CN105160333B - A kind of model recognizing method and identification device - Google Patents
A kind of model recognizing method and identification device Download PDFInfo
- Publication number
- CN105160333B CN105160333B CN201510639752.7A CN201510639752A CN105160333B CN 105160333 B CN105160333 B CN 105160333B CN 201510639752 A CN201510639752 A CN 201510639752A CN 105160333 B CN105160333 B CN 105160333B
- Authority
- CN
- China
- Prior art keywords
- picture
- target vehicle
- weights
- detected
- feature
- 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
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
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)
- Traffic Control Systems (AREA)
Abstract
An embodiment of the present invention provides a kind of model recognizing method, the method includes:Obtain picture to be detected;The picture to be detected is detected using the first default grader;If containing target vehicle in the picture to be detected, the target vehicle in the picture to be detected is extracted;Registration process is carried out to the target vehicle, so that the angle between the headstock direction and the vertical direction of the target area of the target vehicle is less than predetermined threshold value;Feature extraction is carried out to the target vehicle after the registration process, to obtain M feature, the M is the integer more than 1;Classified to the M feature using the second default grader;The vehicle of the target vehicle is determined according to the result of the classification.The embodiment of the present invention additionally provides a kind of identification device, for identification vehicle.The precision of vehicle cab recognition can be improved through the embodiment of the present invention.
Description
Technical field
The present embodiments relate to intelligent monitoring technology fields, and in particular to a kind of model recognizing method and identification device.
Background technology
In intelligent video monitoring, vehicle cab recognition technology is at an early period of the investigation of public security image and traffic state analysis
Important fission in reason.With the development of information technology, vehicle cab recognition technology has also obtained further development, current vehicle
Identification technology rather than traditional vehicle cab recognition, wherein traditional vehicle cab recognition is only capable of telling the big of vehicle
Type is caused, such as dilly, medium sized vehicle and oversize vehicle.And the vehicle cab recognition technology in meaning is exactly to from vehicle at present
The vehicle feature extracted in vehicle face area image is classified, to determine the brand and model belonging to the vehicle.With computer
The progress of technology, the vehicle cab recognition technology based on vehicle face feature gradually move towards practical, can not only identify vehicle brand, or even can
To identify the series and year money under the vehicle brand, to greatly expand application of the technology in related field.
Currently, although the identification technology based on vehicle feature has been achieved for many achievements in research, due to vehicle kind
Class is various and contour of the vehicle updating decision, while preceding face difference is small between the vehicle of part, to reduce the accurate of vehicle cab recognition.
Invention content
An embodiment of the present invention provides a kind of model recognizing method and identification devices, to improve the identification essence of vehicle cab recognition
Degree.
First aspect of the embodiment of the present invention provides a kind of model recognizing method, including:
Obtain picture to be detected;
The picture to be detected is detected using the first default grader;
If containing target vehicle in the picture to be detected, the target vehicle in the picture to be detected is extracted;
Registration process is carried out to the target vehicle, so that the headstock direction of the target vehicle and the target area
Vertical direction between angle be less than predetermined threshold value;
Feature extraction is carried out to the target vehicle after the registration process, to obtain M feature, the M is more than 1
Integer;
Classified to the M feature using the second default grader;
The vehicle of the target vehicle is determined according to the result of the classification.
It is described to adopt in conjunction with the embodiment of the present invention in a first aspect, in the first possible embodiment of first aspect
With the first default grader to the picture to be detected be detected including:
Q pictures of the P pictures of vehicle as positive sample and without vehicle will be contained in picture library as negative sample pair
The picture to be detected is detected, wherein the P and Q is the integer more than 1.
In conjunction with the embodiment of the present invention in a first aspect, in second of possible embodiment of first aspect, it is described will
The target vehicle carries out registration process:
Determine the corresponding affine transformation matrix of the target vehicle;
Affine transformation is carried out to the target vehicle according to the affine transformation matrix.
In conjunction with the first aspect of the embodiment of the present invention or second of possible embodiment of first aspect, in first aspect
The third possible embodiment in, the corresponding affine transformation matrix of the determination target vehicle includes:
Obtain the K picture that user selects from picture library, the vehicle for the vehicle for including in the picture i in the K picture
Angle between head direction and the vertical direction of the picture i is less than the predetermined threshold value, wherein the picture i is the K
Any picture in picture, the K are the integer more than or equal to 1;
Extract O feature respectively from the K picture, the O is the integer more than 1;
Vector regression processing is supported to the K*O feature extracted, to obtain affine transformation matrix.
It is described right in conjunction with the embodiment of the present invention in a first aspect, in the 4th kind of possible embodiment of first aspect
Target area after the registration process carries out feature extraction, includes to obtain M feature:
The target vehicle is divided into the region of K1 the first default sizes, the region of a second default sizes of K2 respectively
The region of size is preset with K3 third, wherein the K1, the K2 and the K3 are to be worth mutually different positive integer;
The corresponding K1 histogram in region, the K2 a second for obtaining the K1 the first default sizes respectively is default big
The corresponding K2 histogram in small region and the K3 third preset the corresponding K3 histogram in region of size;
K1 weights of the K1 histogram, K2 weights of the K2 histogram and institute are configured according to preset order
State K3 weights of K3 histogram, the K1 weights are all higher than 0 and equal, and the K2 weights are all higher than 0 and equal, institute
It states K3 weights and is all higher than 0 and equal, and the K1 weights and the K2 weights and the K3 weights sum are 1;
The histograms of oriented gradients HOG features for extracting the target vehicle respectively, to obtain presetting with the K1 first
The corresponding K1 HOG feature in region of size, the K2 HOG corresponding with the K2 regions of the second default size
Feature and the K3 HOG feature corresponding with the default region of size of the K3 third;
By the K1 weights and K1 HOG features, the K2 weights and the K2 HOG features and the K3
A weights and K3 HOG features composition characteristic vector;
It uses third to preset grader to be trained described eigenvector to obtain M feature.
Second aspect of the embodiment of the present invention provides a kind of identification device, including:
First acquisition unit, for obtaining picture to be detected;
Detection unit, picture to be detected for being got to the first acquisition unit using the first default grader into
Row detection;
First extraction unit contains target vehicle, extraction if being detected for the detection unit in the picture to be detected
Target vehicle in the picture to be detected;
First processing units, the target vehicle for being extracted to first extraction unit carry out registration process, so that
Angle between the headstock direction and the vertical direction of the target area of the target vehicle is less than predetermined threshold value;
Second extraction unit is carried for carrying out feature to the target vehicle after the first processing units registration process
It takes, to obtain M feature, the M is the integer more than 1;
Taxon, the M feature for being extracted to second extraction unit using the second default grader are divided
Class;
First determination unit, the vehicle for determining the target vehicle according to the result of the classification of the taxon.
In conjunction with the second aspect of the embodiment of the present invention, in the first possible embodiment of second aspect, the inspection
Unit is surveyed to be specifically used for:
Q pictures of the P pictures of vehicle as positive sample and without vehicle will be contained in picture library as negative sample pair
The picture to be detected is detected, wherein the P and Q is the integer more than 1.
In conjunction with the second aspect of the embodiment of the present invention, in the first possible embodiment of second aspect, described
One processing unit includes:
Second determination unit determines the corresponding affine transformation matrix of target vehicle of the first extraction unit extraction;
Affine transformation unit, the affine transformation matrix for being determined according to second determination unit is to the target vehicle
Carry out affine transformation.
In conjunction with the second aspect of the embodiment of the present invention or second of possible embodiment of second aspect, in second aspect
The third possible embodiment in, second determination unit includes:
Second acquisition unit, the K picture selected from picture library for obtaining user, the picture i in the K picture
In include vehicle headstock direction and the vertical direction of the picture i between angle be less than the predetermined threshold value, wherein institute
It is any picture in the K picture to state picture i, and the K is the integer more than or equal to 1;
Third extraction unit, for extracting O feature, institute respectively from the K picture that the second acquisition unit obtains
It is the integer more than 1 to state O;
Second processing unit, the K*O feature for being extracted to the third extraction unit are supported vector regression
Processing, to obtain affine transformation matrix.
In conjunction with the second aspect of the embodiment of the present invention, in the 4th kind of possible embodiment of second aspect, described
Two extraction units include:
The target vehicle that first extraction unit extracts is divided into K1 first default big by division unit for respectively
Small region, K2 the second default sizes region and the default size of K3 third region, wherein the K1, the K2 and
The K3 is to be worth mutually different positive integer;
Third acquiring unit, the region for obtaining the K1 the first default sizes that the division unit divides respectively
The corresponding K2 histogram in region of a second default size of corresponding K1 histogram, the K2 and the K3 third are default
The corresponding K3 histogram in region of size;
The K1 of dispensing unit, the K1 histogram for being got according to preset order configuration third acquiring unit is a
Weights, the K2 histogram K2 weights and the K3 histogram K3 weights, the K1 weights be all higher than 0 and
Equal, the K2 weights are all higher than 0 and equal, and the K3 weights are all higher than 0 and equal, and the K1 weights with it is described
K2 weights and the K3 weights sum are 1;
4th extraction unit, the direction gradient histogram of the target vehicle for extracting the first extraction unit extraction respectively
HOG features are schemed, to obtain and the K1 corresponding K1 HOG features in the regions of the first default size and the K2
The corresponding K2 HOG feature in region of second default size and K3 corresponding with the default region of size of the K3 third
A HOG features;
Component units, for the K1 weights and the K1 HOG features, the K2 weights and the K2 are a
HOG features and the K3 weights and K3 HOG features composition characteristic vector;
Training unit, for use third preset grader feature vector that the component units are formed be trained with
Obtain M feature.
Implement the embodiment of the present invention, has the advantages that:
Picture to be detected is obtained through the embodiment of the present invention;The picture to be detected is carried out using the first default grader
Detection;If containing target vehicle in the picture to be detected, the target vehicle in the picture to be detected is extracted;To the target carriage
Carry out registration process so that the angle between the headstock direction and the vertical direction of the target area of the target vehicle
Less than predetermined threshold value;Feature extraction is carried out to the target vehicle after the registration process, to obtain M feature, the M is big
In 1 integer;Classified to the M feature using the second default grader;According to the determination of the result of the classification
The vehicle of target vehicle.In the embodiment of the present invention, target vehicle is extracted, registration process is carried out to the target vehicle, and in the base
Feature extraction is carried out to the target vehicle after registration process on plinth, and classification processing is carried out to the feature extracted, thus, it can
Improve the precision of vehicle cab recognition.
Description of the drawings
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are some embodiments of the invention, for ability
For the those of ordinary skill of domain, without creative efforts, it can also be obtained according to these attached drawings other attached
Figure.
Fig. 1 is a kind of first embodiment flow diagram of model recognizing method provided in an embodiment of the present invention;
Fig. 2 is a kind of second embodiment flow diagram of model recognizing method provided in an embodiment of the present invention;
Fig. 3 is a kind of 3rd embodiment flow diagram of model recognizing method provided in an embodiment of the present invention;
Fig. 4 a are a kind of first embodiment structural schematic diagrams of identification device provided in an embodiment of the present invention;
Fig. 4 b are a kind of another structural schematic diagrams of first embodiment of identification device provided in an embodiment of the present invention;
Fig. 4 c are a kind of another structural schematic diagrams of first embodiment of identification device provided in an embodiment of the present invention;
Fig. 5 is a kind of second embodiment structural schematic diagram of identification device provided in an embodiment of the present invention.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, the every other implementation that those of ordinary skill in the art are obtained without creative efforts
Example, shall fall within the protection scope of the present invention.
The identification device of vehicle cab recognition described by the embodiment of the present invention may include video matrix, monitoring device, with insighted
Other devices etc. of other vehicle function, above-mentioned identification device are only citings, and non exhaustive, are filled including but not limited to above-mentioned identification
It sets.
Further, model recognizing method described in the embodiment of the present invention can be additionally used in other target identifications, and such as people moves
Object or other objects, repeat no more in this embodiment.
Referring to Fig. 1, being a kind of first embodiment flow diagram of model recognizing method provided in an embodiment of the present invention.
The method of target detection described in the present embodiment, includes the following steps:
S101, picture to be detected is obtained.
In the specific implementation, picture to be detected can be an independent picture, alternatively, picture to be detected can be video a frame or
Person's multiframe picture, wherein may include one or more vehicle even 0 vehicle in picture to be detected.The vehicle may include but
It is not limited only to:Compact car, micro- vehicle, compact vehicle, medium vehicle, advanced vehicle, deluxe carmodel, three-box car type, CDV vehicles,
MPV vehicles, SUV etc..The vehicle can also be:Audi Q7, BMW X3, Audi A8, Audi A4, Audi A4L etc..
S102, the picture to be detected is detected using the first default grader.
In the specific implementation, the first default grader, which can be multiple graders, carries out cascade cascade classifier, it is preferable that the
One default grader can be histograms of oriented gradients (English:Histogram of Gradient, abbreviation:) and supporting vector HOG
Machine (English:Support Vector Machine, abbreviation:SVM) the cascade classifier formed.Utilize the first default grader
Picture to be detected is detected, to which target area can be obtained.Optionally, identification device can will contain vehicle in picture library
Q pictures of the P pictures as positive sample and without vehicle the picture to be detected is detected as negative sample,
In, P and Q are the integer more than 1.
Optionally, the embodiment of the present invention can detect vehicle, but the case where capable of also being come together to more vehicles carries out
Detection, in specific implementation process, in order to improve the accuracy rate of identification, the case where can coming the multiple vehicles detected together, regards
For interference.Therefore, it to be excluded by every means.In embodiments of the present invention, more vehicles in a width picture can be come together
The case where referred to as more vehicles, the case where more vehicles, is then also considered as negative sample and is excluded.In implementation process, negative sample can be reduced
This size, to, the probability for excluding more vehicles is improved, for example, original positive and negative sample-size size is 64*128, specific implementation
In, the negative sample size containing more vehicles can be reduced to 60*120, in this way, can preferably exclude more vehicles simultaneously, to,
Vehicle can be accurately detected.
If containing target vehicle in S103, the picture to be detected, the target vehicle in the picture to be detected is extracted.
, can be to the mapping to be checked in the specific implementation, if identification device, which detects in picture to be detected, contains target vehicle
Piece carries out image segmentation, to extract the target vehicle in the identification device.Preferably, there was only 1 vehicle in picture to be detected
In the case of, image segmentation can be carried out to the picture, to extract the target vehicle in the picture.
S104, registration process is carried out to the target vehicle, so that the headstock direction of the target vehicle and the mesh
The angle marked between the vertical direction in region is less than predetermined threshold value.
In the specific implementation, identification device can carry out target vehicle registration process, i.e., the vehicle in target vehicle is carried out
Registration process, to so that the angle between the vertical direction in the headstock direction and goal region of the vehicle is less than predetermined threshold value,
The predetermined threshold value may include but be not limited only to:0.1 degree, 0.01 degree, 1 degree, 5 degree etc..
S105, feature extraction is carried out to the target vehicle after the registration process, to obtain M feature, the M is big
In 1 integer.
In the specific implementation, identification device can extract the feature in the target vehicle after registration process, it is assumed that be extracted M
Feature.Optionally, it can extract the histograms of oriented gradients feature of the target vehicle after registration process.
S106, classified to the M feature using the second default grader.
In the specific implementation, the second default grader can be cascade classifier or an individual grader, second is default
Grader may include but be not limited only to:Grader, the linear branch of support vector machines, adaboost graders, HOG and SVM cascade
Hold vector machine (English:Liblinear SVM) etc..Preferably, the embodiment of the present invention is found during the experiment, nonlinear
SVM has found that object function convergence is too slow when classifying to extensive target, and overall operation speed is excessively slow, therefore, excellent
Selection of land, the embodiment of the present invention choose linear liblinear svm as the second default grader.
S107, the type that the target vehicle is determined according to the result of the classification.
In the specific implementation, the result of classification can be seen that target vehicle is the probability of which vehicle, in the result of classification
In the case of the probability that multiple vehicles are determined, the corresponding vehicle of the value of maximum probability can be determined as to the vehicle of target vehicle.
Picture to be detected is obtained through the embodiment of the present invention;The picture to be detected is carried out using the first default grader
Detection;If containing target vehicle in the picture to be detected, the target vehicle in the picture to be detected is extracted;To the target carriage
Carry out registration process so that the angle between the headstock direction and the vertical direction of the target area of the target vehicle
Less than predetermined threshold value;Feature extraction is carried out to the target vehicle after the registration process, to obtain M feature, the M is big
In 1 integer;Classified to the M feature using the second default grader;According to the determination of the result of the classification
The vehicle of target vehicle.In the embodiment of the present invention, target vehicle is extracted, registration process is carried out to the target vehicle, and in the base
Feature extraction is carried out to the target vehicle after registration process on plinth, and classification processing is carried out to the feature extracted, thus, it can
Improve the precision of vehicle cab recognition.
Referring to Fig. 2, being a kind of second embodiment flow diagram of model recognizing method provided in an embodiment of the present invention.
Model recognizing method described in the present embodiment, includes the following steps:
S201, picture to be detected is obtained.
S202, the picture to be detected is detected using the first default grader.
If containing target vehicle in S203, the picture to be detected, the target vehicle in the picture to be detected is extracted.
S204, the corresponding affine transformation matrix of the target vehicle is determined.
In the specific implementation, identification device can determine the corresponding affine transformation matrix of target vehicle, specifically, identification device can
Obtain K picture being selected from picture library of user, the headstock direction for the vehicle for including in the picture i in the K picture and this
Angle between the vertical direction of picture i is less than predetermined threshold value, wherein picture i is any picture in K picture, K for more than
Or the integer equal to 1, extract O feature respectively from K picture, the O is the integer more than 1, to the K*O extracted
Feature is supported vector regression processing, to obtain affine transformation matrix.
S205, affine transformation is carried out to the target vehicle according to the affine transformation matrix, so that the target carriage
Headstock direction and the vertical direction of the target area between angle be less than predetermined threshold value.
In the specific implementation, identification device can carry out affine transformation according to the affine transformation matrix to target vehicle, thus, it can
So that the angle between the headstock direction and the vertical direction of target area of target vehicle after affine transformation is less than default threshold
Value.The predetermined threshold value may include but be not limited only to:0 degree, 1 degree, 0.5 degree, 0.32 degree etc..
S206, feature extraction is carried out to the target vehicle after the registration process, to obtain M feature, the M is big
In 1 integer.
S207, classified to the M feature using the second default grader.
S208, the vehicle that the target vehicle is determined according to the result of the classification.
In the embodiments of the present invention, target vehicle is extracted, to determining the affine transformation matrix of the target vehicle, according to this
Affine transformation matrix carries out registration process to target vehicle, and carries out feature to the target vehicle after registration process on this basis
Extraction, and classification processing is carried out to the feature extracted, to which the precision of vehicle cab recognition can be improved.
Referring to Fig. 3, being a kind of 3rd embodiment flow diagram of model recognizing method provided in an embodiment of the present invention.
Model recognizing method described in the present embodiment, includes the following steps:
S301, picture to be detected is obtained.
S302, the picture to be detected is detected using the first default grader.
If containing target vehicle in S303, the picture to be detected, the target vehicle in the picture to be detected is extracted.
S304, registration process is carried out to the target vehicle, so that the headstock direction of the target vehicle and the mesh
The angle marked between the vertical direction in region is less than predetermined threshold value.
S305, region, a second default sizes of K2 that the target vehicle is divided into K1 the first default sizes respectively
Region and K3 third preset the region of size, wherein the K1, the K2 and the K3 be worth it is mutually different just whole
Number.
In the specific implementation, target vehicle can be divided into the region of K1 the first default sizes by identification device, wherein example
Such as, K1 is positive integer, it is preferable that K1 may include but be not limited only to:1,4,9,16,25,36 etc..Identification device can be by target carriage
It is divided into the region of K2 the second default sizes, wherein for example, K2 is positive integer, it is preferable that K2 may include but not only limit
In:1,4,9,16,25,36 etc..Target vehicle can be divided into the region of K3 the second default sizes by identification device, wherein
For example, K3 is positive integer, it is preferable that K3 may include but be not limited only to:1,4,9,16,25,36 etc..Wherein, K1, K2 and K3 are mutual
It is unequal.
S306, the corresponding K1 histogram in region for obtaining the K1 the first default sizes respectively, the K2 a second
The corresponding K2 histogram in region of default size and the K3 third preset the corresponding K3 histogram in region of size.
In the specific implementation, identification device can obtain the corresponding K1 histogram in region of K1 the first default sizes respectively,
The corresponding K2 histogram in region of K2 the second default sizes, K3 third preset the corresponding K3 histogram in region of size
Figure.
The K2 power of S307, the K1 weights that the K1 histogram is configured according to preset order, the K2 histogram
K3 weights of value and the K3 histogram, the K1 weights are all higher than 0 and equal, the K2 weights be all higher than 0 and
Equal, the K3 weights are all higher than 0 and equal, and the K1 weights and the K2 weights and the K3 weights sum
It is 1.
In the specific implementation, identification device can configure K1 weights of K1 histogram, the K1 weights according to preset order
It is all higher than 0 and equal, similarly, identification device can configure K3 weights of the K2 weights and K3 histogram of K2 histogram,
Wherein, which is all higher than 0 and equal, and the K3 weights are all higher than 0 and equal, above-mentioned K1 weights, K2 power
Value and K3 weights sum are 1.Under illustration, in the case where K1 is 4,4 histograms, 4 histograms point can be obtained
Not Wei the first histogram, the second histogram, third histogram and the 4th histogram, which is numbered, for example, 1
Indicate that the first histogram, 2 indicate that the second histogram, 3 indicate that third histogram and 4 indicates the 4th histogram, then preset order can
It is 1,2,3 and 4, corresponding 4 weights are assumed to be W1, W2, W3 and W4, wherein the W1=W2=W3=W4 remembers W1+W2+W3+
W4=A1<1.
S308, the histograms of oriented gradients HOG features for extracting the target vehicle respectively, to obtain and the K1 first
The corresponding K1 HOG feature in region of default size, K2 institute corresponding with the K2 regions of the second default size
State HOG features and the K3 HOG feature corresponding with the default region of size of the K3 third.
S309, by the K1 weights and the K1 HOG features, the K2 weights and the K2 HOG features and
The K3 weights and K3 HOG features composition characteristic vector.
S310, it uses third to preset grader to be trained described eigenvector to obtain M feature, the M is big
In 1 integer.
Wherein, it can be that cascade classifier or an individual grader, third preset grader that third, which presets grader,
It may include but be not limited only to:Grader, the linear support vector of support vector machines, adaboost graders, HOG and SVM cascade
Machine etc..
S311, classified to the M feature using the second default grader.
S312, the vehicle that the target vehicle is determined according to the result of the classification.
In the embodiments of the present invention, target vehicle is extracted, to determining the affine transformation matrix of the target vehicle, according to this
Affine transformation matrix carries out registration process to target vehicle, and carries out feature to the target vehicle after registration process on this basis
Extraction, and classification processing is carried out to the feature extracted, to which the precision of vehicle cab recognition can be improved.
Please refer to Fig. 4 a- Fig. 4 c, wherein Fig. 4 a are a kind of first embodiment of identification device provided in an embodiment of the present invention
Structural schematic diagram.Identification device in Fig. 4 a described in the present embodiment, including:First acquisition unit 401, detection unit 402,
First extraction unit 403, first processing units 404, the second extraction unit 405, taxon 406 and the first determination unit 407,
It is specific as follows:
First acquisition unit 401, for obtaining picture to be detected.
Detection unit 402, it is to be detected for being got to the first acquisition unit 401 using the first default grader
Picture is detected.
As a kind of possible embodiment, detection unit 402 is specifically used for:
Q pictures of the P pictures of vehicle as positive sample and without vehicle will be contained in picture library as negative sample pair
The picture to be detected is detected, wherein the P and Q is the integer more than 1.
First extraction unit 403 contains target carriage if being detected for the detection unit 402 in the picture to be detected
, extract the target vehicle in the picture to be detected.
First processing units 404, the target vehicle for being extracted to first extraction unit 403 carry out registration process,
So that the angle between the headstock direction and the vertical direction of the target area of the target vehicle is less than predetermined threshold value.
Second extraction unit 405, it is special for being carried out to the target vehicle after 404 registration process of the first processing units
Sign extraction, to obtain M feature, the M is the integer more than 1.
Taxon 406, the M feature for being extracted to second extraction unit 405 using the second default grader
Classify.
First determination unit 407, the result for the classification according to the taxon 406 determine the target vehicle
Vehicle.
As a kind of possible embodiment, as shown in Figure 4 b, the first processing of the identification device described in Fig. 4 a is single
First 404 include:Second determination unit 4041 and affine transformation unit 4042, it is specific as follows:
Second determination unit 4041 determines the corresponding affine transformation square of target vehicle of the first extraction unit extraction
Battle array;
Affine transformation unit 4042, the affine transformation matrix for being determined according to second determination unit is to the target
Vehicle carries out affine transformation.
Further, the second determination unit 4041 may also include:
Second acquisition unit, the K picture selected from picture library for obtaining user, the picture i in the K picture
In include vehicle headstock direction and the vertical direction of the picture i between angle be less than the predetermined threshold value, wherein institute
It is any picture in the K picture to state picture i, and the K is the integer more than or equal to 1;
Third extraction unit, for extracting O feature, institute respectively from the K picture that the second acquisition unit obtains
It is the integer more than 1 to state O;
Second processing unit, the K*O feature for being extracted to the third extraction unit are supported vector regression
Processing, to obtain affine transformation matrix.
As a kind of possible embodiment, as illustrated in fig. 4 c, second of the identification device described in Fig. 4 a or Fig. 4 b
Extraction unit 405 includes:Division unit 4051, third acquiring unit 4052, dispensing unit 4053, the 4th extraction unit 4054,
Component units 4055 and training unit 4056, it is specific as follows:
Division unit 4051, for the target vehicle that first extraction unit extracts to be divided into K1 first in advance respectively
If the region in the region of size, the region of K2 the second default sizes and the default size of K3 third, wherein the K1, described
The K2 and K3 is to be worth mutually different positive integer;
Third acquiring unit 4052, for obtaining the K1 the first default sizes that the division unit divides respectively
The corresponding K2 histogram in region of a second default size of the corresponding K1 histogram in region, the K2 and the K3 third
The corresponding K3 histogram in region of default size;
Dispensing unit 4053, for according to the K1 histogram that gets of preset order configuration third acquiring unit
K1 weights, the K2 histogram K2 weights and the K3 histogram K3 weights, the K1 weights are big
In 0 and equal, the K2 weights are all higher than 0 and equal, and the K3 weights are all higher than 0 and equal, and the K1 weights
It is 1 with the K2 weights and the K3 weights sum;
4th extraction unit 4054, the direction gradient of the target vehicle for extracting the first extraction unit extraction respectively
Histogram HOG features, with obtain the K1 HOG features corresponding with the regions of the first default size the K1, with it is described
The corresponding K2 HOG feature in region of K2 the second default sizes and corresponding with the default region of size of the K3 third
K3 HOG features;
Component units 4055 are used for the K1 weights and K1 HOG features, the K2 weights and K2
A HOG features and the K3 weights and K3 HOG features composition characteristic vector;
Training unit 4056 is instructed for presetting the feature vector that grader forms the component units using third
Practice to obtain M feature.
Described identification device can obtain picture to be detected through the embodiment of the present invention;Using the first default grader pair
The picture to be detected is detected;If containing target vehicle in the picture to be detected, the mesh in the picture to be detected is extracted
Mark vehicle;Registration process is carried out to the target vehicle, so that the headstock direction of the target vehicle and the target area
Vertical direction between angle be less than predetermined threshold value;Feature extraction is carried out to the target vehicle after the registration process, with
M feature is obtained, the M is the integer more than 1;Classified to the M feature using the second default grader;According to institute
The result for stating classification determines the vehicle of the target vehicle.In the embodiment of the present invention, extract target vehicle, to the target vehicle into
Row registration process, and feature extraction is carried out to the target vehicle after registration process on this basis, and to the feature extracted
Classification processing is carried out, to which the precision of vehicle cab recognition can be improved.
Referring to Fig. 5, being a kind of second embodiment structural schematic diagram of identification device provided in an embodiment of the present invention.This reality
The identification device described in example is applied, including:At least one input equipment 1000;At least one output equipment 2000;At least one
A processor 3000, such as CPU;With memory 4000, above-mentioned input equipment 1000, output equipment 2000,3000 and of processor
Memory 4000 is connected by bus 5000.
Wherein, above-mentioned input equipment 1000 concretely physical button, touch tablet or mouse.
The concretely display screen of above-mentioned output equipment 2000.
Above-mentioned memory 4000 can be high-speed RAM memory or non-labile memory (non-volatile
), such as magnetic disk storage memory.Above-mentioned memory 4000 is above-mentioned input equipment 1000, defeated for storing batch processing code
Go out equipment 2000 and processor 3000 for calling the program code stored in memory 4000, executes following operation:
Above-mentioned input equipment 1000, is used for
Obtain picture to be detected;
The picture to be detected is detected using the first default grader;
If containing target vehicle in the picture to be detected, the target vehicle in the picture to be detected is extracted;
Registration process is carried out to the target vehicle, so that the headstock direction of the target vehicle and the target area
Vertical direction between angle be less than predetermined threshold value;
Feature extraction is carried out to the target vehicle after the registration process, to obtain M feature, the M is more than 1
Integer;
Classified to the M feature using the second default grader;
The vehicle of the target vehicle is determined according to the result of the classification.
As a kind of possible embodiment, above-mentioned processor 3000 uses the first default grader to the mapping to be checked
Piece be detected including:
Q pictures of the P pictures of vehicle as positive sample and without vehicle will be contained in picture library as negative sample pair
The picture to be detected is detected, wherein the P and Q is the integer more than 1.
As a kind of possible embodiment, target vehicle progress registration process is included by above-mentioned processor 3000:
Determine the corresponding affine transformation matrix of the target vehicle;
Affine transformation is carried out to the target vehicle according to the affine transformation matrix.
As a kind of possible embodiment, above-mentioned processor 3000 determines the corresponding affine transformation square of the target vehicle
Battle array include:
Obtain the K picture that user selects from picture library, the vehicle for the vehicle for including in the picture i in the K picture
Angle between head direction and the vertical direction of the picture i is less than the predetermined threshold value, wherein the picture i is the K
Any picture in picture, the K are the integer more than or equal to 1;
Extract O feature respectively from the K picture, the O is the integer more than 1;
Vector regression processing is supported to the K*O feature extracted, to obtain affine transformation matrix.
As a kind of possible embodiment, above-mentioned processor 3000 carries out the target area after the registration process
Feature extraction includes to obtain M feature:
The target vehicle is divided into the region of K1 the first default sizes, the region of a second default sizes of K2 respectively
The region of size is preset with K3 third, wherein the K1, the K2 and the K3 are to be worth mutually different positive integer;
The corresponding K1 histogram in region, the K2 a second for obtaining the K1 the first default sizes respectively is default big
The corresponding K2 histogram in small region and the K3 third preset the corresponding K3 histogram in region of size;
K1 weights of the K1 histogram, K2 weights of the K2 histogram and institute are configured according to preset order
State K3 weights of K3 histogram, the K1 weights are all higher than 0 and equal, and the K2 weights are all higher than 0 and equal, institute
It states K3 weights and is all higher than 0 and equal, and the K1 weights and the K2 weights and the K3 weights sum are 1;
The histograms of oriented gradients HOG features for extracting the target vehicle respectively, to obtain presetting with the K1 first
The corresponding K1 HOG feature in region of size, the K2 HOG corresponding with the K2 regions of the second default size
Feature and the K3 HOG feature corresponding with the default region of size of the K3 third;
By the K1 weights and K1 HOG features, the K2 weights and the K2 HOG features and the K3
A weights and K3 HOG features composition characteristic vector;
It uses third to preset grader to be trained described eigenvector to obtain M feature.
In the specific implementation, input equipment 1000 described in the embodiment of the present invention, output equipment 2000 and processor
3000, which can perform a kind of first embodiment of model recognizing method, second embodiment and third provided in an embodiment of the present invention, implements
Realization method described in example also can perform and be retouched in a kind of first embodiment of identification device provided in an embodiment of the present invention
The realization method for the identification device stated, details are not described herein.
Described identification device can obtain picture to be detected through the embodiment of the present invention;Using the first default grader pair
The picture to be detected is detected;If containing target vehicle in the picture to be detected, the mesh in the picture to be detected is extracted
Mark vehicle;Registration process is carried out to the target vehicle, so that the headstock direction of the target vehicle and the target area
Vertical direction between angle be less than predetermined threshold value;Feature extraction is carried out to the target vehicle after the registration process, with
M feature is obtained, the M is the integer more than 1;Classified to the M feature using the second default grader;According to institute
The result for stating classification determines the vehicle of the target vehicle.In the embodiment of the present invention, extract target vehicle, to the target vehicle into
Row registration process, and feature extraction is carried out to the target vehicle after registration process on this basis, and to the feature extracted
Classification processing is carried out, to which the precision of vehicle cab recognition can be improved.
The embodiment of the present invention also provides a kind of computer storage media, wherein the computer storage media can be stored with journey
Sequence, the program include some or all of any one signal processing method described in above method embodiment step when executing
Suddenly.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiment.
It should be noted that for each method embodiment above-mentioned, for simple description, therefore it is all expressed as a series of
Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the described action sequence because
According to the present invention, certain steps may can be performed in other orders or simultaneously.Secondly, those skilled in the art also should
Know, embodiment described in this description belongs to preferred embodiment, involved action and module not necessarily this hair
Necessary to bright.
In several embodiments provided herein, it should be understood that disclosed device, it can be by another way
It realizes.For example, the apparatus embodiments described above are merely exemplary, for example, said units division, it is only a kind of
Division of logic function, formula that in actual implementation, there may be another division manner, such as multiple units or component can combine or can
To be integrated into another system, or some features can be ignored or not executed.Another point, shown or discussed is mutual
Coupling, direct-coupling or communication connection can be by some interfaces, the INDIRECT COUPLING or communication connection of device or unit,
Can be electrical or other forms.
The above-mentioned unit illustrated as separating component may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, you can be located at a place, or may be distributed over multiple
In network element.Some or all of unit therein can be selected according to the actual needs to realize the mesh of this embodiment scheme
's.
In addition, each functional unit in various embodiments of the present invention can be integrated in a processing unit, can also be
Each unit physically exists alone, can also be during two or more units are integrated in one unit.Above-mentioned integrated unit
Both the form that hardware may be used is realized, can also be realized in the form of SFU software functional unit.
If above-mentioned integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can be stored in a computer read/write memory medium.Based on this understanding, technical scheme of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or network equipment etc., can be specifically the processor in computer equipment) is held
The all or part of step of each embodiment above method of the row present invention.Wherein, storage medium above-mentioned may include:USB flash disk, shifting
Dynamic hard disk, magnetic disc, CD, read-only memory (English:Read-Only Memory, abbreviation:) or random access memory ROM
(English:Random Access Memory, abbreviation:The various media that can store program code such as RAM).
The above, the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although with reference to before
Stating embodiment, invention is explained in detail, it will be understood by those of ordinary skill in the art that:It still can be to preceding
The technical solution recorded in each embodiment is stated to modify or equivalent replacement of some of the technical features;And these
Modification or replacement, the spirit and scope for various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution.
Claims (8)
1. a kind of model recognizing method, which is characterized in that including:
Obtain picture to be detected;
The picture to be detected is detected using the first default grader;
If containing target vehicle in the picture to be detected, the target vehicle in the picture to be detected is extracted;
Registration process is carried out to the target vehicle, so that the headstock direction of the target vehicle and the Vertical Square of target area
Angle between is less than predetermined threshold value;
Feature extraction is carried out to the target vehicle after the registration process, to obtain M feature, the M is whole more than 1
Number;
Classified to the M feature using the second default grader;
The vehicle of the target vehicle is determined according to the result of the classification;
Wherein, the target area to after the registration process carries out feature extraction, includes to obtain M feature:
The target vehicle is divided into the region of K1 the first default sizes, the region of a second default sizes of K2 and K3 respectively
A third presets the region of size, wherein the K1, the K2 and the K3 are to be worth mutually different positive integer;
A second default size of the corresponding K1 histogram in region, the K2 of the K1 the first default sizes is obtained respectively
The corresponding K2 histogram in region and the K3 third preset the corresponding K3 histogram in region of size;
K1 weights of the K1 histogram, K2 weights of the K2 histogram and the K3 are configured according to preset order
K3 weights of a histogram, wherein the K1 weights are all higher than 0 and equal, and the K2 weights are all higher than 0 and equal,
The K3 weights are all higher than 0 and equal, and the K1 weights and the K2 weights and the K3 weights sum are 1;
The histograms of oriented gradients HOG features for extracting the target vehicle respectively, to obtain and the K1 the first default sizes
The corresponding K1 in the region HOG feature, a HOG features of K2 corresponding with the K2 regions of the second default size
The K3 HOG feature corresponding with the default region of size of the K3 third;
By the K1 weights and K1 HOG features, the K2 weights and the K2 HOG features and the K3 power
Value and K3 HOG features composition characteristic vector;
It uses third to preset grader to be trained described eigenvector to obtain M feature.
2. the method as described in claim 1, which is characterized in that described to use the first default grader to the picture to be detected
Be detected including:
Q pictures of the P pictures of vehicle as positive sample and without vehicle will be contained in picture library as negative sample to described
Picture to be detected is detected, wherein the P and Q is the integer more than 1.
3. the method as described in claim 1, which is characterized in that described to include by target vehicle progress registration process:
Determine the corresponding affine transformation matrix of the target vehicle;
Affine transformation is carried out to the target vehicle according to the affine transformation matrix.
4. method as claimed in claim 3, which is characterized in that the corresponding affine transformation matrix of the determination target vehicle
Including:
Obtain the K picture that user selects from picture library, the headstock side for the vehicle for including in the picture i in the K picture
It is less than the predetermined threshold value to the angle between the vertical direction of the picture i, wherein the picture i is the K picture
In any picture, the K is integer more than or equal to 1;
Extract O feature respectively from the K picture, the O is the integer more than 1;
Vector regression processing is supported to the K*O feature extracted, to obtain affine transformation matrix.
5. a kind of identification device, which is characterized in that including:
First acquisition unit, for obtaining picture to be detected;
Detection unit, for being examined to the picture to be detected that the first acquisition unit is got using the first default grader
It surveys;
First extraction unit contains target vehicle, described in extraction if being detected for the detection unit in the picture to be detected
Target vehicle in picture to be detected;
First processing units, the target vehicle for being extracted to first extraction unit carries out registration process, so that described
Angle between the headstock direction and the vertical direction of target area of target vehicle is less than predetermined threshold value;
Second extraction unit, for carrying out feature extraction to the target vehicle after the first processing units registration process, with
M feature is obtained, the M is the integer more than 1;
Taxon, the M feature for being extracted to second extraction unit using the second default grader are classified;
First determination unit, the vehicle for determining the target vehicle according to the result of the classification of the taxon;
Wherein, second extraction unit includes:
Division unit, for the target vehicle that first extraction unit extracts to be divided into a first default sizes of K1 respectively
Region, the region of K2 the second default sizes and the default size of K3 third region, wherein the K1, the K2 and described
K3 is to be worth mutually different positive integer;
Third acquiring unit, the region for obtaining the K1 the first default sizes that the division unit divides respectively correspond to
K1 histogram, the K2 the second default sizes the corresponding K2 histogram in region and the default size of the K3 third
The corresponding K3 histogram in region;
Dispensing unit, K1 power of the K1 histogram for being got according to preset order configuration third acquiring unit
It is worth, K3 weights of the K2 weights and the K3 histogram of the K2 histogram, wherein the K1 weights are all higher than
0 and equal, the K2 weights are all higher than 0 and equal, and the K3 weights are all higher than 0 and equal, and the K1 weights with
The K2 weights and the K3 weights sum are 1;
4th extraction unit, the histograms of oriented gradients of the target vehicle for extracting the first extraction unit extraction respectively
HOG features, with obtain the K1 HOG features corresponding with the regions of the first default size the K1, with the K2 the
The corresponding K2 HOG feature in region of two default sizes and K3 corresponding with the default region of size of the K3 third are a
The HOG features;
Component units, for the K1 weights and the K1 HOG features, the K2 weights and the K2 HOG are special
The K3 weights of seeking peace are vectorial with the K3 HOG features composition characteristic;
Training unit is trained for using third to preset the feature vector that grader forms the component units to obtain
M feature.
6. identification device as claimed in claim 5, which is characterized in that the detection unit is specifically used for:
Q pictures of the P pictures of vehicle as positive sample and without vehicle will be contained in picture library as negative sample to described
Picture to be detected is detected, wherein the P and Q is the integer more than 1.
7. identification device as claimed in claim 5, which is characterized in that the first processing units include:
Second determination unit determines the corresponding affine transformation matrix of target vehicle of the first extraction unit extraction;
Affine transformation unit, the affine transformation matrix for being determined according to second determination unit carry out the target vehicle
Affine transformation.
8. identification device as claimed in claim 7, which is characterized in that second determination unit includes:
Second acquisition unit, the K picture selected from picture library for obtaining user wrap in the picture i in the K picture
Angle between the headstock direction of the vehicle contained and the vertical direction of the picture i is less than the predetermined threshold value, wherein the figure
Piece i is any picture in the K picture, and the K is the integer more than or equal to 1;
Third extraction unit, for extracting O feature respectively from the K picture that the second acquisition unit obtains, the O is
Integer more than 1;
Second processing unit, the K*O feature for being extracted to the third extraction unit are supported vector regression processing,
To obtain affine transformation matrix.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510639752.7A CN105160333B (en) | 2015-09-30 | 2015-09-30 | A kind of model recognizing method and identification device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510639752.7A CN105160333B (en) | 2015-09-30 | 2015-09-30 | A kind of model recognizing method and identification device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105160333A CN105160333A (en) | 2015-12-16 |
CN105160333B true CN105160333B (en) | 2018-08-17 |
Family
ID=54801185
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510639752.7A Active CN105160333B (en) | 2015-09-30 | 2015-09-30 | A kind of model recognizing method and identification device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105160333B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108229468B (en) | 2017-06-28 | 2020-02-21 | 北京市商汤科技开发有限公司 | Vehicle appearance feature recognition and vehicle retrieval method and device, storage medium and electronic equipment |
US10140553B1 (en) * | 2018-03-08 | 2018-11-27 | Capital One Services, Llc | Machine learning artificial intelligence system for identifying vehicles |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120288191A1 (en) * | 2011-05-12 | 2012-11-15 | Fuji Jukogyo Kabushiki Kaisha | Environment recognition device and environment recognition method |
CN202563526U (en) * | 2012-03-22 | 2012-11-28 | 北京尚易德科技有限公司 | Transportation vehicle detection and recognition system based on video |
CN103761723A (en) * | 2014-01-22 | 2014-04-30 | 西安电子科技大学 | Image super-resolution reconstruction method based on multi-layer supporting vectors |
-
2015
- 2015-09-30 CN CN201510639752.7A patent/CN105160333B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120288191A1 (en) * | 2011-05-12 | 2012-11-15 | Fuji Jukogyo Kabushiki Kaisha | Environment recognition device and environment recognition method |
CN202563526U (en) * | 2012-03-22 | 2012-11-28 | 北京尚易德科技有限公司 | Transportation vehicle detection and recognition system based on video |
CN103761723A (en) * | 2014-01-22 | 2014-04-30 | 西安电子科技大学 | Image super-resolution reconstruction method based on multi-layer supporting vectors |
Non-Patent Citations (3)
Title |
---|
一种基于SIFT特征提取的车牌定位方法;姬峰宽,付永庆;《应用科技》;20120215;第39卷(第1期);第2节1-9段 * |
人脸对齐算法研究;宿佳宁;《中国优秀硕士学位论文全文数据库 信息科技辑》;20130215(第02期);I138-1872 * |
车牌检测及汽车类型分类方法研究;姜谊;《中国优秀硕士学位论文全文数据库 信息科技辑》;20101015(第10期);第2.1、3.4、4.2、6.2.2节,第3.1.1节第1段,第3.2节第1段 * |
Also Published As
Publication number | Publication date |
---|---|
CN105160333A (en) | 2015-12-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Tudor Ionescu et al. | How hard can it be? Estimating the difficulty of visual search in an image | |
Li et al. | DeepMitosis: Mitosis detection via deep detection, verification and segmentation networks | |
CN104424466B (en) | Method for checking object, body detection device and image pick up equipment | |
CN105224951B (en) | A kind of vehicle type classification method and sorter | |
CN104680144B (en) | Based on the lip reading recognition methods and device for projecting very fast learning machine | |
CN105512683A (en) | Target positioning method and device based on convolution neural network | |
CN110209866A (en) | A kind of image search method, device, equipment and computer readable storage medium | |
US8306318B2 (en) | Image processing apparatus, image processing method, and computer readable storage medium | |
CN106295666B (en) | Method and device for acquiring classifier and detecting object and image processing equipment | |
CN105718937B (en) | Multi-class object classification method and system | |
WO2016133908A1 (en) | Barcode tag detection in side view sample tube images for laboratory automation | |
CN105303163B (en) | A kind of method and detection device of target detection | |
CN110348392B (en) | Vehicle matching method and device | |
CN107274679A (en) | Vehicle identification method, device, equipment and computer-readable recording medium | |
CN105184286A (en) | Vehicle detection method and detection device | |
CN105160333B (en) | A kind of model recognizing method and identification device | |
CN105809181A (en) | Logo detection method and device | |
CN103530646A (en) | Complex-object detection using a cascade of classifiers | |
KR101584091B1 (en) | Multi-view object detection method using shared local features | |
WO2017069741A1 (en) | Digitized document classification | |
US9798932B2 (en) | Video extraction method and device | |
KR101582983B1 (en) | Apparatus and method for fast object detection using scale invariant feature vector | |
Creusen et al. | A semi-automatic traffic sign detection, classification, and positioning system | |
Estrada et al. | Appearance-based keypoint clustering | |
JP5901054B2 (en) | Object detection method and object detection apparatus using the method |
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 |