CN105205486B - A kind of automobile logo identification method and device - Google Patents
A kind of automobile logo identification method and device Download PDFInfo
- Publication number
- CN105205486B CN105205486B CN201510586228.8A CN201510586228A CN105205486B CN 105205486 B CN105205486 B CN 105205486B CN 201510586228 A CN201510586228 A CN 201510586228A CN 105205486 B CN105205486 B CN 105205486B
- Authority
- CN
- China
- Prior art keywords
- logo
- region
- fusion
- candidate region
- confidence level
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The application provides a kind of automobile logo identification method and device, this method comprises: training logo classifier;Obtain license plate position favored area at the beginning of to determine logo;Favored area at the beginning of logo is detected using logo classifier, obtains the first logo candidate region;Calculate the first logo confidence level of the first logo candidate region;Logo position confidence level is calculated according to the position of the first logo candidate region;It is filtered out apart from central axes closer first logo candidate region according to logo position confidence level as the second logo candidate region;Second logo candidate region is identified, the second logo confidence level is obtained;Fusion is carried out to the second logo candidate region and generates fusion candidate region;According to the fusion confidence level of the corresponding fusion candidate region of the first logo confidence level, logo position confidence level and the second logo confidence calculations;Select the fusion highest fusion candidate region of confidence level as the logo identified.The vehicle-logo recognition rate under complex scene can be improved in the application.
Description
Technical field
This application involves technical field of video monitoring more particularly to a kind of automobile logo identification method and devices.
Background technique
Logo is the important information of vehicle, is the significant image of vehicle, and vehicle-logo recognition can propose vehicle monitoring and tracking
For strong Informational support.But since logo is compared with small, similitude is big, and be illuminated by the light, the reasons such as background, shape influence, difficult
To accurately identify.
Current vehicle-logo recognition technology depends on the precise positioning of logo, and vehicle-logo location mostly use image procossing with
And mode identification technology, the vehicle-logo location mode is to extreme sensitivities such as surrounding's texture of logo, angle rotation, inclinations, therefore,
It is difficult to accomplish to be accurately positioned under complex scene, and then causes the discrimination of logo low.
Summary of the invention
In view of this, the application provides a kind of automobile logo identification method and device.
Specifically, the application is achieved by the following technical solution:
The application provides a kind of automobile logo identification method, this method comprises:
Using logo detection algorithm training logo classifier;
Obtain the license plate location information in image to be detected;
Favored area at the beginning of determining logo according to the license plate location information;
The logo classifier obtained using training carries out logo detection to favored area at the beginning of the logo, obtains several first vehicles
Mark candidate region;
Calculate the first logo confidence level of each the first logo candidate region;
Corresponding logo position confidence level is calculated according to the position of the first logo candidate region;
It is filtered out at the beginning of the logo according to logo position confidence level from several first logos candidate region
The central axes of favored area closer first logo candidate region is as the second logo candidate region;
Second logo candidate region is identified using machine learning algorithm, obtains the second logo candidate region
Second logo confidence level;
Region fusion is carried out to multiple second logos candidate region and generates multiple fusion candidate regions;
According to the first logo confidence level, the logo position confidence level of the second logo candidate region for generating fusion candidate region
And second the corresponding fusion candidate region of logo confidence calculations fusion confidence level;
Select the fusion highest fusion candidate region of confidence level as the logo identified.
The application also provides a kind of vehicle-logo recognition device, which includes:
Training unit, for using logo detection algorithm training logo classifier;
Acquiring unit, for obtaining the license plate location information in image to be detected;
Determination unit, favored area at the beginning of for determining logo according to the license plate location information;
Detection unit, the logo classifier for being obtained using training carry out logo detection to favored area at the beginning of the logo,
Obtain several first logos candidate region;
First computing unit, for calculating the first logo confidence level of each the first logo candidate region;
Second computing unit, for calculating corresponding logo position confidence level according to the position of the first logo candidate region;
Screening unit, for according to logo position confidence level from several first logos candidate region filter out away from
Central axes closer first logo candidate region from favored area at the beginning of the logo is as the second logo candidate region;
Recognition unit obtains second for identifying using machine learning algorithm to second logo candidate region
Second logo confidence level of logo candidate region;
Integrated unit generates multiple fusion candidate regions for carrying out region fusion to multiple second logos candidate region;
Third computing unit, for the first logo confidence according to the second logo candidate region for generating fusion candidate region
The fusion confidence level of the corresponding fusion candidate region of degree, logo position confidence level and the second logo confidence calculations;
Selecting unit, for selecting the fusion highest fusion candidate region of confidence level as the logo identified.
Precise positioning of the application independent of logo it can be seen from above description, but it is based on deep learning algorithm,
Vehicle-logo recognition is carried out by the way of a variety of confidence level Weighted Fusions, improves the vehicle-logo recognition rate under complex scene.
Detailed description of the invention
Fig. 1 is a kind of automobile logo identification method flow chart shown in one exemplary embodiment of the application;
Fig. 2 is the positive sample and negative sample example shown in one exemplary embodiment of the application;
Fig. 3 is the logo primary election area schematic shown in one exemplary embodiment of the application;
Fig. 4 is a kind of underlying hardware structure of vehicle-logo recognition device place equipment shown in one exemplary embodiment of the application
Schematic diagram;
Fig. 5 is a kind of structural schematic diagram of vehicle-logo recognition device shown in one exemplary embodiment of the application.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the application.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the application.
It is only to be not intended to be limiting the application merely for for the purpose of describing particular embodiments in term used in this application.
It is also intended in the application and the "an" of singular used in the attached claims, " described " and "the" including majority
Form, unless the context clearly indicates other meaning.It is also understood that term "and/or" used herein refers to and wraps
It may be combined containing one or more associated any or all of project listed.
It will be appreciated that though various information, but this may be described using term first, second, third, etc. in the application
A little information should not necessarily be limited by these terms.These terms are only used to for same type of information being distinguished from each other out.For example, not departing from
In the case where the application range, the first information can also be referred to as the second information, and similarly, the second information can also be referred to as
One information.Depending on context, word as used in this " if " can be construed to " ... when " or " when ...
When " or " in response to determination ".
Current vehicle-logo recognition technology mainly uses image procossing and mode identification technology, in the base of license plate recognition technology
On plinth, using the relative positional relationship of license plate and logo, just positioning is carried out to logo;It recycles based on Haar feature
Adaboost algorithm carries out logo detection, obtains several doubtful car mark regions;Then using based on HOG (Histogram of
Oriented Gradient, histograms of oriented gradients) feature SVM (Support Vector Machine, support vector machines)
Algorithm screens doubtful car mark region, chooses the maximum region of confidence level as vehicle-logo location region;Finally to determining
Vehicle-logo location region carries out vehicle-logo recognition.
During above-mentioned vehicle-logo recognition, it usually needs train multiple logos according to the difference of logo length-width ratio and classify
Device, and due to length-width ratio difference, necessarily cause to be detected using different scale, different step-lengths, different sliding windows, therefore,
Need to consume a large amount of system resource.Secondly as the inherent shortcoming of HOG algorithm, is difficult the problems such as processing blocks, is incomplete, and
The algorithm is quite sensitive to image noise.In addition, choosing discriminant approach of the maximum region of confidence level as vehicle-logo location region
Too simple violence be easy to cause erroneous judgement.
In view of the above-mentioned problems, the embodiment of the present application proposes a kind of automobile logo identification method, essence of this method independent of logo
Certainly position, but it is based on deep learning algorithm, vehicle-logo recognition is carried out with the mode of a variety of confidence level Weighted Fusions.
It is one embodiment flow chart of the application automobile logo identification method, the embodiment is to vehicle-logo recognition process referring to Fig. 1
It is described.
Step 101, using logo detection algorithm training logo classifier.
Therefore the embodiment of the present application carries out vehicle-logo recognition using logo classifier during vehicle-logo recognition need to instruct in advance
Practice mark classifier.Training (the step 101) of the logo classifier need to only be completed before logo detects (step 104).
The inspection such as Adaboost algorithm, HOG algorithm, DPM (Deformable Parts Model, deformable member model) algorithm specifically can be used
Method of determining and calculating carries out logo classifier training.
In a preferred embodiment, the training of logo classifier is carried out using Adaboost algorithm, specifically:
The positive sample and negative sample of logo pattern are obtained, positive sample is the sample comprising logo pattern, and negative sample is not wrap
The sample of the pattern containing logo.In the embodiment of the present application, all positive samples are demarcated according to the ratio of width to height of 1:1, such as Fig. 2 (a)
It is shown, for being unsatisfactory for the logo of 1:1 the ratio of width to height, part of the car mark region with identification can be intercepted as positive sample, for example,
Audi's logo can choose the ring in Fourth Ring as positive sample, as shown in Fig. 2 (b).
In addition, the negative sample in the present embodiment is divided into two parts, a part is only comprising logo just favored area (logo primary election
Region be license plate area just above, can be discussed in detail in subsequent descriptions) negative sample, hereinafter referred to as the first negative sample, such as
Shown in Fig. 2 (c), which mainly considers that the embodiment of the present application is initial based on logo during subsequent vehicle-logo recognition
Region carries out vehicle-logo recognition, therefore, selects the first favored area of logo that can effectively exclude as negative sample non-in the first favored area of logo
Car mark region improves the detection efficiency of logo classifier.Another part negative sample is the negative sample comprising whole front face, is such as schemed
Shown in 2 (d), hereinafter referred to as the second negative sample.Although the first negative sample has very strong elimination ability, as at the beginning of logo
The range of favored area is smaller, and the textural characteristics in the region are limited, when logo type increases, is located at the logo just constituency
The otherness very little of car mark region and non-car mark region within the scope of domain, the simple logo classifier generated using the first negative sample
The ineffective of vehicle-logo recognition is carried out, the embodiment of the present application increases the second negative sample, which includes whole Chinese herbaceous peony
Face, textural characteristics are abundant, it is possible to increase the difference of car mark region and non-car mark region is conducive to vehicle-logo recognition, therefore, passes through increase
Second negative sample can promote the convergence of logo classifier.Certainly, either the first negative sample or the second negative sample are both needed to scratch
Take down logo pattern.
After getting training required positive sample and negative sample, the embodiment of the present application is negative according to positive sample and first first
Sample training obtains preceding N grades of strong classifiers, rear M grades of strong classifier is obtained further according to positive sample and the training of the second negative sample, by preceding N
Grade strong classifier and the cascade of rear M grades of strong classifier generate logo classifier.As previously mentioned, before being obtained according to the training of the first negative sample
N grades of strong classifiers can effectively exclude the non-car mark region in the first favored area of logo;It is obtained according to the training of the second negative sample latter M grades strong
Classifier can promote the convergence of logo classifier, and details are not described herein.
Step 102, the license plate location information in image to be detected is obtained.
Existing more mature license plate recognition technology identification license plate can be used, obtain the position letter of license plate in image to be detected
Breath, details are not described herein.
Step 103, favored area at the beginning of determining logo according to the license plate location information.
It can be found by largely observing, logo is usually located at license plate area just above, but its locating height has differences,
For example, some pony car logos are usually located at adjacent domain above license plate, and some oversize vehicle logos are usually located at license plate
Top upper zone (such as truck, medium truck).By the statistics discovery to a large amount of logo positions, logo height is usually not
More than twice license plate width.Therefore, the embodiment of the present application determines the logo being located above license plate according to above-mentioned statistical result
First favored area, the lower edge of favored area at the beginning of the logo connects with license plate upper edge, the height on side edge is license plate width 2 times,
As shown in Figure 3.The logo just favored area can cover the overwhelming majority logo position, and as at the beginning of the logo favored area only with vehicle
The width of board is high related with position, and therefore, under varying environment and different angle, the first favored area of the logo is relative to entire front face
Variation is smaller, provides the relatively stable and lesser detection zone of range for subsequent vehicle-logo recognition.
