CN109448000A - A kind of dividing method of road sign image - Google Patents
A kind of dividing method of road sign image Download PDFInfo
- Publication number
- CN109448000A CN109448000A CN201811177533.1A CN201811177533A CN109448000A CN 109448000 A CN109448000 A CN 109448000A CN 201811177533 A CN201811177533 A CN 201811177533A CN 109448000 A CN109448000 A CN 109448000A
- Authority
- CN
- China
- Prior art keywords
- image
- region
- value
- fingerpost
- mser
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- 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
- G06T2207/10004—Still image; Photographic image
Abstract
The present invention relates to the technical field of image processing in machine vision.A kind of dividing method of road sign image, Step 1: feature extraction;Step 2: extracting fingerpost image geometry feature and character pitch from the region MSER, geometrical characteristic then is extracted on the basis of MSER characteristic value, incongruent feature is deleted;Step 3: extracting the Blob feature of fingerpost image, the progressive scanning picture since bianry image first trip counts the non-zero cluster in fingerpost bianry image.The present invention is split the fingerpost in road especially in urban road, especially assist driving is even unmanned to provide road traffic sign detection and identification lays the foundation for the development of intelligent transportation.
Description
Technical field
The present invention relates to the technical field of image processing in machine vision.
Background technique
Traffic sign in specification traffic behavior, instruction condition of road surface, ensures road as important road safety affiliated facility
Road effect, guidance pedestrian and safe driving etc. play an important role.Automatic recognition of traffic signs is the important of intelligent transportation
Component part, the traffic sign on road send recognition result information to driving by preparatory detection, automatic identification in time
Member, can make to drive becomes more safe and light, plays the positive effect of safe driving, greatly reduces accident rate.Road
In traffic sign be chronically exposed to open air, due to adverse weather conditions or artificial destruction, the color of sign face can occur
It degenerates, be stained;Blocking by other objects (such as building, branch, billboard) when being flagged with by urban road, causes to hand over
Logical mark is imperfect;Under outdoor environment, illumination condition be variation and it is uncontrollable, illumination can be with time, season, weather
Change and change, so that collected Traffic Sign Images also generate corresponding change;The movement of vehicle can make the image taken
The reasons such as blur degradation are generated, the detection of traffic sign and identification can be made to bring many difficulties.
Traffic sign is broadly divided into caution sign, prohibitory sign, Warning Mark, fingerpost, tourism distinctive emblem and road
Construction safety mark six class, each classification have its specific color and shape.Therefore currently, the related algorithm of traffic sign is with face
Color and shape, which are characterized, to be detected and is classified, and research object is with caution sign, prohibitory sign, three type of Warning Mark
Based on type.And fingerpost transmits road direction, place, range information, fingerpost shape is mostly rectangle, and content is
Text and character.As the road construction in city is more and more, the fingerpost in urban road increases severely, each crossroad
The pavement of road such as mouth, crossroad, ring road crossing, the crossing Jie Xiang are equipped with fingerpost.The information of fingerpost is very rich,
Comprising road name, geographic direction, location information, target range etc., these Informational Expressions are Chinese character, letter, number and direction
Symbol, detection and identification difficulty are larger.Along with city vehicle is accelerated even more huge, fingerpost is for vehicle drive
The safety of personnel drives, comfortable driving, easily driving has a decisive role.
The basic thought of MSER (most stable extremal region) detection method is: in input gray level image, finding internal picture
Plain gray scale all greater than or less than its surrounding pixel gray value local image region, and define interior pixels gray scale and week
Enclosing the difference between pixel grey scale, directly proportional to the stability of regional area (stability is higher to be detected repeatedly in different images
Probability it is higher).MSER has invariance for the affine transformation of image, when threshold value changes within the scope of some, MSER's
Area does not change with the variation of threshold value, does not need to carry out smooth pretreatment to image, it is possible to detect under a variety of scales
MSER both can detecte small size MSER, may also detect that the features such as large scale MSER.Information for fingerpost is
Steady component in collected road image, various characters and symbol are wrongly written or mispronounced character in this partial information, have stable headway.Again
In conjunction with the rectangular characteristic of the quadrangle of fingerpost, using the geometrical characteristic and character pitch of MSER, fingerpost, to fingerpost
Target information in will carries out coarse segmentation.
Blob in computer vision, which refers in image, has one piece of company composed by Similar color, Texture eigenvalue
Logical region.Blob, which analyzes (Blob Analysis), to be analyzed the connected domain of same pixel in image.Its process is in fact
It is that image is carried out to binaryzation, segmentation obtains foreground and background, connected region detection is then carried out, to obtain the mistake of Blob block
Journey.Blob analysis tool can isolate target from background, and can calculate the quantity of target, position, shape, direction and
Size can also provide the topological structure between related spot.Single pixel is not analyzed one by one during processing, it is right
The row of image is operated.Every a line of image all indicates adjacent target zone with run length coding, RLC (RLE).This calculation
Method substantially increases the speed of processing compared with algorithm pixel-based.It is thick by being carried out to collected fingerpost image
After segmentation, after the removal of non-targeted information, only it is left individual fingerpost.Next it can use Blob analysis, to showing the way
Each information of mark extracts segmentation.It prepares for all kinds of characters and Symbol recognition in next step.
Summary of the invention
The technical problems to be solved by the present invention are: how to solve the problems, such as to detect and identify that difficulty is big in way-finding sign.
The technical scheme adopted by the invention is that: a kind of dividing method of road sign image, according to following step
It is rapid to carry out
Step 1: feature extraction, using the fingerpost image in digital camera acquisition real road, by collected finger
Road sign will image graphic gray processing forms gray scale picture I, using MSER feature extraction, after extracting MSER characteristic value, by MSER
Area marking comes out;
Step 2: fingerpost image geometry feature and character pitch are extracted from the region MSER, then in MSER feature
Geometrical characteristic is extracted on the basis of value, incongruent feature is deleted, and fingerpost image geometry feature includes Euler's numbers, minimum square
Shape and its area, area pixel point number, zone boundary length, eccentricity, secondly, character pitch is converted using Euclidean distance,
Every gives the Euclidean distance of subset to the plane in Calculation Plane;
Step 3: the Blob feature of fingerpost image is extracted, and the progressive scanning picture since bianry image first trip, statistics
Non-zero cluster in fingerpost bianry image;It calculates in every row, the starting column and termination column address of non-zero cluster, and row ground
Location;Mark non-zero cluster Label value are as follows: 1,2,3 ..., and record it is of equal value right, it is of equal value to referring to for marking same target
All Label values of Blob replace all values of centering of equal value with the smallest Label value, according to continuous and unique Label
It is worth, the target number in statistical picture;Label calculates the area and center-of-mass coordinate of each Blob, the region Blob after completing
Value range is [5 400], gives up and falls to be unsatisfactory for the Blob of threshold value, the Blob for being unsatisfactory for threshold value is the lesser Blob of some areas
It is as caused by noise, last record storage meets the area, mass center, ROI region translation specifications information of the Blob of threshold value.
As a kind of preferred embodiment: the detailed step of MSER feature extraction in step 1 are as follows:
One, gray scale picture I is regarded as a kind of mapping, I:Region in D representative image, Z represent square
Battle array, S represent the pixel point value of gray level image, and S is total order, and S={ 0,1 ..., 255 }, neighborhood relationships meet,Using 8 neighborhoods, i.e., ifWhen, then p, q ∈ D be exactly it is adjacent, be expressed as
PAq, A are image region, and p, q are the coordinate value of arbitrary point pixel in A, and i is positive integer, pi,qiFor p in A, in q value range
The coordinate value at any point, the value range of d p, q are positive integer;
Two, region Q is a continuation subset of region D, for any p, q ∈ Q, all there is communication path a p, a1,
a2,…,an..., q, so that pAa1,…,aiAai+1,…,anAq,a1,a2,…anAll continuous positive integers between p and q;
Three, the boundary of region Q With at least one pixel phase in Q
Neighbour, butIt is not belonging to region Q i.e.
Four, extremal region Q ' is denoted asFor all p ∈ Q,Meet in maximum gray areas
Imax(p) > Imax(q) or in minimal gray region Imin(p) < Imin(q) region, Imax(p)、Imax(q) refer to p, q in I
Max pixel value, Imin(p)、Imin(q) refer to the minimum pixel value of p, q in I;
Five, most stable extremal region MSER: Q ' is set1..., Q 'i-1, Q 'iFor nested extremal region One
A sequence.If | Q 'i+Δ\Q′i-Δ|/Qi' in i*There are local minimums at place, then extremal region Q 'i*It is exactly MSER, i*It is embedding
Cover a certain layer of extremal region;| | for the gesture of set;Δ ∈ S is parameter, indicates small grey scale change.
As a kind of preferred embodiment: in step 2, extracting fingerpost image geometry and be characterized in using in MATLAB platform
Regionprops function extract, character pitch is converted using Euclidean distance, in Calculation Plane every it is given to the plane
The detailed process of the Euclidean distance of subset are as follows: enable I':For a bianry image, Ω={ 0 ..., 1 } ×
{ 0 ..., 1 }, I' are the bianry image of image in the region MSER, and Z ' is matrix, and 0 is associated as stain, and 1 is associated with white point, preceding
The set Ο of scape is made of all white points, meets Ο={ p ∈ Ω } | and the supplementary set of I ' (p)=1, set Ο is made of all stains
As background, I ' (p) refer to the pixel value of p in I'.The beneficial effects of the present invention are: the present invention is to especially city road in road
Fingerpost in road is split, and especially assists driving is even unmanned to provide traffic mark for the development of intelligent transportation
Will detection and identification lay the foundation.
Specific embodiment
A kind of dividing method of road sign image is carried out according to following step
Step 1: feature extraction, using the fingerpost image in digital camera acquisition real road, by collected finger
Road sign will image graphic gray processing forms gray scale picture I, using MSER feature extraction, after extracting MSER characteristic value, by MSER
Area marking comes out;
One, gray scale picture I is regarded as a kind of mapping, I:Region in D representative image, Z represent square
Battle array, S represent the pixel point value of gray level image, and S is total order, and S={ 0,1 ..., 255 }, neighborhood relationships meet,Using 8 neighborhoods, i.e., ifWhen, then p, q ∈ D be exactly it is adjacent, be expressed as
PAq, A are image region, and p, q are the coordinate value of arbitrary point pixel in A, and i is positive integer, pi,qiFor p in A, in q value range
The coordinate value at any point, the value range of d p, q are positive integer;
Two, region Q is a continuation subset of region D, for any p, q ∈ Q, all there is communication path a p, a1,
a2,…,an..., q, so that pAa1 ..., aiAai+1,…,anAq,a1,a2,…anAll continuous positive integers between p and q;
Three, the boundary of region Q With at least one pixel phase in Q
Neighbour, butIt is not belonging to region Q i.e.
Four, extremal region Q ' is denoted asFor all p ∈ Q,Meet in maximum gray areas
Imax(p) > Imax(q) or in minimal gray region Imin(p) < Imin(q) region, Imax(p)、Imax(q) refer to p, q in I
Max pixel value, Imin(p)、Imin(q) refer to the minimum pixel value of p, q in I;
Five, most stable extremal region MSER: Q ' is set1..., Q 'i-1, Q 'iFor nested extremal region One
A sequence.If | Q 'i+Δ\Q′i-Δ|/Qi' in i*There are local minimums at place, then extremal region Q 'i*It is exactly MSER, i*It is embedding
Cover a certain layer of extremal region;| | for the gesture of set;Δ ∈ S is parameter, indicates small grey scale change.
MSER algorithm has invariance, covariance, stability, area size's variability, for the fingerpost of natural environment
For will detection, this algorithm has extraordinary robustness.MSER by carrying out pixel gray level to mark gray level image first
Value sequence, generally uses bucket sort algorithm, specifically first by each number rough sort in sequence into different casees, in each case certainly
Dynamic sequence, finally strings together these casees again;Then pixels statistics are carried out.Pixel is according to ascending sort, the pixel after sequence
It is put into image, obtains the incremental list of gray scale, the list of join domain and area are tieed up using efficient merging-lookup algorithm
Shield;Then minimum interconversion rate is found in search.Use Q'iIndicate any connected region in the corresponding bianry image of binarization threshold
Domain, when binarization threshold is in transformation, connected region is also correspondingly converted, and is finally considered as with little change rate region
The region MSER.After extracting MSER characteristic value to gray scale fingerpost image, MSER area marking is come out.
Step 2: fingerpost image geometry feature and character pitch are extracted from the region MSER, then in MSER feature
Geometrical characteristic is extracted on the basis of value, incongruent feature is deleted, and fingerpost image geometry feature includes Euler's numbers, minimum square
Shape and its area, area pixel point number, zone boundary length, eccentricity, judge whether it is finger using these geometrical characteristics
Road sign will can detect the target area of fingerpost to the maximum extent.Secondly, character pitch is converted using Euclidean distance,
Every gives the Euclidean distance of subset to the plane in Calculation Plane;Fingerpost image geometry is extracted to be characterized in using MATLAB
Regionprops function in platform extracts.In view of text, letter, number and the direction symbol etc. in fingerpost
Information, these information are in fingerpost always in a metastable region, such as geographical location, road name, distance
Information, directional information etc., each character can have certain spacing in the every part of these target informations.Character pitch is using European
Range conversion, the detailed process of every Euclidean distance for giving subset to the plane in Calculation Plane are as follows: enable I':For a bianry image, Ω={ 0 ..., 1 } × { 0 ..., 1 }, I' is two of image in the region MSER
It is worth image, Z ' is matrix, and 0 is associated as stain, and 1 is associated with white point, and the set Ο of prospect is made of all white points, meets Ο
={ p ∈ Ω } | the supplementary set of I ' (p)=1, set Ο is made of all stains Ji Wei background, and I ' (p) refers to the pixel value of p in I'.
From the point of view of range conversion, background dot is referred to as point of interest or characteristic point.It, will be to fingerpost after being converted according to Euclidean distance
Target area is more accurate.Finally, the position [X Y Width Height] according to the rectangle callout box of fingerpost is corrected
So that the rectangle geometry frame of fingerpost more gears to actual circumstances.To this step, the target area coarse segmentation of fingerpost is completed.
Step 3: Blob parser can mutually separate the foreground and background of image, any prospect shape is then provided
Morphological parameters.It is not the single pixel of analysis in this treatment process, what it was analyzed is the row of block.This method with
Algorithm pixel-based is compared, and processing speed can be accelerated.It can be provided for machine vision applications the spot in image quantity,
Position, area, perimeter, shape and direction can also provide the topological structure between related spot.Blob feature extraction is according to even
The pixel for meeting 4 neighborhoods or 8 neighborhoods is labeled as identical value by logical domain concept, and the connected domain extracted is exactly a Blob.
8 fields are used herein.Extract the Blob feature of fingerpost image, the progressive scanning picture since bianry image first trip, statistics
Non-zero cluster in fingerpost bianry image;It calculates in every row, the starting column and termination column address of non-zero cluster, and row ground
Location;Mark non-zero cluster Label value are as follows: 1,2,3 ..., and record it is of equal value right, it is of equal value to referring to for marking same target
All Label values of Blob replace all values of centering of equal value with the smallest Label value, according to continuous and unique Label
It is worth, the target number in statistical picture;Label calculates the area and center-of-mass coordinate of each Blob, the region Blob after completing
Value range is [5 400], gives up and falls to be unsatisfactory for the Blob of threshold value, the Blob for being unsatisfactory for threshold value is the lesser Blob of some areas
It is as caused by noise, last record storage meets the area, mass center, ROI region translation specifications information of the Blob of threshold value.
Before extracting feature by Blob, mathematical morphology correction first is carried out to the bianry image for the image that shows the way, is completed
After Blob feature mark, the zonule ROI is split, and then text, number, letter or the direction in fingerpost are accorded with
Number information is split, and establishes good ground foundation for identification from now on.
Claims (3)
1. a kind of dividing method of road sign image, it is characterised in that: carried out according to following step
Step 1: feature extraction, using the fingerpost image in digital camera acquisition real road, by collected fingerpost
Will image graphic gray processing forms gray scale picture I, using MSER feature extraction, after extracting MSER characteristic value, by the region MSER
It marks out and;
Step 2: extract fingerpost image geometry feature and character pitch from the region MSER, the then mould in database
Plate comparison judges whether it is fingerpost, and fingerpost image geometry feature includes Euler's numbers, minimum rectangle and its area, region
Pixel number, zone boundary length, eccentricity, character pitch are converted using Euclidean distance, are put down to this for every in Calculation Plane
Face gives the Euclidean distance of subset;
Step 3: extracting the Blob feature of fingerpost image, the progressive scanning picture since bianry image first trip, statistics shows the way
Indicate the non-zero cluster in bianry image;It calculates in every row, the starting column and termination column address and row address of non-zero cluster;
Mark non-zero cluster Label value are as follows: 1,2,3 ..., and record it is of equal value right, it is of equal value to referring to for marking same target Blob's
All Label values replace all values of centering of equal value with the smallest Label value, according to continuous and unique Label value, system
Count the target number in image;Label calculates the area and center-of-mass coordinate of each Blob, the region Blob value after completing
Range be [5 400], give up and fall to be unsatisfactory for the Blob of threshold value, be unsatisfactory for threshold value Blob be the lesser Blob of some areas be by
Caused by noise, last record storage meets the area, mass center, ROI region translation specifications information of the Blob of threshold value.
2. a kind of dividing method of road sign image according to claim 1, it is characterised in that: in step 1
The detailed step of MSER feature extraction are as follows:
One, gray scale picture I is regarded as a kind of mapping, I:Region in D representative image, Z represent matrix, S
The pixel point value of gray level image is represented, S is total order, and S={ 0,1 ..., 255 }, neighborhood relationships meet,It adopts
With 8 neighborhoods, i.e., ifWhen, then p, q ∈ D be exactly it is adjacent, be expressed as pAq, A is image area
Domain, p, q are the coordinate value of arbitrary point pixel in A, and i is positive integer, pi,qiFor p in A, the coordinate at any point in q value range
Value, the value range of d p, q are positive integer;
Two, region Q is a continuation subset of region D, for any p, q ∈ Q, all there is communication path a p, a1,a2,…,
an..., q, so that pAa1,…,aiAai+1,…,anAq,a1,a2,…anAll continuous positive integers between p and q;
Three, the boundary of region Q It is adjacent at least one pixel in Q,
ButIt is not belonging to region Q i.e.
Four, extremal region Q ' is denoted asFor all p ∈ Q,Meet in maximum gray areas Imax(p)
> Imax(q) or in minimal gray region Imin(p)<Imin(q) region, Imax(p)、Imax(q) refer to the maximum picture of p, q in I
Element value, Imin(p)、Imin(q) refer to the minimum pixel value of p, q in I;
Five, most stable extremal region MSER: Q ' is set1..., Q 'i-1, Q 'iFor nested extremal region A sequence
Column.If | Q 'i5Δ\Q′i-Δ|/Qi' in i*There are local minimums at place, then extremal region Q 'i*It is exactly MSER, i*For nested pole
It is worth a certain layer in region;| | for the gesture of set;Δ ∈ S is parameter, indicates small grey scale change.
3. a kind of dividing method of road sign image according to claim 1, it is characterised in that: in step 2,
It extracts fingerpost image geometry to be characterized in extracting using the regionprops function in MATLAB platform, character pitch
It is converted using Euclidean distance, the detailed process of every Euclidean distance to the given subset of the plane in Calculation Plane are as follows: enable I':For a bianry image, Ω={ 0 ..., 1 } × { 0 ..., 1 }, I' is two of image in the region MSER
It is worth image, Z ' is matrix, and 0 is associated as stain, and 1 is associated with white point, and the set Ο of prospect is made of all white points, meets Ο
={ p ∈ Ω } | the supplementary set of I ' (p)=1, set Ο is made of all stains Ji Wei background, and I ' (p) refers to the pixel value of p in I'.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811177533.1A CN109448000B (en) | 2018-10-10 | 2018-10-10 | Segmentation method of traffic direction sign image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811177533.1A CN109448000B (en) | 2018-10-10 | 2018-10-10 | Segmentation method of traffic direction sign image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109448000A true CN109448000A (en) | 2019-03-08 |
CN109448000B CN109448000B (en) | 2021-07-30 |
Family
ID=65546224
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811177533.1A Active CN109448000B (en) | 2018-10-10 | 2018-10-10 | Segmentation method of traffic direction sign image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109448000B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110633758A (en) * | 2019-09-20 | 2019-12-31 | 四川长虹电器股份有限公司 | Method for detecting and locating cancer region aiming at small sample or sample unbalance |
CN110930450A (en) * | 2019-12-11 | 2020-03-27 | 清远职业技术学院 | Coal gangue positioning method based on image threshold segmentation and BLOB analysis method |
CN111783651A (en) * | 2020-06-30 | 2020-10-16 | 北京百度网讯科技有限公司 | Pavement element identification method and device, electronic equipment and storage medium |
CN111950397A (en) * | 2020-07-27 | 2020-11-17 | 腾讯科技(深圳)有限公司 | Text labeling method, device and equipment for image and storage medium |
CN112763496A (en) * | 2020-12-24 | 2021-05-07 | 苏州赛众自动化科技有限公司 | Mobile phone battery surface defect detection device and detection method thereof |
CN113327012A (en) * | 2021-04-28 | 2021-08-31 | 东南大学 | Urban public transport index calculation method based on RGB color space and Monte Carlo method |
CN113487688A (en) * | 2021-05-18 | 2021-10-08 | 浙江大华技术股份有限公司 | Road sign detection method and device and computer readable storage medium |
CN113505766A (en) * | 2021-09-09 | 2021-10-15 | 北京智源人工智能研究院 | Image target detection method and device, electronic equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090232358A1 (en) * | 2008-03-11 | 2009-09-17 | Geoffrey Mark Timothy Cross | Method and apparatus for processing an image |
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 |
CN107688806A (en) * | 2017-08-21 | 2018-02-13 | 西北工业大学 | A kind of free scene Method for text detection based on affine transformation |
CN108038481A (en) * | 2017-12-11 | 2018-05-15 | 江苏科技大学 | A kind of combination maximum extreme value stability region and the text positioning method of stroke width change |
-
2018
- 2018-10-10 CN CN201811177533.1A patent/CN109448000B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090232358A1 (en) * | 2008-03-11 | 2009-09-17 | Geoffrey Mark Timothy Cross | Method and apparatus for processing an image |
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 |
CN107688806A (en) * | 2017-08-21 | 2018-02-13 | 西北工业大学 | A kind of free scene Method for text detection based on affine transformation |
CN108038481A (en) * | 2017-12-11 | 2018-05-15 | 江苏科技大学 | A kind of combination maximum extreme value stability region and the text positioning method of stroke width change |
Non-Patent Citations (3)
Title |
---|
SUBRATA KUMAR KUNDU ET.AL: "Speed Limit Sign Recognition Using MSER and Artificial Neural Networks", 《2015 IEEE 18TH INTERNATIONAL CONFERENCE ON INTELLIGENT TRANSPORTATION SYSTEMS》 * |
王淑丹 等: "基于边缘增强型MSER特征的道路标志检测", 《计算机时代》 * |
赵慧: "基于Blob的运动目标检测与跟踪算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110633758A (en) * | 2019-09-20 | 2019-12-31 | 四川长虹电器股份有限公司 | Method for detecting and locating cancer region aiming at small sample or sample unbalance |
CN110930450A (en) * | 2019-12-11 | 2020-03-27 | 清远职业技术学院 | Coal gangue positioning method based on image threshold segmentation and BLOB analysis method |
CN111783651A (en) * | 2020-06-30 | 2020-10-16 | 北京百度网讯科技有限公司 | Pavement element identification method and device, electronic equipment and storage medium |
CN111783651B (en) * | 2020-06-30 | 2024-01-12 | 北京百度网讯科技有限公司 | Pavement element identification method, pavement element identification device, electronic equipment and storage medium |
CN111950397A (en) * | 2020-07-27 | 2020-11-17 | 腾讯科技(深圳)有限公司 | Text labeling method, device and equipment for image and storage medium |
CN111950397B (en) * | 2020-07-27 | 2021-10-22 | 腾讯科技(深圳)有限公司 | Text labeling method, device and equipment for image and storage medium |
CN112763496A (en) * | 2020-12-24 | 2021-05-07 | 苏州赛众自动化科技有限公司 | Mobile phone battery surface defect detection device and detection method thereof |
CN113327012A (en) * | 2021-04-28 | 2021-08-31 | 东南大学 | Urban public transport index calculation method based on RGB color space and Monte Carlo method |
CN113327012B (en) * | 2021-04-28 | 2022-03-11 | 东南大学 | Urban public transport index calculation method based on RGB color space and Monte Carlo method |
CN113487688A (en) * | 2021-05-18 | 2021-10-08 | 浙江大华技术股份有限公司 | Road sign detection method and device and computer readable storage medium |
CN113505766A (en) * | 2021-09-09 | 2021-10-15 | 北京智源人工智能研究院 | Image target detection method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN109448000B (en) | 2021-07-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109448000A (en) | A kind of dividing method of road sign image | |
Wu et al. | Rapid localization and extraction of street light poles in mobile LiDAR point clouds: A supervoxel-based approach | |
CN105373794B (en) | A kind of licence plate recognition method | |
Singh et al. | A study of image segmentation algorithms for different types of images | |
CN110378293B (en) | Method for producing high-precision map based on live-action three-dimensional model | |
CN102722707A (en) | License plate character segmentation method based on connected region and gap model | |
Sultani et al. | Automatic pavement object detection using superpixel segmentation combined with conditional random field | |
CN107516077A (en) | Traffic sign information extracting method based on laser point cloud and image data fusion | |
CN102364496A (en) | Method and system for identifying automobile license plates automatically based on image analysis | |
CN107392141A (en) | A kind of airport extracting method based on conspicuousness detection and LSD straight-line detections | |
CN101154271A (en) | License plate character segmentation method based on fast area labeling algorithm and license plate large-spacing locating method | |
CN108875803A (en) | A kind of detection of harmful influence haulage vehicle and recognition methods based on video image | |
Li et al. | Inverse perspective mapping based urban road markings detection | |
CN105678315A (en) | Method of generating training image for automated vehicle object recognition system | |
CN103336961A (en) | Interactive natural scene text detection method | |
CN112434706B (en) | High-precision traffic element target extraction method based on image point cloud fusion | |
EP3770799A1 (en) | A method of identifying topographic features | |
Saha et al. | i LPR: An indian license plate recognition system | |
CN109685061A (en) | The recognition methods of mathematical formulae suitable for structuring | |
CN106845458A (en) | A kind of rapid transit label detection method of the learning machine that transfinited based on core | |
Ng et al. | Scalable Feature Extraction with Aerial and Satellite Imagery. | |
CN110046618B (en) | License plate recognition method based on machine learning and maximum extremum stable region | |
CN105719296A (en) | High speed binary connected domain marking method based on address-event expression | |
Luft | Automatic georeferencing of historical maps by geocoding | |
Andrey et al. | Road guidance sign recognition in urban areas by structure |
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 |