CN106778560A - A kind of model recognizing method based on FHOG features and Linear SVM - Google Patents

A kind of model recognizing method based on FHOG features and Linear SVM Download PDF

Info

Publication number
CN106778560A
CN106778560A CN201611093783.8A CN201611093783A CN106778560A CN 106778560 A CN106778560 A CN 106778560A CN 201611093783 A CN201611093783 A CN 201611093783A CN 106778560 A CN106778560 A CN 106778560A
Authority
CN
China
Prior art keywords
fhog
vehicle
gradient
matrix
cell
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.)
Granted
Application number
CN201611093783.8A
Other languages
Chinese (zh)
Other versions
CN106778560B (en
Inventor
王海滨
马胜涛
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.)
In Department Of Science And Technology (beijing) Co Ltd Realism
Original Assignee
In Department Of Science And Technology (beijing) Co Ltd Realism
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 In Department Of Science And Technology (beijing) Co Ltd Realism filed Critical In Department Of Science And Technology (beijing) Co Ltd Realism
Priority to CN201611093783.8A priority Critical patent/CN106778560B/en
Publication of CN106778560A publication Critical patent/CN106778560A/en
Application granted granted Critical
Publication of CN106778560B publication Critical patent/CN106778560B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Probability & Statistics with Applications (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a kind of model recognizing method based on FHOG features and Linear SVM, it is characterised in that including step：A) vehicle detection and car plate detection, determine vehicle target region and car plate target area；B) license plate area is filled, whole license plate area is filled with car plate background colour；C) vehicle FHOG features, construction feature matrix are extracted；D) FHOG eigenmatrixes are launched by row to build FHOG characteristic vectors；E) target vehicle is recognized using FHOG characteristic vectors and the Linear SVM for training.

