CN103295021A - Method and system for detecting and recognizing feature of vehicle in static image - Google Patents

Method and system for detecting and recognizing feature of vehicle in static image Download PDF

Info

Publication number
CN103295021A
CN103295021A CN2012100428091A CN201210042809A CN103295021A CN 103295021 A CN103295021 A CN 103295021A CN 2012100428091 A CN2012100428091 A CN 2012100428091A CN 201210042809 A CN201210042809 A CN 201210042809A CN 103295021 A CN103295021 A CN 103295021A
Authority
CN
China
Prior art keywords
vehicle
image
training
detection
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN2012100428091A
Other languages
Chinese (zh)
Inventor
胡楠
邹国平
朱建明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
BEIJING MINGRI FASHION INFORMATION TECHNOLOGY Co Ltd
Original Assignee
BEIJING MINGRI FASHION INFORMATION TECHNOLOGY Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by BEIJING MINGRI FASHION INFORMATION TECHNOLOGY Co Ltd filed Critical BEIJING MINGRI FASHION INFORMATION TECHNOLOGY Co Ltd
Priority to CN2012100428091A priority Critical patent/CN103295021A/en
Publication of CN103295021A publication Critical patent/CN103295021A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Image Analysis (AREA)

Abstract

The invention belongs to the field of recognizing types of vehicles through computer image processing, and relates to a method and system for detecting and recognizing a feature and a brand of a vehicle in a static image by means of a digital picture processing technique. A method for detecting the vehicle in the static image is combined with a vehicle logo detecting method which is based on an AdaBoost framework in the method. The method and system for detecting and recognizing the feature of the vehicle in the static images comprises a training part and a detecting part. The training part includes the following steps of manufacturing a vehicle logo sample, collecting an image containing the vehicle logo from the Internet, positioning the vehicle logo, and extracting the vehicle logo image based on position information; calculating a sample feature, constructing 5 different rectangular features with each rectangular feature corresponding to one Haar feature; training a cascade classifier, inputting the training sample acquired from the last step and conducting training, and finally connecting strong classifiers and multiple corresponding weak classifiers obtained in training in series. The detecting part includes the following steps: loading the image to be detected, converting the image into a grey-scale image and conducting histogram equalization, loading the vehicle logo classifiers which include threshold values of the strong classifiers and the weak classifiers and rectangular feature information corresponding to the selected features, conducting cascade vehicle logo detection with the detected image firstly passing the detection of the former strong classifiers. If the detected image is not the vehicle logo image, the detected image can be excluded at the front end, and only the vehicle logo can finally pass the detection of the strong classifiers at various different levels.