Step 104, the logo classifier obtained using training carries out logo detection to favored area at the beginning of the logo, if obtaining
Dry first logo candidate region.
The training of logo classifier is completed by step 101, using the logo classifier to the vehicle in image to be detected
It marks just favored area and carries out logo detection, exclude non-car mark region, get multiple doubtful car mark regions, hereinafter referred to as first
Logo candidate region.The the first logo candidate region obtained by above-mentioned logo detection of classifier is in true logo or similar logo
Surrounding has apparent aggregation, lays the foundation for subsequent sections fusion.
Step 105, the first logo confidence level of each the first logo candidate region is calculated.
After getting the first logo candidate region, the confidence level of each the first logo candidate region is calculated, it is simple below
Referred to as the first logo confidence level, specific calculating process are as follows:
FFstWeight=K/P formula (1)
Wherein, K is the number that the Weak Classifier of the first logo candidate region is capable of detecting when in logo classifier;P is vehicle
Mark the total number of Weak Classifier in classifier;FFstWeightFor the first logo confidence level of the first logo candidate region.
In foregoing description it is found that logo classifier is made of multiple cascade strong classifiers, and strong classifier is by training
The multiple Weak Classifiers composition generated in the process.Above-mentioned formula (1) exactly using the Weak Classifier generated in training process as
Calculate the basis of the first logo candidate region.During logo classifier training, each Weak Classifier has a correspondence
Threshold value, each Weak Classifier can have a corresponding output valve to the area to be tested of input, by the output valve with it is right
The threshold value answered is compared, so that it is determined that whether area to be tested can pass through the detection of weak typing.If area to be tested can
Add one with the Weak Classifier number by the detection of current Weak Classifier, then counted, when area to be tested passes through all weak point
Final statistical result, i.e. K value can be obtained after class device (P Weak Classifier) detection.In confirmation, the area to be tested passes through logo
After the detection of classifier, which can be used as the first logo candidate region and calculates corresponding first vehicle according to formula (1)
Mark confidence level.A possibility that first logo confidence level is bigger, then corresponding first logo candidate region is logo is bigger.
Step 106, corresponding logo position confidence level is calculated according to the position of the first logo candidate region.
Since the embodiment of the present application does not carry out precise positioning to car mark region, by algorithm of target detection pre-
If logo when just favored area carries out logo detection, may will detect that structural texture characteristic area similar with logo, and
This feature cluster around regions a large amount of first logo candidate region.
In order to exclude above structure texture characteristic area similar with logo, the embodiment of the present application be based on logo position make into
One step investigation.Although the horizontal position of logo is usually located at vehicle it is well known that the size of logo, highly having differences
On central axes, therefore, the embodiment of the present application calculates the logo of each the first logo candidate region using the position characteristic of logo
Position confidence level.
The calculating process of logo position confidence level is as follows:
Formula (2)
Wherein, D is distance of the first logo candidate region central point to the first favored area central axes of logo;W is the first logo
The width of candidate region;FLocWeightFor the logo position confidence level of the first logo candidate region.
It needs to add explanation a bit, the vehicle for logo not on central axes, still vehicle can be calculated using formula (2)
Cursor position confidence level, this is needed in logo classifier training, selects to be located in the first favored area of logo some bright on central axes
Aobvious characteristic area replaces true logo as positive sample, can equally exclude the bigger similar car mark region of position deviation.
Step 107, it is filtered out from several first logos candidate region apart from institute according to logo position confidence level
The central axes closer first logo candidate region of the first favored area of logo is stated as the second logo candidate region.
Specifically, the logo position confidence level that the embodiment of the present application is calculated according to formula (2) judges that current first logo is waited
Whether favored area can be used as the second logo candidate region and carries out subsequent vehicle-logo recognition.
In a preferred embodiment, the logo position confidence level F being calculatedLocWeightWhen between 0 to 1,
Think that corresponding first logo candidate region is located near central axes, can be used as the second logo candidate region and further identify;And
Work as FLocWeightWhen less than 0, illustrate that corresponding first logo candidate region farther out, should reject first logo time apart from central axes
Favored area.It is respectively positioned near central axes by the second logo candidate region that this step is screened, and more intensively clustered
In different height, subsequent vehicle-logo recognition range is reduced.
Step 108, second logo candidate region is identified using machine learning algorithm, obtains the second logo and waits
Second logo confidence level of favored area.
There are many existing machine learning algorithm, for example, CNN (Convolutional Neural Networks, convolution mind
Through network) algorithm, SVM algorithm, BOW (Bag of words, bag of words) algorithm etc..The embodiment of the present application is right by taking CNN algorithm as an example
Make further identification in the second logo candidate region.
When carrying out CNN classifier training, the training sample of CNN can increase on the basis of preceding aim detection algorithm sample
Add incomplete part, rotation, the logo sample translated, to enhance the robustness of CNN classifier identification logo.
In this step, set by the second logo that the identification of CNN classifier can get each the second logo candidate region
Reliability FSndWeight, the acquisition methods of the second logo confidence level are the prior art, and details are not described herein.
Step 109, region fusion is carried out to multiple second logos candidate region and generates multiple fusion candidate regions.
Since above-mentioned recognizer has error, and the second logo candidate region identified generally more intensively clusters
Around true logo or similar logo, therefore, the embodiment of the present application is close with size to position and machine learning result is consistent
The second logo candidate region merged.After carrying out S wheel fusion to the second logo candidate region of acquisition, S are generated
Candidate region is merged, specific fusion process is as follows:
Execute new round mixing operation: selection one did not merged with other logo candidate regions and was not chosen as initially melting
Original fusion region of the second logo candidate region in region as new round mixing operation is closed, which is current
Take turns the first intermediate integration region of mixing operation.