Description

A kind of model recognizing method based on FHOG features and Linear SVM
Technical field
A kind of model recognizing method based on FHOG features and Linear SVM of the present invention, designed image treatment, machine is regarded Feel, the field such as pattern-recognition.
Background technology
Vehicle cab recognition is the key technology in field of intelligent monitoring, and the vehicle cab recognition technology based on computer vision is to utilize The various features of vehicle image, the identification vehicle of vehicle target of active, brand and model；Target based on HOG features and SVM Identification technology comparative maturity, but HOG feature calculations amount is big, and intrinsic dimensionality is higher, in addition as various new car systems are continuous Occur, the data volume in a more comprehensive model data storehouse is inevitable very huge, consequently, it is possible to for using Non-linear Kernel letter For several SVM classifiers, its training and recognition speed cannot all meet requirement of real-time.
The content of the invention
The purpose of the application there are provided a kind of model recognizing method based on FHOG features and Linear SVM, and it is special Levy and be, including step：A) vehicle detection and car plate detection, determine vehicle target region and car plate target area；B) car is filled Board region, whole license plate area is filled with car plate background colour；C) vehicle FHOG features, construction feature matrix are extracted；D) will FHOG eigenmatrixes are launched to build FHOG characteristic vectors by row；E) known using FHOG characteristic vectors and the LinearSVM for training Other target vehicle.
Preferably, the vehicle detection of the step a is that background modeling and AdaBoost are detected into phase with detection method of license plate With reference to.
Preferably, the method for construction feature matrix is in the step c：
C1 gradient operator [- 1,0,1] and [- 1,0,1]) are used respectivelyTInput picture to m × n does convolution algorithm, obtains x, Gradient component G on y directionsx(x,y),Gy(x, y), and gradient magnitude M (x, y) and phase angle theta (x, y) are calculated, if input picture is Multichannel image, then take the Grad of amplitude maximum as the gradient of the point；
C2 the square cell that several length of sides are s) is divided the image into；
C3 it is) [0, π] by phase angle Interval Maps, is divided into b bin, the ladder of each cell is calculated using Tri linear interpolation Degree direction histogram, histogram data is saved as the Matrix C of (m/s) × (n/s) × b1, matrix element C1(i, j, k) represents i-th K-th value of bin in row jth row cell；
C4 it is) [0,2 π] by phase angle Interval Maps, is divided into 2b bin, calculates each cell's using Tri linear interpolation Gradient orientation histogram, histogram data is saved as the Matrix C of (m/s) × (n/s) × (2b)2, matrix element C2(i, j, k) table Show k-th value of bin in the i-th row jth row cell；
C5 the FHOG eigenmatrixes F of (m/s) × (n/s) × (3b+4)) is built：
Wherein No,p,q(i, j, k), o ∈ { 1,2 }, p, q ∈ { -1,1 } representing matrix Co, o ∈ { 1,2 } are at position (i, j, k) The normalization factor at place, is defined as
Tα(υ) represents that parameter is the truncation funcation of α, and it is defined as
Preferably, LinearSVM training object function used is in the step e：
Wherein ω is the coefficient of grader, and T represents transposition,The classification of presentation class device is missed Difference, yi∈ { 0,1 } is i-th class label of training sample, and C is penalty factor, represents the attention degree to error in classification.
It should be appreciated that foregoing description substantially and follow-up description in detail are exemplary illustration and explanation, should not As the limitation to claimed content of the invention.
Brief description of the drawings
With reference to the accompanying drawing enclosed, the present invention more purpose, function and advantages are by by the as follows of embodiment of the present invention Description is illustrated, wherein：
Fig. 1 shows a kind of flow of model recognizing method based on gradient orientation histogram feature of the invention Figure.
Specific embodiment
By reference to one exemplary embodiment, the purpose of the present invention and function and the side for realizing these purposes and function Method will be illustrated.However, the present invention is not limited to one exemplary embodiment as disclosed below；Can by multi-form come It is realized.The essence of specification is only to aid in various equivalent modifications Integrated Understanding detail of the invention.
Hereinafter, embodiments of the invention will be described with reference to the drawings.In the accompanying drawings, identical reference represents identical Or similar part, or same or like step.
As shown in figure 1, a kind of reality of model recognizing method based on FHOG features and Linear SVM that the present invention is provided Applying step is：
Step 110：The method being combined using background modeling and target detection carries out vehicle detection and car plate detection, it is determined that Vehicle target and car plate target region；
According to one embodiment of present invention, the vehicle detection of the step a and detection method of license plate are by background modeling It is combined with AdaBoost detections.
Step 120：In order to reduce the influence of characters on license plate, license plate area is filled with car plate background colour.
Step 130：Vehicle image feature is extracted, FHOG eigenmatrixes are built, its step is：
Step 131：Gradient operator [- 1,0,1] and [- 1,0,1] are used respectivelyTInput picture to m × n does convolution algorithm, X is obtained, the gradient component G on y directionsx(x,y),Gy(x, y), and gradient magnitude M (x, y) and phase angle theta (x, y) are calculated, if input Image is multichannel image, then take the Grad of amplitude maximum as the gradient of the point
Step 132：Divide the image into the square cell that several length of sides are s；
Step 133：It is [0, π] by phase angle Interval Maps, is divided into b bin, each cell is calculated using Tri linear interpolation Gradient orientation histogram, histogram data is saved as the Matrix C of (m/s) × (n/s) × b1, matrix element C1(i, j, k) is represented K-th value of bin in i-th row jth row cell；
Step 134：It is [0,2 π] by phase angle Interval Maps, is divided into 2b bin, each is calculated using Tri linear interpolation The gradient orientation histogram of cell, histogram data is saved as the Matrix C of (m/s) × (n/s) × (2b)2, matrix element C2(i, J, k) represent k-th value of bin in the i-th row jth row cell；
Step 135：Build the FHOG eigenmatrixes F of (m/s) × (n/s) × (3b+4)：
Wherein No,p,q(i, j, k), o ∈ { 1,2 }, p, q ∈ { -1,1 } representing matrix Co, o ∈ { 1,2 } are at position (i, j, k) The normalization factor at place, is defined as
Tα(υ) represents that parameter is the truncation funcation of α, and it is defined as
Step 140：FHOG eigenmatrixes are launched by row to build FHOG characteristic vectors；
Step 150：Target vehicle is recognized using FHOG characteristic vectors and the LinearSVM for training.
According to one embodiment of present invention, LinearSVM training object function used is in the step e：
Wherein ω is the coefficient of grader, and T represents transposition,The classification of presentation class device is missed Difference, yi∈ { 0,1 } is i-th class label of training sample, and C is penalty factor, represents the attention degree to error in classification.
Compared with traditional HOG features, intrinsic dimensionality is substantially reduced FHOG features, in the case where recognition effect is suitable significantly Improve recognition speed；The Linear SVM classifiers built using linear kernel function are non-linear due to that need not be carried out to feature Mapping, training and recognition efficiency are higher, and experiment shows, in high-capacity database, using linear kernel function and using non- The recognition effect of linear kernel function is of slight difference.
With reference to the explanation of the invention and practice that disclose here, other embodiment of the invention is for those skilled in the art All will be readily apparent and understand.Illustrate and embodiment is to be considered only as exemplary, true scope of the invention and purport are equal It is defined in the claims.

