CN108182431A - A kind of traffic sign recognition method based on maximum stable extremal region and genetic optimization SVM - Google Patents

A kind of traffic sign recognition method based on maximum stable extremal region and genetic optimization SVM Download PDF

Info

Publication number
CN108182431A
CN108182431A CN201810212287.2A CN201810212287A CN108182431A CN 108182431 A CN108182431 A CN 108182431A CN 201810212287 A CN201810212287 A CN 201810212287A CN 108182431 A CN108182431 A CN 108182431A
Authority
CN
China
Prior art keywords
image
hog
value
region
traffic sign
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.)
Withdrawn
Application number
CN201810212287.2A
Other languages
Chinese (zh)
Inventor
高振国
钱坤
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.)
Dalian University of Technology
Original Assignee
Dalian University of Technology
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 Dalian University of Technology filed Critical Dalian University of Technology
Priority to CN201810212287.2A priority Critical patent/CN108182431A/en
Publication of CN108182431A publication Critical patent/CN108182431A/en
Withdrawn legal-status Critical Current

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
    • G06V20/582—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
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/09—Recognition of logos

Landscapes

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

Abstract

The present invention provides a kind of traffic sign recognition methods based on maximum stable extremal region and genetic optimization SVM, belong to technical field of image processing.With method of the piecemeal HOG feature vectors as edges of regions to be identified detection and image segmentation, it can inhibit to translate to a certain extent and rotate the influence brought, and reduce the interference that image is brought by intensity of illumination variation.Meanwhile piecemeal HOG is greatly reduced compared to tradition HOG dimensions, promotes operation efficiency.The improvement genetic optimization optimized parameter searching algorithm based on self-adaptive cross operation is used in the Classification and Identification stage, calculate optimal SVM classifier parameter, the fallibility and a large amount of of machine training for avoiding handmarking take, the advantage of comprehensive each method, the requirement of accuracy and real-time is preferably balanced, realizes the automatic detection and identification of traffic sign.The invention identifies the test pictures in German traffic mark examination criteria database, has obtained better effects.