Execute work as front-wheel mixing operation: select one have neither part nor lot in when the second logo candidate region of front-wheel mixing operation as
Region to be fused;Position, width, height and the machine recognition result of intermediate integration region and region to be fused are obtained respectively
(result after machine learning algorithm identifies);Calculate current fusion threshold value;According to fusion threshold value, intermediate integration region and to
Position, width, height and the machine recognition result of integration region judge whether intermediate integration region and region to be fused meet
Fusion conditions;When intermediate integration region and region to be fused meet fusion conditions, by intermediate integration region and region to be fused
It is merged, as new intermediate integration region.
Judge whether that each second logo candidate region has participated in the mixing operation when front-wheel;It is executed if it is not, returning
When front-wheel mixing operation;If so, judging whether the second logo candidate region for being also not chosen as original fusion region;If nothing,
Then there is currently intermediate integration region be fusion candidate region;New round mixing operation is executed if so, then returning.
The calculating process of above-mentioned fusion threshold value is as follows:
DDelta=θ × MIN (iRectWdt1, iRectWdt2) formula (3)
Wherein, θ is threshold value adjustment factor, for example, θ=0.3;IRectWdt1 is the width of intermediate integration region;
IRectWdt2 is the width in region to be fused;MIN (iRectWdt1, iRectWdt2) be take intermediate integration region width and to
The minimum value of integration region width;DDelta is fusion threshold value.
It is above-mentioned judge among integration region and region to be fused whether meet fusion conditions process it is as follows:
IType1=iType2 formula (4)
| iRectX1-iRectX2 |≤dDelta formula (5)
| iRectY1-iRectY2 |≤dDelta formula (6)
| iRectX1+iRectWdt1-iRectX2-iRectWdt2 |≤dDelta formula (7)
| iRectY1+iRectHgt1-iRectY2-iRectHgt2 |≤dDelta formula (8)
Wherein, iType1 is the machine recognition result of intermediate integration region;IType2 is the machine recognition in region to be fused
As a result;(iRectX1, iRectY1) is the position of intermediate integration region;(iRectX2, iRectY2) is the position in region to be fused
It sets;IRectWdt1 is the width of intermediate integration region;IRectWdt2 is the width in region to be fused;IRectHgt1 is centre
The height of integration region;IRectHgt2 is the height in region to be fused;DDelta is fusion threshold value.
As can be seen from the above formula that formula (4) indicates the machine recognition result of intermediate integration region and region to be fused
It is identical, for example, being identified as the logo that " runs quickly ";Formula (5) and formula (6) indicate the position of intermediate integration region and region to be fused
It sets close;Formula (7) and formula (8) indicate that the size in intermediate integration region and region to be fused is close.
When intermediate integration region and region to be fused meet simultaneously, machine recognition result is identical, position is close, size is close
When, confirm that intermediate integration region and region to be fused meet fusion conditions.
Step 110, according to the first logo confidence level of the second logo candidate region for generating fusion candidate region, logo position
Set the fusion confidence level of confidence level and the corresponding fusion candidate region of the second logo confidence calculations.
After the fusion by step 109, however it remains multiple fusion candidate regions, it is final to determine to need to continue screening
Logo.
Used recognizer all has respective advantage and disadvantage during vehicle-logo recognition, for example, CNN algorithm is to flat
Shifting, scaling, inclination or the deformation of his total form have height invariance;Adaboost algorithm is to accurate candidate regions
The recognition confidence in domain is higher, and the recognition confidence of the candidate region offset for position is slightly lower, the candidate region of erroneous detection
Recognition confidence it is lower.Certainly, being also not excluded for certain candidate regions has pole with the sample in training library in structure and form
Big similitude.Therefore, multiple obtained in the embodiment of the present application combination above process in order to improve the accuracy of vehicle-logo recognition
The fusion confidence level of each fusion candidate region of confidence calculations.
Merge the calculating process of confidence level are as follows:
Formula (9)
Wherein, t is the number for constituting the second logo candidate region of present fusion candidate region;FSndWeightIt (i) is i-th
Second logo confidence level of a second logo candidate region;FFstWeightIt (i) is the first vehicle of i-th of second logo candidate regions
Mark confidence level;FLocWeightIt (i) is the logo position confidence level of i-th of second logo candidate regions;α, β are respectively the first logo
The weight coefficient of confidence level and logo position confidence level, and alpha+beta=1;FWeightFor the fusion confidence of present fusion candidate region
Degree.
The fusion confidence level that candidate region is merged it can be seen from formula (9) is constitute the fusion candidate region second
The confidence level of logo candidate region cumulative and.Wherein, when calculating the confidence level of each the second logo candidate region, to table
The the first logo confidence level and logo position confidence level for showing the second logo candidate region and true logo similarity degree are weighted
Processing takes α=0.35 for example, in a preferred embodiment, β=0.65, and recognition effect is preferable.
Step 111, select the fusion highest fusion candidate region of confidence level as the logo identified.
It is also higher to merge a possibility that higher fusion candidate region of confidence level is logo.
The application is not necessarily to carry out precise positioning to logo it can be seen from foregoing description, but by being obtained to depth recognition
The a variety of confidence levels taken are weighted fusion, obtain final recognition result.The application is suitable for the logo under complex scene and knows
Not, discrimination is higher.
Corresponding with the embodiment of aforementioned automobile logo identification method, present invention also provides the embodiments of vehicle-logo recognition device.
The embodiment of the application vehicle-logo recognition device can be using on an electronic device.Installation practice can pass through software
It realizes, can also be realized by way of hardware or software and hardware combining.Taking software implementation as an example, as on a logical meaning
Device, be that corresponding computer program instructions are formed in the processor run memory by equipment where it.From hardware
For level, as shown in figure 4, for a kind of hardware structure diagram of the application vehicle-logo recognition device place equipment, in addition to shown in Fig. 4
Except processor, network interface and memory, the practical function of equipment in embodiment where device generally according to the equipment
Can, it can also include other hardware, this is repeated no more.