Claims (4)

1. a kind of model recognizing method based on FHOG features and Linear SVM, it is characterised in that including step：
A) vehicle detection and car plate detection, determine vehicle target region and car plate target area；
B) license plate area is filled, whole license plate area is filled with car plate background colour；
C) vehicle FHOG features, construction feature matrix are extracted；
D) FHOG eigenmatrixes are launched by row to build FHOG characteristic vectors；
E) target vehicle is recognized using FHOG characteristic vectors and the Linear SVM for training.
2. method according to claim 1, it is characterised in that：The vehicle detection of the step a is with detection method of license plate Background modeling is combined with AdaBoost detections.
3. method according to claim 1, it is characterised in that：The method of construction feature matrix is in the step c：
C1 gradient operator [- 1,0,1] and [- 1,0,1]) are used respectivelyTInput picture to m × n does convolution algorithm, obtains x, y directions On gradient component Gx(x,y),Gy(x, y), and gradient magnitude M (x, y) and phase angle theta (x, y) are calculated, if input picture is more logical Road image, then take the Grad of amplitude maximum as the gradient of the point；
C2 the square cell that several length of sides are s) is divided the image into；
C3 it is) [0, π] by phase angle Interval Maps, is divided into b bin, the gradient side of each cell is calculated using Tri linear interpolation To histogram, histogram data is saved as the Matrix C of (m/s) × (n/s) × b1, matrix element C1(i, j, k) represents the i-th row jth K-th value of bin in row cell；
C4 it is) [0,2 π] by phase angle Interval Maps, is divided into 2b bin, the gradient of each cell is calculated using Tri linear interpolation Direction histogram, histogram data is saved as the Matrix C of (m/s) × (n/s) × (2b)2, matrix element C2(i, j, k) represents i-th K-th value of bin in row jth row cell；
C5 the FHOG eigenmatrixes F of (m/s) × (n/s) × (3b+4)) is built：
Wherein No,p,q(i, j, k), o ∈ { 1,2 }, p, q ∈ { -1,1 } representing matrix Co, o ∈ { 1,2 } are at position (i, j, k) place Normalization factor, is defined as
Tα(υ) represents that parameter is the truncation funcation of α, and it is defined as
。
4. method according to claim 1, it is characterised in that：Target used by the training of Linear SVM in the step e Function is：
Wherein ω is the coefficient of grader, and T represents transposition,The error in classification of presentation class device, yi∈ { 0,1 } is i-th class label of training sample, and C is penalty factor, represents the attention degree to error in classification.
CN201611093783.8A 2016-12-01 2016-12-01 Vehicle type recognition method based on FHOG characteristics and Linear SVM Active CN106778560B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611093783.8A CN106778560B (en) 2016-12-01 2016-12-01 Vehicle type recognition method based on FHOG characteristics and Linear SVM

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611093783.8A CN106778560B (en) 2016-12-01 2016-12-01 Vehicle type recognition method based on FHOG characteristics and Linear SVM

Publications (2)

Publication Number Publication Date
CN106778560A true CN106778560A (en) 2017-05-31
CN106778560B CN106778560B (en) 2020-08-04

Family

ID=58882768

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611093783.8A Active CN106778560B (en) 2016-12-01 2016-12-01 Vehicle type recognition method based on FHOG characteristics and Linear SVM

Country Status (1)

Country Link
CN (1) CN106778560B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111476271A (en) * 2020-03-10 2020-07-31 杭州易现先进科技有限公司 Icon identification method, device, system, computer equipment and storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080025568A1 (en) * 2006-07-20 2008-01-31 Feng Han System and method for detecting still objects in images
US20120027263A1 (en) * 2010-08-02 2012-02-02 Sony Corporation Hand gesture detection
CN104063722A (en) * 2014-07-15 2014-09-24 国家电网公司 Safety helmet identification method integrating HOG human body target detection and SVM classifier
CN105426905A (en) * 2015-11-02 2016-03-23 国网山东省电力公司电力科学研究院 Robot barrier identification method based on gradient histogram and support vector machine
CN105574895A (en) * 2016-01-05 2016-05-11 浙江博天科技有限公司 Congestion detection method during the dynamic driving process of vehicle
CN106056086A (en) * 2016-06-02 2016-10-26 江苏华通晟云科技有限公司 Vehicle brand and model identification method based on fast learning framework

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080025568A1 (en) * 2006-07-20 2008-01-31 Feng Han System and method for detecting still objects in images
US20120027263A1 (en) * 2010-08-02 2012-02-02 Sony Corporation Hand gesture detection
CN104063722A (en) * 2014-07-15 2014-09-24 国家电网公司 Safety helmet identification method integrating HOG human body target detection and SVM classifier
CN105426905A (en) * 2015-11-02 2016-03-23 国网山东省电力公司电力科学研究院 Robot barrier identification method based on gradient histogram and support vector machine
CN105574895A (en) * 2016-01-05 2016-05-11 浙江博天科技有限公司 Congestion detection method during the dynamic driving process of vehicle
CN106056086A (en) * 2016-06-02 2016-10-26 江苏华通晟云科技有限公司 Vehicle brand and model identification method based on fast learning framework

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
BIN LENG ET AL: "An improved pedestrians detection algorithm using HOG and ViBe", 《2013 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND BIOMIMETICS (ROBIO)》 *
PEDRO F.FELZENSZWALB ET AL: "Object Detection with Discriminatively Trained Part-Based Models", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 *
STEFAN BADURA ET AL: "Insight into make and model recognition of cars-SVM and FHOG design", 《THE 3RD INTERNATIONAL VIRTUAL RESEARCH CONFERENCE IN TECHNICAL DISCIPLINES》 *
吴长侠: "低空对地运动车辆检测与运动特性分析", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
李星等: "基于HOG特征和SVM的前向车辆识别方法", 《计算机科学》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111476271A (en) * 2020-03-10 2020-07-31 杭州易现先进科技有限公司 Icon identification method, device, system, computer equipment and storage medium
CN111476271B (en) * 2020-03-10 2023-07-21 杭州易现先进科技有限公司 Icon identification method, device, system, computer equipment and storage medium

Also Published As

Publication number Publication date
CN106778560B (en) 2020-08-04

Similar Documents

Publication Publication Date Title
CN102982349B (en) A kind of image-recognizing method and device
Hu et al. Fast detection of multiple objects in traffic scenes with a common detection framework
Aziz et al. Traffic sign recognition based on multi-feature fusion and ELM classifier
CN109918969B (en) Face detection method and device, computer device and computer readable storage medium
CN104392241B (en) A kind of head pose estimation method returned based on mixing
CN102708370B (en) Method and device for extracting multi-view angle image foreground target
CN102999926B (en) A kind of image vision significance computational methods merged based on low-level image feature
CN105005989A (en) Vehicle target segmentation method under weak contrast
CN107464245B (en) Image structure edge positioning method and device
CN116188999B (en) Small target detection method based on visible light and infrared image data fusion
CN105069774A (en) Object segmentation method based on multiple-instance learning and graph cuts optimization
CN102663401A (en) Image characteristic extracting and describing method
CN103839066A (en) Feature extraction method based on biological vision
CN107895021B (en) image recognition method and device, computer device and computer readable storage medium
CN101645135A (en) Method and system for human detection
CN104050460A (en) Pedestrian detection method with multi-feature fusion
CN103295026B (en) Based on the image classification method of space partial polymerization description vectors
CN106778560A (en) A kind of model recognizing method based on FHOG features and Linear SVM
CN105069455B (en) A kind of method and device of invoice official seal filtering
Jang et al. Object classification using CNN for video traffic detection system
Mannan et al. Optimized segmentation and multiscale emphasized feature extraction for traffic sign detection and recognition
Vashisth et al. Histogram of Oriented Gradients based reduced feature for traffic sign recognition
CN105160677B (en) A kind of multi-characteristic image suitable for space non-cooperative target splits localization method
Frejlichowski Application of the polar–fourier greyscale descriptor to the automatic traffic sign recognition
CN115035331A (en) Small sample target detection method and system applied to small unmanned aerial vehicle-mounted platform

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant