CN106919910B - Traffic sign identification method based on HOG-CTH combined features - Google Patents
Traffic sign identification method based on HOG-CTH combined features Download PDFInfo
- Publication number
- CN106919910B CN106919910B CN201710075888.9A CN201710075888A CN106919910B CN 106919910 B CN106919910 B CN 106919910B CN 201710075888 A CN201710075888 A CN 201710075888A CN 106919910 B CN106919910 B CN 106919910B
- Authority
- CN
- China
- Prior art keywords
- traffic sign
- cth
- hog
- gradient
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
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
- G06V20/582—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of traffic signs
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/09—Recognition of logos
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a traffic sign identification method based on HOG-CTH combined characteristics, which comprises the following steps: training a classifier model by using a training sample; detecting a traffic sign image in the positioning live-action picture; extracting HOG-CTH fusion characteristic vectors of the positioned traffic sign images; and identifying the traffic sign type of the test picture. The traffic sign identification method based on the HOG-CTH combined characteristics can reduce the calculation complexity to a great extent, can obtain higher identification rate and has stronger robustness.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a traffic sign recognition method based on HOG-CTH combined features.
Background
The development of the automobile industry greatly facilitates the work and life of people, but the traffic safety problem is also avoided. One approach to solve the traffic safety problem is to accurately and effectively set up road traffic signs, so as to provide drivers with abundant information such as prohibition, warning, indication and the like, thereby reducing traffic accidents. In order to ensure timely and accurate transmission of Traffic Sign information, a Traffic Sign Recognition (TSR) system has received attention from various researchers. The traffic sign recognition system has wide application prospect, is mainly applied to a plurality of fields such as driving assistance, traffic sign maintenance, unmanned driving and the like, relates to a plurality of disciplines such as machine vision, mode recognition, image processing, digital signal processing, artificial intelligence, communication and information technology and the like, and is a typical mode recognition application system as well as face recognition and target tracking.
The traffic sign recognition system mainly comprises two basic technical links: firstly, detecting a traffic sign, including positioning the traffic sign and necessary preprocessing; and secondly, the classification and the identification of the traffic signs, including the feature extraction and the classification of the traffic signs. In the detection stage, methods such as threshold segmentation and template matching are mainly adopted; in the classification stage, solutions are mainly sought from the aspects of learners and image features, namely, suitable learners are adopted to extract suitable image features to complete the final traffic sign identification task. Generally, the traffic sign identification method at the present stage has low identification accuracy and long operation time, and is difficult to meet the requirement of vehicle-mounted instantaneity.
Disclosure of Invention
Aiming at the problems and the defects in the prior art, the invention aims to provide a traffic sign identification method based on HOG-CTH combined characteristics.
In order to achieve the purpose, the invention adopts the following technical scheme:
a traffic sign identification method based on HOG-CTH combined features comprises the following steps:
s1: training classifier models using training samples
Step S1-1: determining a training set;
step S1-2: respectively extracting the HOG (histogram of oriented gradient) features and the CTH (statistical transformation histogram) features of the images of the training samples in the training set determined in the step S1-1, performing refinement on the CTH features, and connecting the HOG and the CTH features in series to obtain HOG-CTH fusion features;
step S1-3: training the HOG-CTH fusion feature vector obtained in the step S1-2 by using a linear Support Vector Machine (SVM) algorithm to obtain an SVM traffic sign classifier;
s2: detecting and positioning traffic sign image in live-action picture
Step S2-1: carrying out color segmentation on the traffic sign in the natural scene in the HSV color space;
step S2-2: performing morphological image processing on the region subjected to color segmentation in the step S2-1;
step S2-3: positioning and cutting out the divided traffic sign images;
s3: identifying traffic sign images using classifier models
Step S3-1: carrying out graying processing on the traffic sign image in the detected live-action image in the S2;
step S3-2: extracting HOG-CTH fusion characteristics of the traffic sign image subjected to gray processing;
step S3-3: and identifying the type of the traffic sign by using the trained SVM traffic sign classifier in the S1.
The step S1-2 is specifically as follows:
the gamma correction method is adopted to carry out color space normalization on the input traffic sign image, the input traffic sign image is firstly converted into a gray level image, and I is set g (x, y) is the gray value of the pixel point of the (x, y) coordinate, and the gamma compression formula:
I g (x,y)=I g (x,y) gamma take gamma =0.5 (1)
Calculating the gradient amplitude and direction of each pixel of the traffic sign image, and performing convolution calculation on the traffic sign image by using a 3 multiplied by 3Sobel template, wherein I is g The gradient of (x, y) is:
G x (x,y)、G y (x, y) represent the gradients of the horizontal and vertical edge detection respectively, and the gradient magnitude G (x, y) and the direction theta (x, y) of the pixel point (x, y) are shown as formula (4) and formula (5):
the gradients of all color channels of the color images can be respectively calculated, and the value with the maximum amplitude is selected as the gradient of the pixel point;
counting gradient histogram in cell, dividing image window area into uniformly distributed cell units, each cell unit containing 4 × 4 pixels, and dividing the cell unit into multiple cellsThe gradient direction of the cell is averagely divided into 9 bin intervals, then histogram statistics is carried out on gradient values of all pixels in each cell unit in each bin interval respectively, and direction gradient histograms in the cell units are counted respectively;
normalizing the histogram of gradient directions in the region block, forming a block by every 2 multiplied by 2 cells, forming a 36-dimensional characteristic vector by one block, and normalizing the whole block by using an L2-norm, wherein the formula is shown in formula (6):
wherein v is a feature vector, | v | | non-calculation 2 2 norm of v, epsilon represents a constant to avoid denominator of 0;
and (4) forming HOG feature vectors through serialization, and carrying out serialization processing on the HOG features of all block blocks to form the final HOG feature vectors of training samples or detection window images.
Calculating the statistical transformation of the pixel points, and sequentially numbering clockwise: p is a radical of 0 ~p 7 (ii) a Obtaining a symbol difference value between pixel points according to the amplitude relation:
T=t(s(p c -p 0 ),s(p c -p 1 ),…,s(p c -p 7 )) (7)
wherein:
in the formula, the t () function represents a joint distribution function of the symbol difference values; s (p) c -p i ) Representing the current pixel point p c And the ith neighborhood point p i A sign difference value therebetween; at this moment, the current pixel point p c Is calculated by equation (9):
in the formula, N is the number of neighborhood pixels, and R is the CTH calculation radius;
fine quantization processing: introducing the concept of 0-1 transformation, wherein 0 to 1 or 1 to 0 is regarded as one 0-1 transformation, and the central pixel point p c The degree of 0-1 transformation u of (a) can be calculated by the following equation (10):
unifying the CTH characteristic value according to the 0-1 transformation times of the pixel points, as shown in formula (11):
obtaining corresponding characteristics by counting the histogram of the sparse CTH characteristic values;
counting a block histogram and carrying out normalization processing, and carrying out normalization processing on a CTH histogram in the block by adopting an L2-norm so as to fuse two types of features of HOG and CTH to obtain a HOG-CTH fusion feature vector; and putting the HOG-CTH fusion feature vector into an SVM algorithm of a linear support vector machine for training to obtain an SVM traffic sign classifier.
Has the advantages that: compared with the prior art, the method is used for identifying the traffic sign based on the HOG-CTH fusion characteristics, in the detection stage, a traffic sign detection method based on an HSV color space is used for carrying out color threshold segmentation in the HSV space, then the segmented area is filled and expanded, a series of characteristics are marked, then candidate areas are screened by calculating whether the characteristics meet the conditions, and finally the area of the traffic sign is cut from an original image to finish the detection. The method has constant scale, can perform reliable traffic sign detection in a complex traffic sign scene, and is convenient for the next step of identification. In the identification stage, the gradient feature and the texture feature are combined, the description performance is richer than that of a single feature, and the limitation of the single feature can be made up, so that the identification rate is improved. When the identification rate is improved by adopting the fusion features, the CTH features are finely quantized, so that the feature dimension is greatly reduced, the identification time is greatly shortened, and the robustness of the system is improved.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of the HOG-CTH combined feature extraction in the present invention;
fig. 3 is a CTH characteristic calculation chart in the present invention.
Detailed Description
The technical solution of the present invention is described in detail below with reference to the accompanying drawings.
As shown in fig. 1, a traffic sign identification method based on HOG-CTH combination features includes the following steps:
step A: selecting a training sample, wherein a national traffic sign database is selected as a positive sample, a sample set contains 43 different traffic signs, each traffic sign is a category, category labels are set for the 43 different traffic signs, and a plurality of image sets which do not contain the traffic signs are randomly shot as negative samples. And B, step B: as shown in FIG. 2, the HOG (Histogram of ordered) of the training sample in step A is extracted
(ii) a Gradient feature;
step B1: the gamma correction method is adopted to carry out color space normalization on the input image, and square root gamma normalization can well eliminate the influence of integral illumination and contrast of the image. Since the color information has little effect, it is usually converted into a gray scale image, set I g (x, y) is the gray value of the pixel point of the (x, y) coordinate, and the gamma compression formula:
I g (x,y)=I g (x,y) gamma take gamma =0.5 (1)
And step B2: calculating the gradient amplitude and direction of each pixel of the image, and performing convolution calculation on the traffic sign image by using a 3 multiplied by 3Sobel template, so as to obtain the value I g The gradient of (x, y) is:
G x (x,y)、G y (x, y) represent the gradients of the horizontal and vertical edge detection respectively, and the gradient magnitude G (x, y) and the direction theta (x, y) of the pixel point (x, y) are shown as formula (4) and formula (5):
the gradients of all color channels of the color images can be respectively calculated, and the value with the maximum amplitude is selected as the gradient of the pixel point;
and step B3: counting gradient histogram in cell, dividing image window area into uniformly distributed cell (cell) units, each cell unit comprises 4 × 4 pixels, and counting gradient histogram in each cell unitThe gradient direction of the cell is averagely divided into 9 intervals (bins), then the gradient values of all pixels in each cell unit are subjected to histogram statistics in each bin interval respectively, so that a 9-dimensional feature vector is obtained by one cell unit, and directional gradient histograms in the cell units are respectively counted;
and step B4: normalizing the histogram of gradient directions in the region block, forming a block (block) by every 2 × 2 cell units, forming a 36-dimensional feature vector by using one block, and normalizing the whole block by using an L2-norm, as shown in formula (6):
wherein v is a feature vector, | v | | non-calculation 2 Represents the norm of order 2 of v, and epsilon represents a small constant to avoid a denominator of 0;
and step B5: and (4) forming HOG feature vectors through serialization, and forming the final HOG feature vectors of training samples or detection window images by carrying out serialization processing on the HOG features of all block blocks.
And C: extracting the CTH (CENTs Transform Histogram, CENT RIST) characteristics of the training sample in the step A;
step C1: calculating the statistical transformation of the pixel point, as shown in (a) of FIG. 3, for the current calculated pixel point p c And the space relation graph between the adjacent points is sequentially numbered clockwise as follows: p is a radical of 0 ~p 7 (ii) a Fig. 3 (b) shows the gray scale of the current pixel. Based on the amplitude relationship, the pixel can be obtainedThe symbol difference value between them is shown in equations (7) and (8):
T=t(s(p c -p 0 ),s(p c -p 1 ),…,s(p c -p 7 )) (7)
wherein:
in the formula, the t () function represents a joint distribution function of the symbol difference values; s (p) c -p i ) Representing the current pixel point p c And the ith neighborhood point p i The sign difference value between them. As shown in fig. 3 (c), the corresponding sign difference values are for all neighborhood points. Fig. 3 (d) shows a weight template for CTH feature calculation, where the current pixel point p c The corresponding CTH value of (a) can be calculated by equation (9):
in the formula, N is the number of the neighborhood pixels, R is the CTH calculation radius, and the calculated CTH value of the current pixel is 47. It can be seen that the CTH value is only related to the relative relationship between the pixels, and not to the specific pixel amplitude. The corresponding image features can be obtained by calculating a statistical histogram of the CTH values.
And step C2: in the fine quantization process, in order to avoid an excessively large dimension of the CTH feature, it is necessary to thin the CTH feature. First, the concept of a 0-1 transform is introduced. A 0-1 transformation is considered to be a 0-1 transformation from 0 to 1 or from 1 to 0. Thus, the center pixel point p c The degree of 0-1 transformation u of (a) can be calculated by the following equation (10):
at this time, the CTH value can be unified according to the 0-1 transformation times of the pixel points, as shown in equation (11):
corresponding characteristics can be obtained by counting the histogram of the sparse CTH value;
and C3: the histogram of the block is counted and normalized, in order to fuse HOG and CTH characteristics, the invention adopts a unified normalization scheme for the HOG characteristics and the CTH characteristics, namely, the normalization processing is carried out on the CTH histogram in the block by adopting an L2-norm, thereby combining/fusing the two characteristics.
Step D: and putting the combined features into a linear Support Vector Machine (SVM) algorithm for training to obtain the traffic sign classifier.
And E, step E: performing color segmentation on the traffic sign under a natural scene under an HSV (Hue, saturation) color space; let (r, g, b) be the red, green and blue coordinates of a certain color, respectively, and be the real number in the interval [0,1], let max equal to the maximum value among r, g and b, min be the minimum value, and the RGB-to-HSV spatial conversion formula is:
V=max (14)
to facilitate the selection of the threshold, H, S, V is normalized to [0,255]. The threshold value for the segmentation of the red, yellow and blue colors is as follows:
red: h e (0,10) U (240,255), S e (40,255), V e (30,255)
Yellow: h e (18,45), S e (148,255), V e (66,255)
Blue color: h e (140,255), S e (60,255), V e (20,255)
Judging whether the current pixel point is an interested color point (red, yellow and blue) by utilizing the threshold value; if the point is the interest point, the pixel value is set to be white, otherwise, the pixel value is set to be black, and therefore the binarization of the image is completed.
Step F: e, positioning and cutting out a traffic sign image, and performing morphological closing operation on the image obtained in the step E to fill up the hole; expanding the binary image, filling the inner cavity, and labeling each separated part of a communicated object, wherein 8-way communication is adopted; a series of attributes of the labeled region are then measured, using 3 features: sequentially calculating the total number of pixels in each region of the image, the center of mass (gravity center) of each region and the minimum rectangle containing the corresponding region; then according to the area of each filling block, the largest 3 filling blocks are found and stored.
Taking one of the filling blocks as an example, M is the minimum value of the length and the width of the minimum rectangle circumscribing the region, X is the abscissa of the gravity center of the region, Y is the ordinate of the gravity center, H and L are the row number and the column number of the pixels of the original image, respectively, T is the name of the binary image just before labeling, a is the total number of the pixels of the region, as a traffic sign candidate region, the following five sets of conditions must be satisfied:
A/(H×L)>0.01 (19)
according to the conditions, at most 3 mark areas are determined, then, according to the horizontal and vertical coordinates of the pixel at the upper left corner of the corresponding rectangle of each area and the length and width of the rectangle, the original image is cut, and the cut image is the traffic mark image detected after positioning.
Step G: and extracting HOG-CTH fusion feature vectors of the cut and positioned traffic sign images, and putting the fusion feature vectors into a trained SVM classifier for classification and identification.
The invention discloses a traffic sign identification method based on HOG-CTH combined characteristics, which comprises the following steps: training a classifier model by using a training sample; detecting a traffic sign image in the positioning live-action picture; extracting HOG-CTH fusion characteristic vectors of the positioned traffic sign images; and identifying the traffic sign type of the test picture. The invention can reduce the calculation complexity to a great extent, and can also obtain higher recognition rate and stronger robustness.
Claims (2)
1. A traffic sign identification method based on HOG-CTH combined features is characterized by comprising the following steps:
s1: training classifier models using training samples
Step S1-1: determining a training set;
step S1-2: respectively extracting the HOG (histogram of oriented gradient) features and the CTH (statistical transformation histogram) features of the images of the training samples in the training set determined in the step S1-1, performing refinement on the CTH features, and connecting the HOG and the CTH features in series to obtain HOG-CTH fusion features;
step S1-3: training the HOG-CTH fusion feature vector obtained in the step S1-2 by using a linear Support Vector Machine (SVM) algorithm to obtain an SVM traffic sign classifier;
s2: detecting and positioning traffic sign image in live-action picture
Step S2-1: carrying out color segmentation on the traffic sign in the natural scene in the HSV color space;
step S2-2: performing morphological image processing on the region subjected to color segmentation in the step S2-1;
step S2-3: positioning and cutting out the divided traffic sign images;
s3: identifying traffic sign images using classifier models
Step S3-1: carrying out graying processing on the traffic sign image in the detected live-action image in the S2;
step S3-2: extracting HOG-CTH fusion characteristics of the traffic sign image subjected to gray processing;
step S3-3: and identifying the type of the traffic sign by using the trained SVM traffic sign classifier in the S1.
2. The method for recognizing the traffic sign based on the HOG-CTH combined features as claimed in claim 1, wherein the step S1-2 is specifically as follows:
the gamma correction method is adopted to carry out color space normalization on the input traffic sign image, the input traffic sign image is firstly converted into a gray level image, and I is set g (x, y) is the gray value of the pixel point of the (x, y) coordinate, and the gamma compression formula:
I g (x,y)=I g (x,y) gamma take gamma =0.5 (1)
Calculating the gradient amplitude and direction of each pixel of the traffic sign image, and performing convolution calculation on the traffic sign image by using a 3 multiplied by 3Sobel template, wherein I is g The gradient of (x, y) is:
G x (x,y)、G y (x, y) represent the gradient of the horizontal and vertical edge detection respectively, and the magnitude G (x, y) and direction θ (x, y) of the gradient of the pixel point (x, y) are shown in the formula (4) and the formula (5):
the gradients of all color channels of the color images can be respectively calculated, and the value with the maximum amplitude is selected as the gradient of the pixel point;
counting gradient histogram in cell, dividing image window area into uniformly distributed cell units, each cell unit comprises 4 × 4 pixels, and each cell unit is internally provided with a plurality of pixelsThe gradient direction of the cell is averagely divided into 9 bin intervals, then histogram statistics is carried out on gradient values of all pixels in each cell unit in each bin interval respectively, and direction gradient histograms in the cell units are counted respectively;
normalizing the histogram of gradient directions in the region block, forming a block by every 2 multiplied by 2 cells, forming a 36-dimensional characteristic vector by one block, and normalizing the whole block by using an L2-norm, wherein the formula is shown in formula (6):
wherein v is a feature vector, | v | | calucity 2 2 norm of v, epsilon represents a constant to avoid denominator of 0;
forming HOG characteristic vectors through serialization, and carrying out serialization processing on HOG characteristics of all block blocks to form final HOG characteristic vectors of training samples or detection window images;
calculating the statistical transformation of the pixel points, and sequentially numbering clockwise: p is a radical of formula 0 ~p 7 (ii) a Obtaining a symbol difference value between pixel points according to the amplitude relation:
T=t(s(p c -p 0 ),s(p c -p 1 ),…,s(p c -p 7 )) (7)
wherein:
in the formula, the t () function represents a joint distribution function of the symbol difference values; s (p) c -p i ) Representing the current pixel point p c And the ith neighborhood point p i A sign difference value therebetween; at this moment, the current pixel point p c Is calculated by equation (9):
in the formula, N is the number of neighborhood pixels, and R is the CTH calculation radius;
fine quantization processing: introducing the concept of 0-1 transformation, wherein 0 to 1 or 1 to 0 is regarded as one 0-1 transformation, and the central pixel point p c The degree of 0-1 transformation u of (a) can be calculated by the following equation (10):
unifying the CTH characteristic value according to the 0-1 transformation times of the pixel points, as shown in formula (11):
obtaining corresponding characteristics by counting a histogram of the sparse CTH characteristic values;
counting a block histogram and carrying out normalization processing, and carrying out normalization processing on a CTH histogram in a block by adopting an L2-norm so as to fuse two types of characteristics, namely HOG and CTH, to obtain a HOG-CTH fusion characteristic vector; and putting the HOG-CTH fusion feature vector into an SVM algorithm of a linear support vector machine for training to obtain an SVM traffic sign classifier.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2016103191156 | 2016-05-12 | ||
CN201610319115 | 2016-05-12 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106919910A CN106919910A (en) | 2017-07-04 |
CN106919910B true CN106919910B (en) | 2023-03-24 |
Family
ID=59453581
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710075888.9A Active CN106919910B (en) | 2016-05-12 | 2017-02-13 | Traffic sign identification method based on HOG-CTH combined features |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106919910B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107622229B (en) * | 2017-08-29 | 2021-02-02 | 中山大学 | Video vehicle re-identification method and system based on fusion features |
CN108009472B (en) * | 2017-10-25 | 2020-07-21 | 五邑大学 | Finger back joint print recognition method based on convolutional neural network and Bayes classifier |
CN109190451B (en) * | 2018-07-12 | 2021-07-06 | 广西大学 | Remote sensing image vehicle detection method based on LFP characteristics |
CN109086687A (en) * | 2018-07-13 | 2018-12-25 | 东北大学 | The traffic sign recognition method of HOG-MBLBP fusion feature based on PCA dimensionality reduction |
CN110801208B (en) * | 2019-11-27 | 2022-04-05 | 东北师范大学 | Tooth crack detection method and system |
CN113420633B (en) * | 2021-06-18 | 2022-04-12 | 桂林电子科技大学 | Traffic sign identification method based on UM enhancement and SIFT feature extraction |
CN113674270B (en) * | 2021-09-06 | 2023-11-14 | 深邦智能科技集团(青岛)有限公司 | Tire pattern consistency detection system and method thereof |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102881160A (en) * | 2012-07-18 | 2013-01-16 | 广东工业大学 | Outdoor traffic sign identification method under low-illumination scene |
CN103366190A (en) * | 2013-07-26 | 2013-10-23 | 中国科学院自动化研究所 | Method for identifying traffic sign |
CN103824081A (en) * | 2014-02-24 | 2014-05-28 | 北京工业大学 | Method for detecting rapid robustness traffic signs on outdoor bad illumination condition |
CN104992165A (en) * | 2015-07-24 | 2015-10-21 | 天津大学 | Extreme learning machine based traffic sign recognition method |
-
2017
- 2017-02-13 CN CN201710075888.9A patent/CN106919910B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102881160A (en) * | 2012-07-18 | 2013-01-16 | 广东工业大学 | Outdoor traffic sign identification method under low-illumination scene |
CN103366190A (en) * | 2013-07-26 | 2013-10-23 | 中国科学院自动化研究所 | Method for identifying traffic sign |
CN103824081A (en) * | 2014-02-24 | 2014-05-28 | 北京工业大学 | Method for detecting rapid robustness traffic signs on outdoor bad illumination condition |
CN104992165A (en) * | 2015-07-24 | 2015-10-21 | 天津大学 | Extreme learning machine based traffic sign recognition method |
Also Published As
Publication number | Publication date |
---|---|
CN106919910A (en) | 2017-07-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106919910B (en) | Traffic sign identification method based on HOG-CTH combined features | |
CN108108761B (en) | Rapid traffic signal lamp detection method based on deep feature learning | |
CN107545239B (en) | Fake plate detection method based on license plate recognition and vehicle characteristic matching | |
CN110705639B (en) | Medical sperm image recognition system based on deep learning | |
CN108090459B (en) | Traffic sign detection and identification method suitable for vehicle-mounted vision system | |
CN104573685B (en) | A kind of natural scene Method for text detection based on linear structure extraction | |
CN105809138A (en) | Road warning mark detection and recognition method based on block recognition | |
CN108921120B (en) | Cigarette identification method suitable for wide retail scene | |
CN104134079A (en) | Vehicle license plate recognition method based on extremal regions and extreme learning machine | |
CN103544484A (en) | Traffic sign identification method and system based on SURF | |
CN103413147A (en) | Vehicle license plate recognizing method and system | |
CN104408424A (en) | Multiple signal lamp recognition method based on image processing | |
CN102915544A (en) | Video image motion target extracting method based on pattern detection and color segmentation | |
CN112464731B (en) | Traffic sign detection and identification method based on image processing | |
CN109145964B (en) | Method and system for realizing image color clustering | |
CN107330365A (en) | Traffic sign recognition method based on maximum stable extremal region and SVM | |
CN108073940B (en) | Method for detecting 3D target example object in unstructured environment | |
CN111860509A (en) | Coarse-to-fine two-stage non-constrained license plate region accurate extraction method | |
CN113256624A (en) | Continuous casting round billet defect detection method and device, electronic equipment and readable storage medium | |
Xing et al. | Traffic sign detection and recognition using color standardization and Zernike moments | |
CN108664969A (en) | Landmark identification method based on condition random field | |
CN104182769A (en) | Number plate detection method and system | |
CN113283405A (en) | Mask detection method and device, computer equipment and storage medium | |
CN113888754A (en) | Vehicle multi-attribute identification method based on radar vision fusion | |
CN107392115B (en) | Traffic sign identification method based on hierarchical feature extraction |
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 |