Referring to FIG. 5, for the structural schematic diagram of the vehicle-logo recognition device in the application one embodiment.Vehicle-logo recognition dress
It sets including training unit 501, acquiring unit 502, determination unit 503, detection unit 504, the first computing unit 505, second meter
Unit 506, screening unit 507, recognition unit 508, integrated unit 509, third computing unit 510 and selecting unit 511 are calculated,
Wherein:
Training unit 501, for using logo detection algorithm training logo classifier;
Acquiring unit 502, for obtaining the license plate location information in image to be detected;
Determination unit 503, favored area at the beginning of for determining logo according to the license plate location information;
Detection unit 504, the logo classifier for being obtained using training carry out logo inspection to favored area at the beginning of the logo
It surveys, obtains several first logos candidate region;
First computing unit 505, for calculating the first logo confidence level of each the first logo candidate region;
Second computing unit 506, for calculating corresponding logo position confidence according to the position of the first logo candidate region
Degree;
Screening unit 507, for being screened according to logo position confidence level from several first logos candidate region
Central axes closer first logo candidate region out apart from favored area at the beginning of the logo is as the second logo candidate region;
Recognition unit 508 obtains for being identified to second logo candidate region using machine learning algorithm
Second logo confidence level of two logo candidate regions;
Integrated unit 509 generates multiple fusion candidate regions for carrying out region fusion to multiple second logos candidate region
Domain;
Third computing unit 510, for the first logo according to the second logo candidate region for generating fusion candidate region
The fusion confidence level of the corresponding fusion candidate region of confidence level, logo position confidence level and the second logo confidence calculations;
Selecting unit 511, for selecting the fusion highest fusion candidate region of confidence level as the logo identified.
Further, first computing unit 505, specifically:
FFstWeight=K/P
Wherein,
K is the number that the Weak Classifier of the first logo candidate region is capable of detecting when in the logo classifier;
P is the total number of Weak Classifier in the logo classifier;
FFstWeightFor the first logo confidence level of first logo candidate region.
Further, second computing unit 506, specifically:
Wherein,
D is distance of the first logo candidate region central point to the first favored area central axes of logo;
W is the width of the first logo candidate region;
FLocWeightFor the logo position confidence level of the first logo candidate region.
Further, the integrated unit 509, comprising:
Original fusion region selection module was not merged and was not chosen as with other logo candidate regions for selecting one
Original fusion region of the second logo candidate region in original fusion region as new round mixing operation, the original fusion area
Domain is the first intermediate integration region when front-wheel mixing operation;
Region selection module to be fused, for selecting one to have neither part nor lot in when the second logo candidate region of front-wheel mixing operation
As region to be fused;
Data obtaining module, for obtain respectively the position of the intermediate integration region and the region to be fused, width,
Height and machine recognition result;
Threshold calculation module, for calculating current fusion threshold value;
Judgment module is merged, for according to the fusion threshold value, the intermediate integration region and the region to be fused
Position, width, height and machine recognition result judge whether the intermediate integration region and the region to be fused meet and melt
Conjunction condition;
Region Fusion Module will for when the intermediate integration region and the region to be fused meet fusion conditions
The intermediate integration region and the region to be fused are merged, as new intermediate integration region;
As a result judgment module, for judging whether that each second logo candidate region has participated in grasping when the fusion of front-wheel
Make;If it is not, then executing region selection module to be fused;If so, judging whether also to be not chosen as the second of original fusion region
Logo candidate region;If nothing, there is currently intermediate integration region be fusion candidate region;If so, then executing original fusion
Region selection module.
Further, the threshold calculation module, specifically:
DDelta=θ × MIN (iRectWdt1, iRectWdt2)
Wherein,
θ is threshold value adjustment factor;
IRectWdt1 is the width of intermediate integration region;
IRectWdt2 is the width in region to be fused;
MIN (iRectWdt1, iRectWdt2) is the minimum value for taking intermediate integration region width and peak width to be fused;
DDelta is fusion threshold value.
Further, the fusion conditions are as follows:
IType1=iType2
|iRectX1-iRectX2|≤dDelta
|iRectY1-iRectY2|≤dDelta
|iRectX1+iRectWdt1-iRectX2-iRectWdt2|≤dDelta
|iRectY1+iRectHgt1-iRectY2-iRectHgt2|≤dDelta
Wherein,
IType1 is the machine recognition result of intermediate integration region;
IType2 is the machine recognition result in region to be fused;
(iRectX1, iRectY1) is the position of intermediate integration region;
(iRectX2, iRectY2) is the position in region to be fused;
IRectWdt1 is the width of intermediate integration region;
IRectWdt2 is the width in region to be fused;
IRectHgt1 is the height of intermediate integration region;
IRectHgt2 is the height in region to be fused;
DDelta is fusion threshold value.
Further, the third computing unit 510, specifically:
Wherein,
T is the number for constituting the second logo candidate region of present fusion candidate region;
FSndWeightIt (i) is the second logo confidence level of i-th of second logo candidate regions;
FFstWeightIt (i) is the first logo confidence level of i-th of second logo candidate regions;
FLocWeightIt (i) is the logo position confidence level of i-th of second logo candidate regions;
α, β are respectively the weight coefficient of the first logo confidence level and logo position confidence level, and alpha+beta=1;
FWeightFor the fusion confidence level of present fusion candidate region.
The function of each unit and the realization process of effect are specifically detailed in the above method and correspond to step in above-mentioned apparatus
Realization process, details are not described herein.
For device embodiment, since it corresponds essentially to embodiment of the method, so related place is referring to method reality
Apply the part explanation of example.The apparatus embodiments described above are merely exemplary, wherein described be used as separation unit
The unit of explanation may or may not be physically separated, and component shown as a unit can be or can also be with
It is not physical unit, it can it is in one place, or may be distributed over multiple network units.It can be according to actual
The purpose for needing to select some or all of the modules therein to realize application scheme.Those of ordinary skill in the art are not paying
Out in the case where creative work, it can understand and implement.