Description

A kind of Traffic Sign Recognition based on maximum stable extremal region and genetic optimization SVM Method
Technical field
The invention belongs to technical field of image processing, apply in intelligent transportation scene.With based on MSER maximum stables The traffic sign extraction algorithm of extremal region is originated from the RGB image containing traffic sign under natural scene to divide extraction.It Afterwards, the feature of traffic sign is extracted with piecemeal HOG gradient orientation histograms.Using based on self-adaptive cross operation Genetic optimization optimized parameter searching algorithm is improved, punishment parameter g optimal in SVM classifier and kernel functional parameter c is found out and carrys out structure The optimal SVM classifier for this problem is made, to identify the traffic sign in image.
Background technology
Image processing techniques is component part important in intelligent transportation field, and the automation traffic sign of efficiently and accurately is known Traffic behavior is participated in while traffic participant specification can not guided, mitigates driver information processing pressure, so as to reduce accident hair Raw probability.At present, traffic mark identification (Traffic Sign Recognition, TSR) system is mainly by being mounted on vehicle The traffic mark information on camera acquisition road on is transmitted to image processing module and carries out Mark Detection and identification, it System will make different counter-measures according to the result of identification afterwards.Detection-phase needs the color and shape according to traffic mark Feature finds and positions the area-of-interest for including traffic mark from the image of acquisition.Cognitive phase will to area-of-interest, Feature is extracted, and classified to these area-of-interests with suitable sorting algorithm with different methods, obtain traffic mark Type information.
With going deep into for research, the theoretical method of many automatic recognition of traffic signs is emerged.Most of identification at present The method that theory is all divided as image using colouring information in detection-phase, still, colouring information by strong light, dim light, Effect can be deteriorated under the influence of situations such as haze, fog, mark self color degrade.In addition, although traffic mark usually has Eye-catching color, naked eyes are readily discernible, but under the City scenarios of modernization (such as colorful light, wall color etc.), make It obtains traffic mark to be difficult to differentiate between with background color, so as to be not easy to find identified areas during Mark Detection.In cognitive phase, Most of existing system is made of the grader that the true picture of manual markings is used to train, this is that a repetition takes, and And the processing procedure easily to malfunction.For this purpose, to evade this manual operation and manual markings training as possible.Simultaneously as traffic Sign recognition system is higher to the requirement of real-time of detection identification, especially in the case where condition of road surface complexity, speed are higher, is System needs to make identification in the shortest possible time, and by result notice driver, so that the time that driver has abundance does Go out reaction.Therefore, in view of color identification is complex environment apply the drawbacks of, the fallibility and machine of manual markings are trained a large amount of Requirement of the time-consuming and system to real-time, effect more stable algorithm more balanced without a kind of performance are handed over to realize The intelligent recognition of logical mark.
Invention content
Currently invention addresses several important performance indexes in Traffic Sign Recognition, it is proposed that one kind is based on maximum stable pole It is worth region and the traffic sign recognition method of genetic optimization SVM, it is therefore intended that optimize the performance of existing method, overcome prior art Deficiency.The methods of mainly using maximum stable extremal region, extracts the picture containing traffic sign segmentation, will be initial Common RGB photos, pre-process into the bianry image conducive to recognition detection；Later feature extraction is carried out using piecemeal HOG；Finally Using the improvement genetic optimization optimized parameter searching algorithm based on self-adaptive cross operation, calculate optimal in SVM classifier Parameter is classified and is identified to the traffic sign in image with the grader after optimization.
Technical scheme of the present invention：
Traffic sign recognition method based on maximum stable extremal region and SVM, step are as follows：
Step 1. color is converted：Interest region is detected, and right using maximum stable extremal region (MSER) algorithm Image carries out gray processing processing.
Step 2. edge detection：The horizontal and vertical First-order Gradient of each pixel in calculating interest region, and as HOG feature vectors carry out edge detection to interest region.And dimension expense is reduced with piecemeal HOG, improve operation efficiency.
Step 3. Classification and Identification：It is calculated with being searched for using the improvement genetic optimization optimized parameter based on self-adaptive cross operation Method finds out punishment parameter g optimal in SVM classifier and kernel functional parameter c to construct the optimal svm classifier for this problem Device.And the shape in interest region is classified and identified using SVM classifier.It is defeated and according to the standard picture in database Go out identified traffic sign.
Beneficial effects of the present invention：Friendship based on maximum stable extremal region and genetic optimization SVM proposed by the invention Logical sign, the prominent stability and robustness for having played maximum stable extremal region algorithm, and it is special with piecemeal HOG The method that sign vector is divided as interest region edge detection and image, this method can inhibit to translate to a certain extent and revolve Turn the influence brought, since its variation for illumination is insensitive, can also reduce what image was brought by intensity of illumination variation Interference.Meanwhile piecemeal HOG is greatly reduced compared to tradition HOG dimensions, computing overhead reduces, improved efficiency.In the Classification and Identification stage Using genetic optimization SVM classifier, the fallibility and a large amount of of machine training for avoiding handmarking take, comprehensive each method Advantage preferably balances the requirement of accuracy and real-time, realizes the automatic detection and identification of traffic sign.
Description of the drawings
Fig. 1 is the flow chart of traffic sign recognition method of the present invention
Fig. 2 is the design sketch of the example recognition flow of the present invention
Specific embodiment
Below in conjunction with attached drawing and technical solution, the specific embodiment further illustrated the present invention.
For above-mentioned 3 steps, detailed description below is carried out to each step：
Step 1：Color is converted
1.1 image normalization：Since traffic sign is mostly red and blue, first formula
Red and blue portion is counted, larger one is chosen as threshold value, red blue normalization is carried out to picture Processing.
1.2 determining MSER regions
Multiple MESR regions can be generated after normalized, with principle exclusive PCR region once.
1) area region bigger than normal and less than normal needs to remove, because area region bigger than normal is particularly likely that roadside shape rule Object then, area region less than normal may be noise, and the ratio that interest region accounts for entire image is between 0.2% to 4% Relatively reasonable.
2) simultaneously, the excessive or too small region of Aspect Ratio is purged, determines traffic sign region for pros Shape, Aspect Ratio range here is between 0.6 to 1.4；
3) judge the interest region area and a upper change rate for being judged as stability region area in the nested region, If change rate is less than a certain smaller threshold value, there may be repeat regions, need to remove, and according to experiment, threshold value is set herein It is 2；
4) calculate the centre coordinate of each MSER, and as center calculation in a smaller range with the presence or absence of multiple MSER regions if it is present explanation has multiple MSER regions to be overlapped, then remove extra MSER regions；
Step 2：Edge detection
Feature extraction based on HOG divides the image into multiple small connected domains first, then calculates these small companies The gradient information of each pixel in logical domain.
2.1 CELL cells divide
Multiple small connected domains are divided an image into, each connected domain is a CELL unit, wherein 4 CELL will Form a new BLOCK unit.
2.2 calculate the gradient information of each pixel of image.
The amplitude and angle value of each pixel gradient in image are calculated, calculation formula is：
What variable θ (x, y) was represented is the angle of gradient；What variable m (x, y) was represented is the amplitude of gradient；Variable H (x, y) What is represented is image transverse gradients；What variable V (x, y) represented is image longitudinal direction ladder breath.
The statistics of the direction gradient of 2.3 images
The directional information that can obtain gradient θ (x, y) according to 2.2, value range are -90 degree between 90 degree.Then It is averagely divided according to the value range of θ (x, y) by 9 deciles, corresponding power is finally calculated according to the bearing data of CELL gradients Value m (x, y).Therefore, by above-mentioned calculating process, the feature vector that a dimension is 9 can be obtained, and a BLOCK is by four A CELL is formed, and therefore, the feature vector dimension of each BLOCK is 36.
The processing of 2.4 feature normalizations
With the presence of the various interference of Traffic Sign Images collected, it is therefore desirable to be standardized place to gradient information Reason.CELL is put into corresponding BLOCK, then each BLOCK is standardized, formula is：
Here, the vector before v is standardization, ε is the normalization constants of a very little, is 0 to prevent divisor.
2.5 HOG integrograms
According to aforementioned four step, by any point (x, y) in image, H (x, y)=[H (x, y) is used1,…,H(x,y )9]TTo represent its corresponding 9 direction gradient information.Therefore its HOG characteristic value can be obtained by formula below：