Description

Method and system for detecting and identifying vehicle characteristics in static picture
Technical Field
The invention relates to a method and a system for detecting and identifying vehicle characteristics in a static picture, belongs to the field of vehicle image identification, and particularly relates to vehicle appearance characteristic detection and vehicle brand identification in the static picture.
Background
With the development of electronic imaging technology and internet, people create pictures, share pictures and obtain pictures more and more conveniently and variously, so that the pictures appearing on the internet are increasing, except for the description of the pictures by characters, a computer does not know the content of the pictures, such as whether a vehicle exists in a certain picture, what characteristics and brand of the vehicle in the picture, and generally, the contents can be known only through human inspection and judgment.
The traditional method for annotating the picture content in the form of keywords cannot be well matched with the corresponding picture, and therefore, the technology that people directly identify the picture content by using a computer according to a certain algorithm is developed. Therefore, even if the internet pictures are not described in advance, the contents of the internet pictures can be known through the image recognition technology, wherein the contents comprise the detection and the recognition of the vehicle.
The existing vehicle characteristic detection and identification technology is mainly applied to videos, such as intelligent traffic monitoring, and moving vehicles can be easily detected through multi-frame information of the videos so as to be positioned and identified; however, these conventional methods are not suitable for use in still pictures, and a method for detecting and identifying vehicles in still pictures is sought. With the development of electronic commerce, more and more commodities are directly facing consumers on the network, including vehicles. If the computer can know whether a vehicle exists in a certain picture in the Internet and know the brand of the vehicle, the popularization of the vehicle by a vehicle seller is facilitated, so that the technology for detecting and identifying the vehicle in the static picture has wide application prospect.
Disclosure of Invention
The invention aims to provide a method for detecting and identifying vehicle characteristics by using static pictures, and provides a system for detecting and identifying vehicle characteristics by using internet pictures, which comprises the following steps: the method comprises the steps of carrying out vehicle detection on a network picture to be detected, judging the posture of a vehicle, carrying out vehicle logo detection in the front area or the rear area of the vehicle so as to identify the vehicle, and the flow is shown in fig. 1.
The invention provides a vehicle detection method in a static picture, which comprises two parts of training and detection.
The training phase comprises the following steps:
(1) manufacturing a vehicle sample: normalizing the input picture, namely performing gamma normalization on each color component of the input image to adapt to the condition that the image is too dark or the contrast is low, wherein the operation adopted by the method is to perform logarithm operation on the color components;
(2) calculating sample characteristics: calculating a Histogram of Oriented Gradient (HOG) feature pyramid of the normalized image;
(3) training a vehicle model: and (3) transmitting the characteristic data set of the sample into an implicit Support Vector machine (LSVM) of a training classifier for learning. A hybrid model of a root model, a component model, and a corresponding deformable component model of the vehicle is generated by learning.
The detection phase comprises the following steps:
(1) loading a picture to be tested: normalizing the input picture, namely performing gamma normalization on each color component of the input image;
(2) and (3) feature calculation: calculating an HOG characteristic pyramid of the network picture to be detected;
(3) loading a vehicle model: loading a data file storing a vehicle model;
(4) vehicle detection: and scanning an area matched with the variable component model on the characteristic pyramid through a multi-scale detection algorithm of a scale space to realize the detection and positioning of the vehicle.
The invention provides a static picture vehicle identification method, in particular to a vehicle logo detection method based on an AdaBoost framework, which comprises two parts of training and detection.
The training phase comprises the following steps:
(1) manufacturing a car logo sample;
(2) calculating sample characteristics;
(3) and training a cascade classifier.
The detection phase comprises the following steps:
(1) loading a picture to be tested;
(2) loading a car logo classifier;
(3) and detecting the cascade vehicle logo.
Drawings
FIG. 1 is a flow diagram of still-picture vehicle feature detection and identification.
Fig. 2 is a schematic composition diagram of a model-deformable part of a training vehicle model 6.
FIG. 3 is a schematic illustration of target deforming member positioning in vehicle feature detection.
FIG. 4 is a schematic diagram of Haar features for sample feature computation.
Fig. 5 is a schematic flow chart of cascade logo detection.
Fig. 6 is a diagram illustrating the result of the cascade emblem detection process shown in fig. 5.
Detailed Description
The following detailed description of the embodiments of the present invention will be provided with reference to the drawings and examples, so that how to apply the technical means to solve the technical problems and achieve the technical effects can be fully understood and implemented.
The existing vehicle detection method only builds a model for the front side of the vehicle, when the vehicle in the picture appears on the side surface or the rear surface, the vehicle is probably not detected, and the existing method solves the problem that the model can only be built for each side surface of the vehicle, so that the calculated amount is increased, and the shielding problem cannot be well avoided. The method of the invention is to respectively establish models (root models) at the front, the back and the side of the vehicle, divide each model into different parts, respectively establish models (part models and deformable part models) for each part and corresponding deformation thereof, and finally fuse the models into a final vehicle model.
The existing vehicle identification method is to locate the vehicle logo of the vehicle by locating the vehicle license plate, and then distinguish the brand of the vehicle by the edge characteristics of the image of the vehicle logo, and the like, and the method depends on the accuracy of license plate location, but many network vehicle pictures have no license plate, such as the vehicle pictures on the vehicle exhibition; in addition, the accuracy, such as illumination and noise, can also be reduced due to the influence of the environment by simply judging various types of car logos through the features. According to the method, on the basis of extracting Haar features, a large number of car logo samples are trained by an AdaBoost method to obtain a cascade classifier for detecting the car logos in the pictures to be detected, and the method can directly position the car logos in the pictures and is not influenced by the environment.
The invention provides a vehicle detection method in a static picture, which comprises two parts of training and detection and is characterized in that:
the training phase comprises the following steps:
(1) manufacturing a vehicle sample:
the gamma normalization is carried out on each color component of the input image so as to adapt to the condition that the image is too dark or the contrast is low. The sample pictures are used for marking the vehicles in the sample pictures by rectangles in advance.
(2) Calculating sample characteristics:
and constructing an image pyramid for the normalized sample image, scanning by windows with certain sizes on each layer, and calculating HOG characteristics in each window to form a characteristic pyramid.
(3) Training a vehicle model:
in order to adapt to different postures and shielding problems of vehicles, 2 template models are established for vehicle models, wherein one template model is a model of a front view angle and a rear view angle of the vehicle (the front view angle and the rear view angle of the vehicle are not greatly different and are represented by one share), and the other template model is a vehicle side model; for each template model, the invention divides it into 6 parts, as shown in fig. 2.
The vehicle model of the present invention contains n =2 sub-templates, using (n + 2) tuples
Figure 901075DEST_PATH_IMAGE001
Is defined in which
Figure 156345DEST_PATH_IMAGE002
Representing a root filter (HOG feature model),
Figure 918764DEST_PATH_IMAGE003
is the model of the ith sub-template, and b is the offset. Each sub-template contains m =6 part models, with m
Figure 569189DEST_PATH_IMAGE004
The triple definition, wherein,
Figure 617916DEST_PATH_IMAGE005
is a filter of the jth component, of size
Figure 986580DEST_PATH_IMAGE006
Figure 552691DEST_PATH_IMAGE007
The fixed position of the j-part relative to the root,is a loss function of the sub-template from the correct position in the vehicle model.
In the image feature pyramid, each filter is located at a position of
Figure 559141DEST_PATH_IMAGE009
Wherein
Figure 415102DEST_PATH_IMAGE010
Indicating that the ith filter is in the characteristic pyramid
Figure 112799DEST_PATH_IMAGE011
The position of the layer, the corresponding response is the sum of the responses of each filter at the respective position minus the vehicle component variation loss, plus the offset:
Figure 737816DEST_PATH_IMAGE012
wherein, H represents a characteristic pyramid,
Figure 534871DEST_PATH_IMAGE013
indicates juxtaposition in H in row-priority order
Figure 986449DEST_PATH_IMAGE014
In the upper left corner
Figure 628783DEST_PATH_IMAGE015
The feature vectors of the sub-windows are,
Figure 232940DEST_PATH_IMAGE016
the offset of the ith component from the fixed position,
Figure 200896DEST_PATH_IMAGE017
is a deformation feature.
The model is simplified to be combined with a classifier to obtain model parameters:
Figure 765869DEST_PATH_IMAGE018
using m-element arrays
Figure 556102DEST_PATH_IMAGE019
Defining a mixture model comprising m sets of models, wherein
Figure 686869DEST_PATH_IMAGE020
Is an object model of group c, wherein. Single object model in hybrid model
Figure 471471DEST_PATH_IMAGE020
At the position of each filter of
Figure 721187DEST_PATH_IMAGE022
Wherein
Figure 175302DEST_PATH_IMAGE023
Is a model
Figure 327804DEST_PATH_IMAGE020
Number of middle partsSimplifying the filter positions of the c-th group of models to
Figure 398528DEST_PATH_IMAGE024
. Model parameter vectors for position response of hybrid models
Figure 920776DEST_PATH_IMAGE025
Sum vector
Figure 354032DEST_PATH_IMAGE026
The dot product of (d). Wherein the vector
Figure 834691DEST_PATH_IMAGE025
Is a concatenation of model parameter vectors in each single target model,
Figure 595974DEST_PATH_IMAGE027
(ii) a And vector
Figure 797279DEST_PATH_IMAGE028
Appearing sparse. Different models of the same target can be obtained by adopting a plurality of groups of models to learn, so that a mixed model is combined.
And the LSVM is adopted to learn the component model, so that hidden information in the image can be fully utilized, and the model can be enriched.
The detection phase comprises the following steps:
(1) loading a picture to be tested:
the input picture is normalized, i.e. gamma normalization is performed for each color component of the input image.
(2) And (3) feature calculation:
and constructing an image pyramid for the normalized image to be detected, scanning by windows with certain sizes on each layer, and calculating HOG characteristics in each window to form a characteristic pyramid.
(3) Loading a vehicle model:
a data file storing a vehicle model is loaded.
(4) Vehicle detection:
computing a stored response of the feature pyramid first layer features with the first model filter,and (3) performing distance conversion on the component filter:
Figure 611969DEST_PATH_IMAGE030
the filter response at the nearby position is expanded by using distance conversion, and the deformation loss of the component is added, so that the detection precision is improved.
Figure 250760DEST_PATH_IMAGE031
Is the maximum distance of the ith part from the root position where the root position filter response is placed at the ith layer (x, y) for that part.
The response for each layer root position is calculated using the sum of the corresponding layer root filter responses plus the transformed and sampled component filter:
Figure 114811DEST_PATH_IMAGE032
whereinThe total number of layers of the feature pyramid.
Calculate the optimal offset of the part:
by root position
Figure 923739DEST_PATH_IMAGE035
At an optimum offset
Figure 325902DEST_PATH_IMAGE036
The corresponding part position is searched, and the positioning of the target deformation part is realized, and the result is shown in figure 3.
Secondly, the method for detecting the car logo based on the AdaBoost frame comprises two parts of training and detecting, wherein the steps of the training stage are as follows:
(1) manufacturing a car logo sample:
the method comprises the steps of collecting pictures containing car logos from a network, calibrating the positions of the car logos, extracting car logo images according to position information, zooming according to the inherent length-width ratio of the car logos, eliminating illumination influences through histogram equalization to serve as a positive sample of the car logos, and taking other parts of the car pictures without the car logos as negative samples.
(2) Calculating sample characteristics:
5 different rectangular features are constructed, each rectangular feature corresponds to a Haar feature (shown in FIG. 4), the Haar feature is defined as the sum of weighted values of the sum of pixel values of the corresponding rectangular region, and the Haar feature is calculated by means of integral images.
The integral image is defined as:
Figure 66773DEST_PATH_IMAGE038
representing pixels in an original image
Figure 555523DEST_PATH_IMAGE039
The sum of all the pixel values at the upper left,
Figure 151589DEST_PATH_IMAGE040
namely to represent
Figure 143816DEST_PATH_IMAGE039
A certain pixel value of the upper left region.
Integral image
Figure 308081DEST_PATH_IMAGE041
Calculating in an incremental manner:
Figure 392449DEST_PATH_IMAGE042
provision for
Figure 401994DEST_PATH_IMAGE043
And then, the integral image can be calculated by traversing the whole image once according to the rows or the columns. The sum of the pixel values of a certain rectangular area in the original image is calculated, and as long as four values are obtained by inquiring the integral image according to the positions of four vertexes of the rectangle, certain addition and subtraction operation of the four values can be equivalent to the sum of the pixel values of the rectangular area.
And (3) Haar feature calculation:
the first feature is that:
Figure 779885DEST_PATH_IMAGE044
wherein
Figure 974106DEST_PATH_IMAGE045
Is the sum of the pixel values of the white rectangular area,
Figure 906290DEST_PATH_IMAGE046
is the sum of the pixel values of the black rectangular area.
Figure 250684DEST_PATH_IMAGE047
The second characteristic:
Figure 92869DEST_PATH_IMAGE048
Figure 67778DEST_PATH_IMAGE049
a third feature:
Figure 759977DEST_PATH_IMAGE051
a fourth feature:
Figure 581302DEST_PATH_IMAGE052
the fifth characteristic:
Figure 538949DEST_PATH_IMAGE054
Figure 225145DEST_PATH_IMAGE055
Figure 900977DEST_PATH_IMAGE056
features are computed on n samples of the input, including m positive samples and n-m negative samples, each sample having two attributes
Figure 607902DEST_PATH_IMAGE057
Where x represents the Haar feature vector of the sampleY represents the class of the sample, and is taken as 1 for positive samples and-1 for negative samples.
(3) Training a cascade classifier:
inputting n training samples obtained in the previous step:
Figure 226282DEST_PATH_IMAGE059
defining a weak classifier:
Figure 897566DEST_PATH_IMAGE060
is a sample
Figure 797706DEST_PATH_IMAGE062
The Haar feature vector of the image,
Figure 419180DEST_PATH_IMAGE063
is an unequal number direction controller, takes a value of +1 or-1,
Figure 600763DEST_PATH_IMAGE064
a threshold is trained for the weak classifier.
Initializing error weights:
Figure 524856DEST_PATH_IMAGE065
in the first placeIn the secondary training:
Figure 845252DEST_PATH_IMAGE067
wherein
Figure 350182DEST_PATH_IMAGE068
Is set total training times;
normalization weight:
Figure 569811DEST_PATH_IMAGE069
for each featureGenerate its corresponding weak classifier
Figure 529994DEST_PATH_IMAGE071
Calculating the error with respect to the current weight:
Figure 30376DEST_PATH_IMAGE072
selecting the one with the smallest error
Figure 296273DEST_PATH_IMAGE073
Weak classifier of
Figure 498584DEST_PATH_IMAGE074
Adding the feature information into a strong classifier, recording rectangular feature information corresponding to the feature at the moment, and updating the weight:
Figure 406497DEST_PATH_IMAGE075
wherein, if it is
Figure 886020DEST_PATH_IMAGE076
A sample
Figure 696719DEST_PATH_IMAGE077
Is correctly classified, then
Figure 261692DEST_PATH_IMAGE078
Otherwise
Figure 973296DEST_PATH_IMAGE079
Forming a final strong classifier:
Figure 697539DEST_PATH_IMAGE080
wherein,
under the set training times T, a strong classifier is generated in each training, a plurality of weak classifiers are selected in the training process, and finally each strong classifier and the corresponding weak classifiers are connected in series to form the final cascade classifier.
The detection stage comprises the following steps:
(1) loading a picture to be tested: converting the image into a gray scale image and carrying out histogram equalization;
(2) loading a car logo classifier: the classifier data obtained by training is stored in a corresponding txt file, wherein the data structure is described as:
n strong classifiers of
Figure 967294DEST_PATH_IMAGE082
Each strong classifier includes a threshold of the strong classifierAnda weak classifier;
wherein, the firstjEach weak classifier includes a threshold of the weak classifier
Figure 839938DEST_PATH_IMAGE085
Direction controller
Figure 113925DEST_PATH_IMAGE086
Coefficient of
Figure 167331DEST_PATH_IMAGE087
And rectangular feature information corresponding to the features selected by the weak classifier: number of rectangles
Figure 849854DEST_PATH_IMAGE088
Type of rectangular featurePosition information and weight of each sub-rectangle
(3) Detecting a cascade vehicle logo: in the cascade classifier, the strong classifier is more complex than one stage, the detection image firstly passes through the detection of the previous strong classifier, if the detection image is not the car logo image, the detection image is eliminated at the front end, only the car logo image can finally pass through the detection of the strong classifier at each stage, a large number of non-car logo images are eliminated when the strong classifier at the previous stages is detected, the process is as shown in fig. 5, and the result is as shown in fig. 6.

Claims (12)

1. The invention aims to provide a method for detecting and identifying vehicle characteristics by using static pictures, and provides a system for detecting and identifying vehicle characteristics by using internet pictures, which comprises the following steps: and carrying out vehicle detection on the network picture to be detected, judging the posture of the vehicle, and carrying out vehicle logo detection in the front area or the rear area of the vehicle so as to identify the vehicle.
2. The invention provides a vehicle detection method in a static picture, which comprises two parts of training and detection.
3. The training phase comprises the following steps: by manufacturing a vehicle sample, normalizing an input picture; then, sample feature calculation is carried out, and an image pyramid is constructed on the normalized image; next, training a vehicle model, and transmitting the characteristic data set of the sample into a training classifier for learning; generating a hybrid model of a root model, a component model, and a corresponding deformable component model of the vehicle by learning; the detection stage comprises: loading a picture to be tested: normalizing the input picture, namely performing gamma normalization on each color component of the input image; and (3) carrying out feature calculation: constructing an image pyramid for the normalized image to be detected; loading a vehicle model, and loading a data file for storing the vehicle model; and finally, vehicle detection is carried out, and an area matched with the variable component model is scanned on the characteristic pyramid to realize the detection and positioning of the vehicle.
4. The invention provides a car logo detection method based on an AdaBoost framework, which comprises two parts of training and detection.
5. The training phase comprises the following steps: making a car logo sample, acquiring a picture containing a car logo from a network, calibrating the position of the car logo, and extracting a car logo image according to position information; calculating sample characteristics, and constructing rectangular characteristics, wherein each rectangular characteristic corresponds to a Haar characteristic; training the cascade classifier, inputting the training sample obtained in the last step and training, and finally connecting the strong classification obtained by training and a plurality of corresponding weak classifiers in series; the detection stage comprises: loading a picture to be tested: converting the image into a gray scale image and carrying out histogram equalization; loading a vehicle logo classifier which comprises thresholds of a strong classifier and a weak classifier and rectangular feature information corresponding to the selected features; and (3) detecting the cascade vehicle logo, wherein the detection image firstly passes through the detection of the strong classifiers in the front, if the detection image is not the vehicle logo image, the detection image is eliminated at the front end, and only the vehicle logo image can finally pass through the detection of the strong classifiers in each stage.
6. The method for making a vehicle sample during a training phase and loading a picture to be tested during a testing phase as claimed in claim 1, wherein: the gamma normalization is carried out on each color component of the input image so as to adapt to the condition that the image is too dark or the contrast is low, the logarithm operation is carried out on the color components, and the vehicle in the sample image is marked by a rectangle in advance.
7. The method of training phase sample feature computation of claim 2, wherein: and scanning the normalized sample on each layer by using a window with a certain size, and calculating the HOG characteristic in each window to form a characteristic pyramid.
8. A method of training a vehicle model during a training phase as claimed in claim 3, wherein: in order to adapt to different postures and shielding problems of vehicles, 2 template models are established for vehicle models, wherein one model is a model of a front view angle and a rear view angle of the vehicle, and the other model is a vehicle side model.
9. The method of inspection phase vehicle inspection of claim 4, wherein: computing a characteristic pyramidlLayer characteristics andithe storage response of each model filter, the distance conversion is carried out on the component filter, the filter response of the nearby position is expanded by utilizing the distance conversion, the deformation loss of the component is added, and the detection precision is improved;D i,j (x,y)is the firstiMaximum distance of each part from root position, wherein root position filter response is placed at the second part corresponding to the partlLayer (A)x,y) Calculating a response for each layer root position using the sum of the corresponding layer root filter responses plus the transformed and sampled component filter; then, the optimal offset of the component is calculated, using the root positionx 0, y 0, l 0) And searching the corresponding part position in the optimal offset to realize the positioning of the target deformation part.
10. A method of making emblem samples for a training phase based on the AdaBoost framework of claim 5, wherein: and extracting the car logo image according to the position information, zooming according to the inherent length-width ratio of the car logo, eliminating the illumination influence through histogram equalization, and taking the car logo image as a positive sample and other parts without the car logo as negative samples.
11. The method of AdaBoost framework-based training phase sample feature computation of claim 6, wherein: defining the Haar characteristics as the sum of weighted values of the sum of pixel values of the corresponding rectangular area, and calculating the Haar characteristics in an integral image mode; integral image SAT (x, y) Representing pixels in the original image (x, y) The sum of all pixel values at the upper left part is calculated in an incremental mode, and then the corresponding integral image can be calculated only by traversing the whole image once according to rows or columns; the sum of the pixel values of a certain rectangular area in the original image is calculated, and as long as four values are obtained by inquiring the integral image according to the positions of four vertexes of the rectangle, certain addition and subtraction operation of the four values can be equivalent to the sum of the pixel values of the rectangular area.
12. The method of training a cascade classifier based on an AdaBoost framework training phase of claim 7, wherein: under the set training times T, a strong classifier is generated in each training, a plurality of weak classifiers are selected in the process, and finally each strong classifier and the corresponding weak classifiers are connected in series to form the final cascade classifier.
CN2012100428091A 2012-02-24 2012-02-24 Method and system for detecting and recognizing feature of vehicle in static image Pending CN103295021A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2012100428091A CN103295021A (en) 2012-02-24 2012-02-24 Method and system for detecting and recognizing feature of vehicle in static image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2012100428091A CN103295021A (en) 2012-02-24 2012-02-24 Method and system for detecting and recognizing feature of vehicle in static image

Publications (1)

Publication Number Publication Date
CN103295021A true CN103295021A (en) 2013-09-11

Family

ID=49095850

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2012100428091A Pending CN103295021A (en) 2012-02-24 2012-02-24 Method and system for detecting and recognizing feature of vehicle in static image

Country Status (1)

Country Link
CN (1) CN103295021A (en)

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103559501A (en) * 2013-10-25 2014-02-05 公安部第三研究所 Vehicle sun visor detecting method and device based on image analysis
CN103984950A (en) * 2014-04-22 2014-08-13 北京联合大学 Moving vehicle stop lamp state recognition method adaptable to day detection
CN105225523A (en) * 2015-10-15 2016-01-06 浙江宇视科技有限公司 A kind of parking space state detection method and device
CN105574490A (en) * 2015-12-10 2016-05-11 金鹏电子信息机器有限公司 Vehicle brand identification method and system based on headlight image characteristics
CN105809184A (en) * 2015-10-30 2016-07-27 哈尔滨工程大学 Vehicle real-time identification tracking and parking space occupancy determining method suitable for gas station
CN105956632A (en) * 2016-05-20 2016-09-21 浙江宇视科技有限公司 Target detection method and device
CN106203446A (en) * 2016-07-05 2016-12-07 中国人民解放军63908部队 Three dimensional object recognition positioning method for augmented reality auxiliary maintaining system
CN106909884A (en) * 2017-01-17 2017-06-30 北京航空航天大学 A kind of hand region detection method and device based on hierarchy and deformable part sub-model
CN107220650A (en) * 2017-05-31 2017-09-29 广州炒米信息科技有限公司 Food image detection method and device
CN107229906A (en) * 2017-05-08 2017-10-03 上海工程技术大学 A kind of automobile overtaking's method for early warning based on units of variance model algorithm
CN107273923A (en) * 2017-06-02 2017-10-20 浙江理工大学 A kind of construction method of weaving face fabric friction sound wave discriminator
WO2018036277A1 (en) * 2016-08-22 2018-03-01 平安科技(深圳)有限公司 Method, device, server, and storage medium for vehicle detection
CN108052955A (en) * 2017-11-03 2018-05-18 中国科学院计算技术研究所 A kind of high-precision braille recognition methods and system
CN109344684A (en) * 2018-08-03 2019-02-15 广东工业大学 Battery-operated motor cycle image detecting method based on units of variance model
CN110633751A (en) * 2019-09-17 2019-12-31 上海眼控科技股份有限公司 Training method of car logo classification model, car logo identification method, device and equipment
CN111080612A (en) * 2019-12-12 2020-04-28 哈尔滨市科佳通用机电股份有限公司 Truck bearing damage detection method

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7715597B2 (en) * 2004-12-29 2010-05-11 Fotonation Ireland Limited Method and component for image recognition
CN102419820A (en) * 2011-08-18 2012-04-18 电子科技大学 Method for rapidly detecting automobile logo in video and image

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7715597B2 (en) * 2004-12-29 2010-05-11 Fotonation Ireland Limited Method and component for image recognition
CN102419820A (en) * 2011-08-18 2012-04-18 电子科技大学 Method for rapidly detecting automobile logo in video and image

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
林志阳 等: "基于Adaboost的车标定位方法", 《计算机工程》, vol. 34, no. 11, 30 June 2008 (2008-06-30), pages 206 - 207 *
胡正平 等: "HOG特征混合模型结合隐SVM的感兴趣目标检测定位算法", 《信号处理》, vol. 27, no. 8, 31 August 2011 (2011-08-31), pages 1206 - 1212 *

Cited By (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103559501A (en) * 2013-10-25 2014-02-05 公安部第三研究所 Vehicle sun visor detecting method and device based on image analysis
CN103984950A (en) * 2014-04-22 2014-08-13 北京联合大学 Moving vehicle stop lamp state recognition method adaptable to day detection
CN103984950B (en) * 2014-04-22 2017-07-14 北京联合大学 A kind of moving vehicle brake light status recognition methods for adapting to detection on daytime
CN105225523A (en) * 2015-10-15 2016-01-06 浙江宇视科技有限公司 A kind of parking space state detection method and device
CN105809184A (en) * 2015-10-30 2016-07-27 哈尔滨工程大学 Vehicle real-time identification tracking and parking space occupancy determining method suitable for gas station
CN105809184B (en) * 2015-10-30 2020-06-16 哈尔滨工程大学 Method for real-time vehicle identification and tracking and parking space occupation judgment suitable for gas station
CN105574490A (en) * 2015-12-10 2016-05-11 金鹏电子信息机器有限公司 Vehicle brand identification method and system based on headlight image characteristics
CN105574490B (en) * 2015-12-10 2019-04-09 金鹏电子信息机器有限公司 Vehicle brand recognition methods and system based on headlight characteristics of image
CN105956632A (en) * 2016-05-20 2016-09-21 浙江宇视科技有限公司 Target detection method and device
CN105956632B (en) * 2016-05-20 2019-06-11 浙江宇视科技有限公司 A kind of method and apparatus detecting target
CN106203446A (en) * 2016-07-05 2016-12-07 中国人民解放军63908部队 Three dimensional object recognition positioning method for augmented reality auxiliary maintaining system
CN106203446B (en) * 2016-07-05 2019-03-12 中国人民解放军63908部队 Three dimensional object recognition positioning method for augmented reality auxiliary maintaining system
WO2018036277A1 (en) * 2016-08-22 2018-03-01 平安科技(深圳)有限公司 Method, device, server, and storage medium for vehicle detection
CN106909884A (en) * 2017-01-17 2017-06-30 北京航空航天大学 A kind of hand region detection method and device based on hierarchy and deformable part sub-model
CN106909884B (en) * 2017-01-17 2020-08-04 北京航空航天大学 Hand region detection method and device based on layered structure and deformable part model
CN107229906A (en) * 2017-05-08 2017-10-03 上海工程技术大学 A kind of automobile overtaking's method for early warning based on units of variance model algorithm
CN107220650A (en) * 2017-05-31 2017-09-29 广州炒米信息科技有限公司 Food image detection method and device
CN107220650B (en) * 2017-05-31 2020-07-17 广州炒米信息科技有限公司 Food image detection method and device
CN107273923A (en) * 2017-06-02 2017-10-20 浙江理工大学 A kind of construction method of weaving face fabric friction sound wave discriminator
CN107273923B (en) * 2017-06-02 2020-09-29 浙江理工大学 Construction method of textile fabric friction sound wave discriminator
CN108052955A (en) * 2017-11-03 2018-05-18 中国科学院计算技术研究所 A kind of high-precision braille recognition methods and system
CN108052955B (en) * 2017-11-03 2021-03-05 中国科学院计算技术研究所 High-precision Braille identification method and system
CN109344684A (en) * 2018-08-03 2019-02-15 广东工业大学 Battery-operated motor cycle image detecting method based on units of variance model
CN110633751A (en) * 2019-09-17 2019-12-31 上海眼控科技股份有限公司 Training method of car logo classification model, car logo identification method, device and equipment
CN111080612A (en) * 2019-12-12 2020-04-28 哈尔滨市科佳通用机电股份有限公司 Truck bearing damage detection method

Similar Documents

Publication Publication Date Title
CN103295021A (en) Method and system for detecting and recognizing feature of vehicle in static image
CN110348319B (en) Face anti-counterfeiting method based on face depth information and edge image fusion
CN109584248B (en) Infrared target instance segmentation method based on feature fusion and dense connection network
CN110443143B (en) Multi-branch convolutional neural network fused remote sensing image scene classification method
CN106650721B (en) A kind of industrial character identifying method based on convolutional neural networks
CN108154102B (en) Road traffic sign identification method
CN106845374B (en) Pedestrian detection method and detection device based on deep learning
CN105046196B (en) Front truck information of vehicles structuring output method based on concatenated convolutional neutral net
CN109684922B (en) Multi-model finished dish identification method based on convolutional neural network
CN112418074A (en) Coupled posture face recognition method based on self-attention
CN107103317A (en) Fuzzy license plate image recognition algorithm based on image co-registration and blind deconvolution
CN104598885B (en) The detection of word label and localization method in street view image
CN109978882A (en) A kind of medical imaging object detection method based on multi-modal fusion
CN110969171A (en) Image classification model, method and application based on improved convolutional neural network
CN111339975A (en) Target detection, identification and tracking method based on central scale prediction and twin neural network
CN110599463B (en) Tongue image detection and positioning algorithm based on lightweight cascade neural network
US8094971B2 (en) Method and system for automatically determining the orientation of a digital image
CN104408728A (en) Method for detecting forged images based on noise estimation
CN109255326A (en) A kind of traffic scene smog intelligent detecting method based on multidimensional information Fusion Features
CN112818850A (en) Cross-posture face recognition method based on progressive neural network and attention mechanism
CN106372658A (en) Vehicle classifier training method
CN105956570A (en) Lip characteristic and deep learning based smiling face recognition method
CN108932471A (en) A kind of vehicle checking method
CN115205155A (en) Distorted image correction method and device and terminal equipment
CN108960005B (en) Method and system for establishing and displaying object visual label in intelligent visual Internet of things

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20130911