The foregoing is merely the preferred embodiments of the application, not to limit the application, all essences in the application
Within mind and principle, any modification, equivalent substitution, improvement and etc. done be should be included within the scope of the application protection.
Claims (14)
1. a kind of automobile logo identification method, which is characterized in that this method comprises:
Using logo detection algorithm training logo classifier;
Obtain the license plate location information in image to be detected;
Favored area at the beginning of determining logo according to the license plate location information;
The logo classifier obtained using training carries out logo detection to favored area at the beginning of the logo, obtains several first logos and waits
Favored area;
Calculate the first logo confidence level of each the first logo candidate region;
Corresponding logo position confidence level is calculated according to the position of the first logo candidate region;
It is filtered out from several first logos candidate region apart from constituency at the beginning of the logo according to logo position confidence level
The central axes in domain closer first logo candidate region is as the second logo candidate region;
Second logo candidate region is identified using machine learning algorithm, obtains the second of the second logo candidate region
Logo confidence level;
According to position, width, height and the machine recognition of the second logo candidate region as a result, to meeting the second of fusion conditions
Logo candidate region is merged, and fusion candidate region is generated;
According to generate the first logo confidence level of the second logo candidate region of fusion candidate region, logo position confidence level and
The fusion confidence level of the corresponding fusion candidate region of second logo confidence calculations;
Select the fusion highest fusion candidate region of confidence level as the logo identified.
2. the method as described in claim 1, which is characterized in that first vehicle for calculating each the first logo candidate region
Mark confidence level, comprising:
The calculation method of first logo confidence level of each first logo candidate region is identical, specifically:
FFstWeight=K/P
Wherein,
K is the number that the Weak Classifier of the first logo candidate region is capable of detecting when in the logo classifier;
P is the total number of Weak Classifier in the logo classifier;
FFstWeightFor the first logo confidence level of first logo candidate region.
3. the method as described in claim 1, which is characterized in that described calculated according to the position of the first logo candidate region corresponds to
Logo position confidence level, comprising:
Wherein,
D is distance of the first logo candidate region central point to the first favored area central axes of logo;
W is the width of the first logo candidate region;
FLocWeightFor the logo position confidence level of the first logo candidate region.
4. the method as described in claim 1, which is characterized in that the position according to the second logo candidate region, width, height
As a result, merging to the second logo candidate region for meeting fusion conditions, candidate region is merged in generation for degree and machine recognition,
Include:
Execute new round mixing operation: selection one did not merged with other logo candidate regions and was not chosen as original fusion area
Original fusion region of the second logo candidate region in domain as new round mixing operation, the original fusion region are to work as front-wheel
The first intermediate integration region of mixing operation;
It executes and works as front-wheel mixing operation: selecting one to have neither part nor lot in the second logo candidate region when front-wheel mixing operation as wait melt
Close region;Position, width, height and the machine recognition knot of the intermediate integration region and the region to be fused are obtained respectively
Fruit;Calculate current fusion threshold value;According to the fusion threshold value, the position of the intermediate integration region and the region to be fused
It sets, width, height and machine recognition result judge whether the intermediate integration region and the region to be fused meet fusion
Condition;When the intermediate integration region and the region to be fused meet fusion conditions, by the intermediate integration region and institute
It states region to be fused to be merged, as new intermediate integration region;
Judge whether that each second logo candidate region has participated in the mixing operation when front-wheel;It is executed currently if it is not, returning
Take turns mixing operation;If so, judging whether the second logo candidate region for being also not chosen as original fusion region;If nothing, when
Preceding existing intermediate integration region is fusion candidate region;New round mixing operation is executed if so, then returning.
5. method as claimed in claim 4, which is characterized in that described to calculate current fusion threshold value, comprising:
DDelta=θ × MIN (iRectWdt1, iRectWdt2)
Wherein,
θ is threshold value adjustment factor;
IRectWdt1 is the width of intermediate integration region;
IRectWdt2 is the width in region to be fused;
MIN (iRectWdt1, iRectWdt2) is the minimum value for taking intermediate integration region width and peak width to be fused;
DDelta is fusion threshold value.
6. method as claimed in claim 4, which is characterized in that the fusion conditions are as follows:
IType1=iType2
|iRectX1-iRectX2|≤dDelta
|iRectY1-iRectY2|≤dDelta
|iRectX1+iRectWdt1-iRectX2-iRectWdt2|≤dDelta
|iRectY1+iRectHgt1-iRectY2-iRectHgt2|≤dDelta
Wherein,
IType1 is the machine recognition result of intermediate integration region;
IType2 is the machine recognition result in region to be fused;
(iRectX1, iRectY1) is the position of intermediate integration region;
(iRectX2, iRectY2) is the position in region to be fused;
IRectWdt1 is the width of intermediate integration region;
IRectWdt2 is the width in region to be fused;
IRectHgt1 is the height of intermediate integration region;
IRectHgt2 is the height in region to be fused;
DDelta is fusion threshold value.
7. the method as described in claim 1, which is characterized in that described candidate according to the second logo for generating fusion candidate region
Melt the corresponding fusion candidate region of first logo confidence level, logo position confidence level and the second logo confidence calculations in region
Close confidence level, comprising:
Wherein,
T is the number for constituting the second logo candidate region of present fusion candidate region;
FSndWeightIt (i) is the second logo confidence level of i-th of second logo candidate regions;
FFstWeightIt (i) is the first logo confidence level of i-th of second logo candidate regions;
FLocWeightIt (i) is the logo position confidence level of i-th of second logo candidate regions;
α, β are respectively the weight coefficient of the first logo confidence level and logo position confidence level, and alpha+beta=1;
FweightFor the fusion confidence level of present fusion candidate region.
8. a kind of vehicle-logo recognition device, which is characterized in that the device includes:
Training unit, for using logo detection algorithm training logo classifier;
Acquiring unit, for obtaining the license plate location information in image to be detected;
Determination unit, favored area at the beginning of for determining logo according to the license plate location information;
Detection unit, the logo classifier for being obtained using training are carried out logo detection to favored area at the beginning of the logo, obtained
Several first logos candidate region;
First computing unit, for calculating the first logo confidence level of each the first logo candidate region;
Second computing unit, for calculating corresponding logo position confidence level according to the position of the first logo candidate region;
Screening unit, for being filtered out from several first logos candidate region apart from institute according to logo position confidence level
The central axes closer first logo candidate region of the first favored area of logo is stated as the second logo candidate region;
Recognition unit obtains the second logo for identifying using machine learning algorithm to second logo candidate region
Second logo confidence level of candidate region;
Integrated unit, for according to position, width, height and the machine recognition of the second logo candidate region as a result, to satisfaction
Second logo candidate region of fusion conditions is merged, and fusion candidate region is generated;
Third computing unit, the first logo confidence level of the second logo candidate region for merging candidate region according to generation,
Logo position confidence level and the fusion confidence level of the corresponding fusion candidate region of the second logo confidence calculations;
Selecting unit, for selecting the fusion highest fusion candidate region of confidence level as the logo identified.
9. device as claimed in claim 8, which is characterized in that first computing unit, specifically:
FFstWeight=K/P
Wherein,
K is the number that the Weak Classifier of the first logo candidate region is capable of detecting when in the logo classifier;
P is the total number of Weak Classifier in the logo classifier;
FFstWeightFor the first logo confidence level of first logo candidate region.
10. device as claimed in claim 8, which is characterized in that second computing unit, specifically:
Wherein,
D is distance of the first logo candidate region central point to the first favored area central axes of logo;
W is the width of the first logo candidate region;
FLocWeightFor the logo position confidence level of the first logo candidate region.
11. device as claimed in claim 8, which is characterized in that the integrated unit, comprising:
Original fusion region selection module, for selecting one not merge and be not chosen as with other logo candidate regions initially
Original fusion region of the second logo candidate region of integration region as new round mixing operation, the original fusion region is
When the first intermediate integration region of front-wheel mixing operation;
Region selection module to be fused, for select one have neither part nor lot in when the second logo candidate region of front-wheel mixing operation as
Region to be fused;
Data obtaining module, for obtaining position, the width, height of the intermediate integration region and the region to be fused respectively
And machine recognition result;
Threshold calculation module, for calculating current fusion threshold value;
Merge judgment module, for according to the position of the fusion threshold value, the intermediate integration region and the region to be fused,
Width, height and machine recognition result judge whether the intermediate integration region and the region to be fused meet fusion item
Part;
Region Fusion Module will be described for when the intermediate integration region and the region to be fused meet fusion conditions
Intermediate integration region and the region to be fused are merged, as new intermediate integration region;
As a result judgment module, for judging whether that each second logo candidate region has participated in the mixing operation when front-wheel;
If it is not, then executing region selection module to be fused;If so, judging whether the second logo for being also not chosen as original fusion region
Candidate region;If nothing, there is currently intermediate integration region be fusion candidate region;If so, then executing original fusion region
Selecting module.
12. device as claimed in claim 11, which is characterized in that the threshold calculation module, specifically:
DDelta=θ × MIN (iRectWdt1, iRectWdt2)
Wherein,
θ is threshold value adjustment factor;
IRectWdt1 is the width of intermediate integration region;
IRectWdt2 is the width in region to be fused;
MIN (iRectWdt1, iRectWdt2) is the minimum value for taking intermediate integration region width and peak width to be fused;
DDelta is fusion threshold value.
13. device as claimed in claim 11, which is characterized in that the fusion conditions are as follows:
IType1=iType2
|iRectX1-iRectX2|≤dDelta
|iRectY1-iRectY2|≤dDelta
|iRectX1+iRectWdt1-iRectX2-iRectWdt2|≤dDelta
|iRectY1+iRectHgt1-iRectY2-iRectHgt2|≤dDelta
Wherein,
IType1 is the machine recognition result of intermediate integration region;
IType2 is the machine recognition result in region to be fused;
(iRectX1, iRectY1) is the position of intermediate integration region;
(iRectX2, iRectY2) is the position in region to be fused;
IRectWdt1 is the width of intermediate integration region;
IRectWdt2 is the width in region to be fused;
IRectHgt1 is the height of intermediate integration region;
IRectHgt2 is the height in region to be fused;
DDelta is fusion threshold value.
14. device as claimed in claim 8, which is characterized in that the third computing unit, specifically:
Wherein,
T is the number for constituting the second logo candidate region of present fusion candidate region;
FSndWeightIt (i) is the second logo confidence level of i-th of second logo candidate regions;
FFstWeightIt (i) is the first logo confidence level of i-th of second logo candidate regions;
FLocWeightIt (i) is the logo position confidence level of i-th of second logo candidate regions;
α, β are respectively the weight coefficient of the first logo confidence level and logo position confidence level, and alpha+beta=1;
FWeightFor the fusion confidence level of present fusion candidate region.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510586228.8A CN105205486B (en) | 2015-09-15 | 2015-09-15 | A kind of automobile logo identification method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510586228.8A CN105205486B (en) | 2015-09-15 | 2015-09-15 | A kind of automobile logo identification method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105205486A CN105205486A (en) | 2015-12-30 |
CN105205486B true CN105205486B (en) | 2018-12-07 |
Family
ID=54953158
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510586228.8A Active CN105205486B (en) | 2015-09-15 | 2015-09-15 | A kind of automobile logo identification method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105205486B (en) |
Families Citing this family (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105608441B (en) * | 2016-01-13 | 2020-04-10 | 浙江宇视科技有限公司 | Vehicle type recognition method and system |
CN105957071B (en) * | 2016-04-26 | 2019-04-12 | 浙江宇视科技有限公司 | A kind of lamp group localization method and device |
CN106339445B (en) * | 2016-08-23 | 2019-06-18 | 东方网力科技股份有限公司 | Vehicle retrieval method and device based on big data |
CN106529424B (en) * | 2016-10-20 | 2019-01-04 | 中山大学 | A kind of logo detection recognition method and system based on selective search algorithm |
CN106503710A (en) * | 2016-10-26 | 2017-03-15 | 北京邮电大学 | A kind of automobile logo identification method and device |
CN106529460A (en) * | 2016-11-03 | 2017-03-22 | 贺江涛 | Object classification identification system and identification method based on robot side |
CN108256404B (en) * | 2016-12-29 | 2021-12-10 | 北京旷视科技有限公司 | Pedestrian detection method and device |
CN107590492B (en) * | 2017-08-28 | 2019-11-19 | 浙江工业大学 | A kind of vehicle-logo location and recognition methods based on convolutional neural networks |
CN108229308A (en) * | 2017-11-23 | 2018-06-29 | 北京市商汤科技开发有限公司 | Recongnition of objects method, apparatus, storage medium and electronic equipment |
CN108171274B (en) * | 2018-01-17 | 2019-08-09 | 百度在线网络技术(北京)有限公司 | The method and apparatus of animal for identification |
CN109409159A (en) * | 2018-10-11 | 2019-03-01 | 上海亿保健康管理有限公司 | A kind of fuzzy two-dimensional code detection method and device |
CN109919154B (en) * | 2019-02-28 | 2020-10-13 | 北京科技大学 | Intelligent character recognition method and device |
CN109697719B (en) * | 2019-03-05 | 2021-12-24 | 北京康夫子健康技术有限公司 | Image quality evaluation method and device and computer readable storage medium |
CN112069862A (en) * | 2019-06-10 | 2020-12-11 | 华为技术有限公司 | Target detection method and device |
CN110852252B (en) * | 2019-11-07 | 2022-12-02 | 厦门市美亚柏科信息股份有限公司 | Vehicle weight-removing method and device based on minimum distance and maximum length-width ratio |
CN113470347B (en) * | 2021-05-20 | 2022-07-26 | 上海天壤智能科技有限公司 | Congestion identification method and system combining bayonet vehicle passing record and floating vehicle GPS data |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN102968646A (en) * | 2012-10-25 | 2013-03-13 | 华中科技大学 | Plate number detecting method based on machine learning |
CN103077384A (en) * | 2013-01-10 | 2013-05-01 | 北京万集科技股份有限公司 | Method and system for positioning and recognizing vehicle logo |
CN103310231A (en) * | 2013-06-24 | 2013-09-18 | 武汉烽火众智数字技术有限责任公司 | Auto logo locating and identifying method |
CN104268596A (en) * | 2014-09-25 | 2015-01-07 | 深圳市捷顺科技实业股份有限公司 | License plate recognizer and license plate detection method and system thereof |
CN104281851A (en) * | 2014-10-28 | 2015-01-14 | 浙江宇视科技有限公司 | Extraction method and device of car logo information |
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 |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6735337B2 (en) * | 2001-02-02 | 2004-05-11 | Shih-Jong J. Lee | Robust method for automatic reading of skewed, rotated or partially obscured characters |
-
2015
- 2015-09-15 CN CN201510586228.8A patent/CN105205486B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN102968646A (en) * | 2012-10-25 | 2013-03-13 | 华中科技大学 | Plate number detecting method based on machine learning |
CN103077384A (en) * | 2013-01-10 | 2013-05-01 | 北京万集科技股份有限公司 | Method and system for positioning and recognizing vehicle logo |
CN103310231A (en) * | 2013-06-24 | 2013-09-18 | 武汉烽火众智数字技术有限责任公司 | Auto logo locating and identifying method |
CN104268596A (en) * | 2014-09-25 | 2015-01-07 | 深圳市捷顺科技实业股份有限公司 | License plate recognizer and license plate detection method and system thereof |
CN104281851A (en) * | 2014-10-28 | 2015-01-14 | 浙江宇视科技有限公司 | Extraction method and device of car logo information |
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 |
Also Published As
Publication number | Publication date |
---|---|
CN105205486A (en) | 2015-12-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105205486B (en) | A kind of automobile logo identification method and device | |
CN108319964B (en) | Fire image recognition method based on mixed features and manifold learning | |
CN106407981B (en) | A kind of licence plate recognition method, apparatus and system | |
Zhou et al. | Principal visual word discovery for automatic license plate detection | |
CN103679168B (en) | Detection method and detection device for character region | |
CN101799867B (en) | Improved detection of people in real world videos and images | |
CN106682586A (en) | Method for real-time lane line detection based on vision under complex lighting conditions | |
CN107103317A (en) | Fuzzy license plate image recognition algorithm based on image co-registration and blind deconvolution | |
CN105740886B (en) | A kind of automobile logo identification method based on machine learning | |
Lee et al. | Shape discovery from unlabeled image collections | |
CN109255375A (en) | Panoramic picture method for checking object based on deep learning | |
CN106355138A (en) | Face recognition method based on deep learning and key features extraction | |
CN103366181A (en) | Method and device for identifying scene integrated by multi-feature vision codebook | |
CN106778687A (en) | Method for viewing points detecting based on local evaluation and global optimization | |
CN106682641A (en) | Pedestrian identification method based on image with FHOG- LBPH feature | |
CN110263712A (en) | A kind of coarse-fine pedestrian detection method based on region candidate | |
CN105224945B (en) | A kind of automobile logo identification method based on joint-detection and identification algorithm | |
CN106250824A (en) | Vehicle window localization method and system | |
CN105279492A (en) | Iris identification method and device | |
CN106503710A (en) | A kind of automobile logo identification method and device | |
CN101984453A (en) | Human eye recognition system and method | |
Ruta et al. | Towards real-time traffic sign recognition by class-specific discriminative features | |
CN106056132A (en) | Garment style identification method based on Fourier descriptor and support vector machine | |
CN107886066A (en) | A kind of pedestrian detection method based on improvement HOG SSLBP | |
CN104156690A (en) | Gesture recognition method based on image space pyramid bag of features |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |