough straight line detection |
| CN105469057A (en) * | 2015-11-27 | 2016-04-06 | 成都神州数码索贝科技有限公司 | Hough line detection and projection-based automobile window detection method |
-
2016
- 2016-07-21 CN CN201610579549.XA patent/CN106250824A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105005759A (en) * | 2015-05-04 | 2015-10-28 | 南京理工大学 | Multi-characteristic fused monitoring image front vehicle window positioning and extracting method |
| CN105404874A (en) * | 2015-11-27 | 2016-03-16 | 成都神州数码索贝科技有限公司 | Vehicle window recognition system based on projection and hough straight line detection |
| CN105469057A (en) * | 2015-11-27 | 2016-04-06 | 成都神州数码索贝科技有限公司 | Hough line detection and projection-based automobile window detection method |
Non-Patent Citations (3)
| Title |
|---|
| 杨鹏: "基于最大局部变化的安全带检测算法", 《浙江工贸职业技术学院学报》 * |
| 王运琼，游志胜: "基于色差均值的快速车窗定位算法", 《计算机应用于软件》 * |
| 骆玉荣 等: "一种自动车窗识别方法的设计与实现", 《计算机技术与应用进展•2007——全国第18届计算机技术与应用（CACIS）学术会议论文集》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018182538A1 (en) * | 2017-03-31 | 2018-10-04 | Agency For Science, Technology And Research | Systems and methods that improve alignment of a robotic arm to an object |
| CN108960228A (en) * | 2017-05-18 | 2018-12-07 | 富士通株式会社 | Detection device, method and the image processing equipment of vehicle |
| CN108133170B (en) * | 2017-10-31 | 2020-07-14 | 浙江浩腾电子科技股份有限公司 | Window positioning method for multi-directional vehicle |
| CN108133170A (en) * | 2017-10-31 | 2018-06-08 | 浙江浩腾电子科技股份有限公司 | A kind of vehicle window localization method of multi-direction vehicle |
| CN108182376A (en) * | 2017-11-15 | 2018-06-19 | 浙江工业大学 | Vehicle window corner detection-based vehicle window positioning method |
| CN108108656B (en) * | 2017-11-15 | 2020-07-07 | 浙江工业大学 | Vehicle window corner detection and multidirectional projection-based vehicle window accurate positioning method |
| CN108108656A (en) * | 2017-11-15 | 2018-06-01 | 浙江工业大学 | Vehicle window corner detection and multidirectional projection-based vehicle window accurate positioning method |
| CN108182376B (en) * | 2017-11-15 | 2020-12-08 | 浙江工业大学 | Vehicle window corner detection-based vehicle window positioning method |
| CN109949331A (en) * | 2019-04-17 | 2019-06-28 | 合肥泰禾光电科技股份有限公司 | Container edge detection method and device |
| CN110182023A (en) * | 2019-05-29 | 2019-08-30 | 京东方科技集团股份有限公司 | Control method, system, device, medium and the equipment of vehicle window transparent region range |
| CN110182023B (en) * | 2019-05-29 | 2021-10-12 | 京东方科技集团股份有限公司 | Control method, system, device, medium and equipment for window transparent area range |
| CN110610519A (en) * | 2019-09-25 | 2019-12-24 | 江苏鸿信系统集成有限公司 | Vehicle window positioning method based on deep learning |
| CN111949307A (en) * | 2020-07-06 | 2020-11-17 | 北京大学 | Optimization method and system of open source project knowledge graph |
| CN111949307B (en) * | 2020-07-06 | 2021-06-25 | 北京大学 | Optimization method and system of open source project knowledge graph |
| CN115359650A (en) * | 2022-07-06 | 2022-11-18 | 浙江大华技术股份有限公司 | Parking position detection method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106250824A (en) | Vehicle window localization method and system | |
| Hsu et al. | Robust license plate detection in the wild | |
| CN105160309B (en) | Three lanes detection method based on morphological image segmentation and region growing | |
| US9846946B2 (en) | Objection recognition in a 3D scene | |
| CN105404886B (en) | Characteristic model generation method and characteristic model generating means | |
| CN106446150B (en) | A kind of method and device of vehicle precise search | |
| Janahiraman et al. | Traffic light detection using tensorflow object detection framework | |
| WO2013065220A1 (en) | Image recognition device, image recognition method, and integrated circuit | |
| US20120093420A1 (en) | Method and device for classifying image | |
| CN104166841A (en) | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network | |
| CN106250838A (en) | vehicle identification method and system | |
| CN103208185A (en) | Method and system for nighttime vehicle detection on basis of vehicle light identification | |
| CN102855500A (en) | Haar and HoG characteristic based preceding car detection method | |
| CN110929593A (en) | Real-time significance pedestrian detection method based on detail distinguishing and distinguishing | |
| CN101984453B (en) | Human eye recognition system and method | |
| CN106128115A (en) | Fusion method for detecting road traffic information based on double cameras | |
| Cui et al. | Vehicle localisation using a single camera | |
| JP2014178862A (en) | Feature point extraction device | |
| Hakim et al. | Implementation of an image processing based smart parking system using Haar-Cascade method | |
| Lee et al. | Near-infrared-based nighttime pedestrian detection using grouped part models | |
| CN104463238B (en) | A kind of automobile logo identification method and system | |
| Zhou et al. | Hybridization of appearance and symmetry for vehicle-logo localization | |
| CN104751197A (en) | Device and method for recognizing faces of drivers during vehicle running on basis of video analysis | |
| CN113343903B (en) | License plate recognition method and system in natural scene | |
| Nguwi et al. | Number plate recognition in noisy image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20161221 |
|
| WD01 | Invention patent application deemed withdrawn after publication |