According to formula (5) it is found that all pixels point to image is handled, the HOG integrograms of complete image are obtained. It obtains HOG integrograms and then HOG integrations is obtained by following calculation formula.
So for CELL arbitrary in image, the HOG on its corresponding four vertex integrations are calculated by formula (6) Value, i.e. A1,A2,A3,A4, then its corresponding HOG vector can be expressed as：
HOG=A4+A1-A2-A3 (7)
By the calculating steps of above-mentioned HOG vectors it is found that the extraction of HOG feature vectors is mainly based upon topography obtains , therefore its specific stronger antijamming capability.
2.6 feature extractions based on piecemeal HOG
By taking the Traffic Sign Images of 40 × 40 sizes as an example, will wherein 8 × 8 sizes image configuration one CELL, every four A CELL forms a BLOCK, since each CELL has 9 dimensional feature vectors, so in each BLOCK containing 36 dimensional features to Amount, then be scanned by 8 pixels of each step-length, at this time the horizontal direction of image and vertically to that will have 4 scanning windows, always Totally 16 blocks with overlapping.And a BLOCK has 36 dimensional feature vectors, then the Traffic Sign Images of 40 × 40 sizes will With 576 dimensional feature vectors.Feature extraction carries out whole image by HOG features, the dimension of obtained feature vector compared with Greatly, the operational efficiency of recognizer can be influenced.Thus with piecemeal HOG, to reduce the dimension of feature vector.
Usual interest region area usually will not be very big, so 16 subgraphs of 4 rows 4 row are classified as, then to 16 sons Figure carries out HOG feature extractions respectively.In this way, it is possible to reduce a large amount of feature vector generated due to overlapping, from And reduce final HOG feature vector dimensions.
Step 3. Classification and Identification
Since Libsvm tool boxes can carry out quick multi-objective predictive and classification feature, it is more suitable for the solution of problems Certainly, so we realize that SVM trains prediction model using Libsvm tool boxes.But the parameter of Libsvm tool boxes acquiescence has The space advanced optimized, because different punishment parameter g, the selection of kernel functional parameter c and kernel function has recognition effect Apparent performance influences.The optimal SVM parameters for such issues that therefore need to calculate adaptation.
Here it is optimal to calculate with the improvement genetic optimization optimized parameter searching algorithm based on self-adaptive cross operation Punishment parameter g and kernel functional parameter c, construct optimal SVM classifier.
The adaptive adjustment and the adaptive adjustment of mutation probability of 3.1 crossover probabilities
Crossover probability P in genetic algorithmCWith mutation probability PmThe setting of value can largely influence genetic algorithm The degree of approach of convergence and optimal solution and true optimal solution.Under normal conditions, crossover probability PCValue it is bigger, then new individual produce Raw speed will be faster, and crossover probability PCValue cross conference and make the high individual configurations that adapt to by rapid damage.For mutation probability Pm, If PmValue it is too small, be just not easy generate new individual；If PmValue it is excessive, algorithm is equivalent to random search algorithm.Here The diversity of group is ensured with the crossover probability and mutation probability that can adaptively adjust：
Wherein：Variable λ1,λ2,λ3,λ4Meaning represent 0~1 constant, variable fmaxMeaning represent adaptive value maximum Value, variableMeaning represent the average value of adaptive value, the meaning of variable f ' represents larger cross-adaptation value, and variable f's contains Justice represents the adaptive value of variation individual.
The improvement of 3.2 crossover operators
Crossover operator can guarantee that the part Optimality in individual can smoothly go down, therefore method used herein is in heredity When chromosome binary coding, two crosspoints are randomly choosed, then the chromosome between crosspoint is swapped. In real coding, using arithmetic crossover operator, arithmetic crossover operator can be expressed as：
WhereinIt is the individual after intersecting,It is prechiasmal individual,It is 0~1 random number.
The improvement of 3.3 mutation operators
Mutation operation can keep the diversity of population, extremely important to the solution of optimal solution.Herein in binary coding Basic bit mutation operation has been used, and non-uniform mutation has been used in real coding.The new gene used in non-uniform mutation Value x 'kFor：
WhereinIt isIn the range of random number；It isIn the range of random number.
The foundation of 3.4 genetic optimization fitness functions
According to the principle of above-mentioned improved adaptive GA-IAGA, following optimization object function is established：
Fobj=funcSVM(g,c,P) (12)
Wherein g is punishment parameter, and c is kernel functional parameter, and P is test sample, FobjFor corresponding discrimination.The two ginsengs Number carries out optimum search by improved adaptive GA-IAGA.Its corresponding fitness function can be then expressed as：
By improving genetic optimization, it is 0.0068 that can obtain parameter g optimal values, and parameter c optimal values are 4.9951.This When, the parameter of SVM support vector machines is done optimal value setting can obtain optimal SVM.

Claims (1)

  1. A kind of 1. traffic sign recognition method based on maximum stable extremal region and genetic optimization SVM, which is characterized in that step It is as follows：
    Step 1：Color is converted
    Step 1a：Determine image red blue normalized threshold, calculation formula is：
    Wherein, R is RED sector, and B is blue portion, and R+G+B is whole for image；ΩRBIt is the normalized threshold value of red blue；
    Step 1b：Using the threshold value calculated, image is normalized, the interest region of prominent identification to be detected；It is right The interest region of prominent identification to be detected, final interest region, as traffic sign region are filtered out with following principle；
    1) area region bigger than normal and less than normal is removed, because area region bigger than normal is particularly likely that the object of roadside regular shape Body, area region less than normal may be noise, and interest region accounts for the ratio of entire image between 0.2% to 4%；
    2) simultaneously, the excessive or too small region of Aspect Ratio is purged, determines traffic sign region for square, length and width Proportional region is between 0.6 to 1.4；
    3) judge the interest region area and a upper change rate for being judged as stability region area in the nested region, if Change rate is less than a certain smaller threshold value, then there may be repeat regions, remove, and threshold value is set as 2；
    4) centre coordinate of each MSER is calculated, and whether there is multiple MSER in a smaller range as center calculation Region if it is present explanation has multiple MSER regions to be overlapped, then removes extra MSER regions；
    Step 2：Edge detection
    2a：CELL cells divide
    Multiple small connected domains are divided an image into, each connected domain is a CELL unit, wherein 4 CELL will be formed One new BLOCK unit；
    2b：Calculate the gradient information of each pixel of image
    The amplitude and angle value of each pixel gradient in image are calculated, calculation formula is：
    Wherein, what variable θ (x, y) was represented is the angle of gradient；What variable m (x, y) was represented is the amplitude of gradient；Variable H (x, y) What is represented is image transverse gradients；What variable V (x, y) represented is image longitudinal direction ladder breath；
    2c：The statistics of the direction gradient of image
    The directional information of gradient θ (x, y) is obtained according to step 2b, value range is -90 degree between 90 degree；Then according to θ (x,y）Value range it is averagely divided to 9 deciles, finally according to the bearing data of CELL gradients calculate corresponding weights m (x, y）；The feature vector that a dimension is 9 is calculated by above-mentioned, and a BLOCK is made of four CELL, therefore, each The feature vector dimension of BLOCK is 36；
    2d：Feature normalization processing
    The Traffic Sign Images collected are standardized gradient information there are various interference；CELL is put into Into corresponding BLOCK, then each BLOCK is standardized, formula is：
    Wherein, the vector before v is standardization, ε is the normalization constants of a very little, is 0 to prevent divisor；
    2e：HOG integrograms
    According to aforementioned four step, by any point (x, y) in image, H (x, y are used）=[H (x, y)1,…,H(x,y)9]TCome Represent its corresponding 9 direction gradient information；Its HOG characteristic value is obtained by formula below：
    According to formula above it is found that all pixels point to image is handled, the HOG integrograms of complete image are obtained；It is obtaining It obtains HOG integrograms and then passes through following calculation formula and obtain HOG integrations；
    So for CELL arbitrary in image, the HOG integrated values on its corresponding four vertex, i.e. A are calculated by above formula1, A2,A3,A4, then its corresponding HOG vector is expressed as：
    HOG=A4+A1-A2-A3
    2f：Feature extraction based on piecemeal HOG
    Interest region is divided into 16 subgraphs that 4 rows 4 arrange, HOG feature extractions are then carried out respectively to 16 subgraphs；Using this side Formula, a large amount of feature vector for reducing due to overlapping and generating, so as to reduce final HOG feature vector dimensions；
    Step 3：Classification and Identification
    3a：The adaptive adjustment and the adaptive adjustment of mutation probability of crossover probability
    Crossover probability P in genetic algorithmCWith mutation probability PmValue setting largely influence genetic algorithm convergence With optimal solution and the degree of approach of true optimal solution；Under normal conditions, crossover probability PCValue it is bigger, then new individual generate speed To be faster, and crossover probability PCValue cross ambassador's height adapt to individual configurations by rapid damage；For mutation probability PmIf Pm's Be worth it is too small, be just not easy generate new individual；If PmValue it is excessive, algorithm is equivalent to random search algorithm；It is adjusted with adaptive Whole crossover probability and mutation probability ensure the diversity of group：
    Wherein：Variable λ1,λ2,λ3,λ4Meaning represent 0~1 constant, variable fmaxMeaning represent adaptive value maximum value, become AmountMeaning represent the average value of adaptive value, the meaning of variable f ' represents larger cross-adaptation value, and the meaning of variable f represents Make a variation individual adaptive value；
    3b：The improvement of crossover operator
    Crossover operator can guarantee that the smooth heredity of part Optimality in individual is gone down, and the method used is compiled in chromosome binary system When code, two crosspoints are randomly choosed, then the chromosome between crosspoint is swapped；In real coding, make With arithmetic crossover operator, arithmetic crossover operator representation is：
    WhereinIt is the individual after intersecting,It is prechiasmal individual,It is 0~1 random number；
    3c：The improvement of mutation operator
    It is operated in binary coding using basic bit mutation, and non-uniform mutation is used in real coding；Non-uniform mutation The middle new gene value x ' usedkFor：
    WhereinIt isIn the range of random number；It is In the range of random number；
    3d：The foundation of genetic optimization fitness function
    According to the principle of above-mentioned improved adaptive GA-IAGA, following optimization object function is established：
    Fobj=funcSVM(g,c,P)
    Wherein g is punishment parameter, and c is kernel functional parameter, and P is test sample, FobjFor corresponding discrimination；The two parameters are led to It crosses improved adaptive GA-IAGA and carries out optimum search, corresponding fitness function is then expressed as：
    By improving genetic optimization, it is 0.0068 to obtain parameter g optimal values, and parameter c optimal values are 4.9951.
CN201810212287.2A 2018-03-15 2018-03-15 A kind of traffic sign recognition method based on maximum stable extremal region and genetic optimization SVM Withdrawn CN108182431A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810212287.2A CN108182431A (en) 2018-03-15 2018-03-15 A kind of traffic sign recognition method based on maximum stable extremal region and genetic optimization SVM

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810212287.2A CN108182431A (en) 2018-03-15 2018-03-15 A kind of traffic sign recognition method based on maximum stable extremal region and genetic optimization SVM

Publications (1)

Publication Number Publication Date
CN108182431A true CN108182431A (en) 2018-06-19

Family

ID=62553616

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810212287.2A Withdrawn CN108182431A (en) 2018-03-15 2018-03-15 A kind of traffic sign recognition method based on maximum stable extremal region and genetic optimization SVM

Country Status (1)

Country Link
CN (1) CN108182431A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108896053A (en) * 2018-07-12 2018-11-27 北京理工大学 A kind of planetary landing optical guidance optimal landmark choosing method
CN109063619A (en) * 2018-07-25 2018-12-21 东北大学 A kind of traffic lights detection method and system based on adaptive background suppression filter and combinations of directions histogram of gradients
CN109446984A (en) * 2018-10-26 2019-03-08 重庆大学 Traffic sign recognition method in natural scene
CN110175513A (en) * 2019-04-15 2019-08-27 浙江工业大学 A kind of guideboard identification attack defense method based on the optimization of multiple target road

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106295686A (en) * 2016-08-01 2017-01-04 安徽农业大学 The tea aroma sorting technique of parameter optimization support vector machine
CN107330365A (en) * 2017-05-27 2017-11-07 深圳市美好幸福生活安全系统有限公司 Traffic sign recognition method based on maximum stable extremal region and SVM
CN107609485A (en) * 2017-08-16 2018-01-19 中国科学院自动化研究所 The recognition methods of traffic sign, storage medium, processing equipment

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106295686A (en) * 2016-08-01 2017-01-04 安徽农业大学 The tea aroma sorting technique of parameter optimization support vector machine
CN107330365A (en) * 2017-05-27 2017-11-07 深圳市美好幸福生活安全系统有限公司 Traffic sign recognition method based on maximum stable extremal region and SVM
CN107609485A (en) * 2017-08-16 2018-01-19 中国科学院自动化研究所 The recognition methods of traffic sign, storage medium, processing equipment

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
刘成云: "行车环境下多特征融合的交通标识检测与识别研究", 《中国博士学位论文全文数据库 信息科技辑》 *
叶阳阳: "交通标志检测和识别算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
王斌 等: "基于MSER和SVM的快速交通标志检测", 《光电子.激光》 *
王斌: "自然环境下道路交通标志的检测与识别", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 *
韩瑞锋: "《遗传算法原理与应用实例》", 31 January 2010, 兵器工业出版社 *
龚文引 等: "《智能算法在高光谱遥感数据处理中的应用》", 30 November 2014, 中国地质大学出版社 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108896053A (en) * 2018-07-12 2018-11-27 北京理工大学 A kind of planetary landing optical guidance optimal landmark choosing method
CN108896053B (en) * 2018-07-12 2021-11-23 北京理工大学 Planet landing optical navigation optimal road sign selection method
CN109063619A (en) * 2018-07-25 2018-12-21 东北大学 A kind of traffic lights detection method and system based on adaptive background suppression filter and combinations of directions histogram of gradients
CN109446984A (en) * 2018-10-26 2019-03-08 重庆大学 Traffic sign recognition method in natural scene
CN110175513A (en) * 2019-04-15 2019-08-27 浙江工业大学 A kind of guideboard identification attack defense method based on the optimization of multiple target road

Similar Documents

Publication Publication Date Title
CN105930791B (en) The pavement marking recognition methods of multi-cam fusion based on DS evidence theory
CN108182431A (en) A kind of traffic sign recognition method based on maximum stable extremal region and genetic optimization SVM
CN103605953B (en) Vehicle interest target detection method based on sliding window search
Chen et al. An improved license plate location method based on edge detection
CN109284669A (en) Pedestrian detection method based on Mask RCNN
CN110349260B (en) Automatic pavement marking extraction method and device
CN102968646B (en) A kind of detection method of license plate based on machine learning
CN109101924A (en) A kind of pavement marking recognition methods based on machine learning
CN103984953A (en) Cityscape image semantic segmentation method based on multi-feature fusion and Boosting decision forest
CN109086687A (en) The traffic sign recognition method of HOG-MBLBP fusion feature based on PCA dimensionality reduction
CN107729812B (en) Method suitable for recognizing vehicle color in monitoring scene
CN107330365A (en) Traffic sign recognition method based on maximum stable extremal region and SVM
CN105005989B (en) A kind of vehicle target dividing method under weak contrast
Le et al. Real time traffic sign detection using color and shape-based features
CN105913041A (en) Pre-marked signal lights based identification method
CN106529532A (en) License plate identification system based on integral feature channels and gray projection
CN111127499A (en) Security inspection image cutter detection segmentation method based on semantic contour information
CN105868734A (en) Power transmission line large-scale construction vehicle recognition method based on BOW image representation model
CN104408424A (en) Multiple signal lamp recognition method based on image processing
CN107563301A (en) Red signal detection method based on image processing techniques
CN102393902A (en) Vehicle color detection method based on H_S two-dimensional histogram and regional color matching
CN104573707A (en) Vehicle license plate Chinese character recognition method based on multi-feature fusion
CN107315990A (en) A kind of pedestrian detection algorithm based on XCS LBP features and cascade AKSVM
CN111915583A (en) Vehicle and pedestrian detection method based on vehicle-mounted thermal infrared imager in complex scene
CN108734170B (en) License plate character segmentation method based on machine learning and template

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
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20180619