CN108520252A - Landmark identification method based on generalised Hough transform and wavelet transformation - Google Patents
Landmark identification method based on generalised Hough transform and wavelet transformation Download PDFInfo
- Publication number
- CN108520252A CN108520252A CN201810361896.4A CN201810361896A CN108520252A CN 108520252 A CN108520252 A CN 108520252A CN 201810361896 A CN201810361896 A CN 201810361896A CN 108520252 A CN108520252 A CN 108520252A
- Authority
- CN
- China
- Prior art keywords
- image
- hough transform
- wavelet
- low
- wavelet transformation
- 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
- 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
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
-
- 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
- G06V10/48—Extraction of image or video features by mapping characteristic values of the pattern into a parameter space, e.g. Hough transformation
-
- 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
- G06V10/52—Scale-space analysis, e.g. wavelet analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The landmark identification method based on generalised Hough transform and wavelet transformation that the invention discloses a kind of, includes the following steps：Step 1：Select standard picture construction feature image library；Step 2：Denoising operation is carried out for the image collected；Step 3：Edge extracting is carried out to the imagery exploitation generalised Hough transform after denoising；Step 4：Imagery exploitation wavelet decomposition after extraction is extracted into image low-frequency information；Step 5：Corresponding characteristic image is established according to image low-frequency information；Step 6：Match cognization is carried out to the characteristic image established and characteristic image library using Euclidean distance.Invention obtains image low-frequency information using wavelet decomposition herein, to construction feature image library, remains most essential, characteristic information that recognition effect is best in image, compares traditional recognition methods, accuracy rate higher.
Description
Technical field
The present invention relates to unmanned technical field more particularly to a kind of roads based on generalised Hough transform and wavelet transformation
Mark recognition methods.
Background technology
In recent years, advanced driving assistance system (ADAS) and unmanned technology are quickly grown, wherein landmark identification is to grind
One of hot spot studied carefully.Common road sign has deceleration, stopping, speed limit etc..Accurate detection and identification road sign can be occurred with less accident
Possibility, increase the safety of driving.For human, visual information is passed to after brain, can be by brain according to existing
Knowledge carry out information processing, and then identify and judge.
For machine, since road sign type is more, discrimination is little and illumination, acquires, blocks shadow with scale
It rings so that it is difficult to accurately identify road sign from complex scene.
The technology that there is now：Since different colours space is for the effect of the Traffic Sign Segment of the colors such as red, yellow, blue,
On color space road sign can be extracted using Threshold segmentation；For the identifying system of red road sign, using red filtering, edge
And the method for closed curve is detected, and feature extraction and target classification, Ke Yida are carried out with expert system and neural network
To target；Image is transformed into Gauss color space from RGB color, using two for the constant amount of viewing angle
Color as image indicates, then to color cluster and then extracts road sign using K-means；Exploitation identification stops road sign
System, which uses the color segmentation method of HSV space in context of detection, in terms of identification, using neural network method.But
It is that these methods always cannot be guaranteed higher accuracy rate, and use with limitation.
This patent designs an accurate general landmark identification method, for improving DAS (Driver Assistant System) and unmanned peace
It is of great significance entirely.
Invention content
The object of the present invention is to provide a kind of combination Gaussian smoothing filter, generalised Hough transforms, efficiently and accurately identify
Road calibration method.
The technical solution adopted by the present invention to solve the technical problems is：
A kind of landmark identification method based on generalised Hough transform and wavelet transformation is provided, is included the following steps：
Step 1：Select standard picture construction feature image library；
Step 2：Denoising operation is carried out for the image collected；
Step 3：Edge extracting is carried out to the imagery exploitation generalised Hough transform after denoising；
Step 4：Imagery exploitation wavelet decomposition after extraction is extracted into image low-frequency information；
Step 5：Corresponding characteristic image is established according to image low-frequency information；
Step 6：Match cognization is carried out to the characteristic image established and characteristic image library using Euclidean distance.
Above-mentioned technical proposal is connect, is specially in step 1：Select at least 20 treated standard pictures as road sign
Characteristic image library.
Above-mentioned technical proposal is connect, step 2 is specially：Letter unrelated in image is slackened or removed using Gaussian smoothing filter
Breath enhances the relevant information in image, realizes image denoising.For the validity of verification method, it is added in standard picture artificial
Noise, and by after denoising image and standard picture compare；
Wherein, Gaussian smoothing filter uses two-dimensional Gaussian function, is divided into the progress of two steps：1) by image and one-dimensional Gaussian function
Number carries out convolution；2) by the convolution results identical one-dimensional Gaussian function convolution vertical with direction.
Above-mentioned technical proposal is connect, edge detection is carried out using generalised Hough transform to the image after denoising in step 3, mainly
Using a kind of generalised Hough transform for all curves to detect the edge of geometry, accelerate to calculate using figure gradient amount
Method speed；It specifically includes：
3-1 drawing template establishments select reference point in image-region, its gradient direction angle φ, examination point are calculated to each marginal point
It is indicated with their line direction angle alpha and wire length r with the relative position of each boundary point, r, α and φ determine a side jointly
They are once arranged in R tables by the size of gradient direction φ and stored by edge point, and storage content is marginal points information (r, α),
Form R tables.
3-2 edge detections carry out edge detection, note according to the principle of generalised Hough transform according to preset threshold value
The marginal points information for recording image to be detected, in the statistics stage, according to different multi-party rotationally-varying, change R tables repositioning ginsengs
Examination point and statistical match number are sentenced for the multiple target object in image when the matching times of anchor point are more than the threshold value of setting
It is image reference point to determine current anchor point；Image is then needed according to coordinate relationship also because rotation scaling will appear multiple reference points
Former effective image.
Above-mentioned technical proposal is connect, step 4 specifically includes：
4-1 is analyzed using Haar small echos, is defined scaling function, is calculated corresponding wavelet function；
4-2 carries out wavelet transformation to signal, and weighted sum is obtained with different scaling functions；
2 power reduction scale of 4-3 wavelet basis functions, the details of the small echo embodiment signal of small scale, and large scale
Small echo embodies the profile of signal；Wavelet decomposition covers entire frequency domain, by choosing suitable filter, is reduced or removed and is extracted
Correlation between different characteristic；
4-4 is for two-dimensional image data, and using quick Mallat wavelet decompositions algorithm, each layer of wavelet transformation all can
By picture breakdown at four parts, low-frequency approximation, vertical high frequency, horizontal high-frequent and the diagonal high frequency section of image are respectively represented.
Above-mentioned technical proposal is connect, step 5 is specially：Wavelet transformation obtains the low frequency and high-frequency information of image, takes several roads
Logo image create training set picture library, using all images carry out wavelet decomposition and preserve its low-frequency wavelet coefficients generation image as
Characteristic image.
Above-mentioned technical proposal is connect, step 6 specifically includes：The image collected forms test set, will be each in test set
Width characteristics of low-frequency image is matched with the low-frequency image in characteristic image library, is then classified into the spy of Euclidean distance minimum
It levies in image, if two same mark informations of graphical representation, expression identifies successfully, otherwise indicates recognition failures.
The landmark identification system based on generalised Hough transform and wavelet transformation that the present invention also provides a kind of, including：
Module is established in characteristic image library, for selecting standard picture construction feature image library；
Denoising module, for carrying out denoising operation for the image collected；
Edge extracting module, for carrying out edge extracting to the imagery exploitation generalised Hough transform after denoising；
Image low-frequency information extraction module, for the imagery exploitation wavelet decomposition extraction image low-frequency information after extracting；
Characteristic image establishes module, for establishing corresponding characteristic image according to image low-frequency information；
Identification module, for carrying out match cognization to the characteristic image established and characteristic image library using Euclidean distance.
The present invention also provides a kind of computer readable storage mediums, including the computer program that can be executed by processor, should
Computer program specifically executes the landmark identification system based on generalised Hough transform and wavelet transformation of above-mentioned technical proposal.
The beneficial effect comprise that：The present invention can detect the curve of arbitrary shape using generalised Hough transform,
There is broader practice compared to traditional Hough transformation detection.Image low-frequency information is obtained using wavelet decomposition, to structure
Characteristic image library is built, most essential, characteristic information that recognition effect is best in image is remained, compares traditional recognition methods,
Accuracy rate higher.
Further, in conjunction with Gaussian smoothing filter, generalised Hough transform and wavelet transformation to the image collected
Reason and identification, are a kind of recognition methods of efficiently and accurately, recognition effect has larger improvement compared with conventional method.
Further, the present invention is had the advantage that using two-dimensional Gaussian function progress smothing filtering：
1), two-dimensional Gaussian function have rotational symmetry, it means that Gaussian smoothing filter subsequent edges detection in not
It can be biased to either direction.
2), Gaussian function is monotropic function, this shows the weighted mean of Gaussian filter neighborhood of pixels to replace the point
Pixel value, and each neighborhood territory pixel point weights are that dullness increases and decreases at a distance from central point with the point.
3), the Fourier transformation frequency spectrum of Gaussian function is single-lobe, it means that smoothed image will not be by unwanted height
Frequency signal is polluted, while remaining most of desired signal.
4), Gaussian filter width has parameter σ to characterize, and σ and smoothness are very simple, and σ is bigger, Gauss
Filter bandwidth is wider, and smoothness is better.
5), due to the separability of Gaussian function, big Gaussian filter can be effectively realized.
Description of the drawings
Present invention will be further explained below with reference to the attached drawings and examples, in attached drawing：
Fig. 1 is that the present invention is based on the landmark identifications of generalised Hough transform and wavelet transformation to calculate method flow diagram；
Fig. 2 is the training set picture library of the present invention；
Fig. 3 is that the present invention utilizes the comparison diagram after Gaussian smoothing filter denoising；
Fig. 4 is the edge image that the present invention is obtained using generalised Hough transform
Fig. 5 is the image that the present invention is obtained using wavelet decomposition；
Fig. 6 is the characteristic image library of the present invention；
Fig. 7 and Fig. 8 is the test result figure of the present invention.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, not
For limiting the present invention.
The embodiment of the present invention calculates method based on the landmark identification of generalised Hough transform and wavelet transformation, as shown in Figure 1, including
Following steps：
Step 1：Select standard picture construction feature image library；
Step 2：Denoising operation is carried out for the image collected；
Step 3：Edge extracting is carried out to the imagery exploitation generalised Hough transform after denoising；
Step 4：Imagery exploitation wavelet decomposition after extraction is extracted into image low-frequency information；
Step 5：Corresponding characteristic image is established according to low-frequency information obtained above；
Step 6：Match cognization is carried out to the characteristic image of above-mentioned foundation and characteristic image library using Euclidean distance；
It is specifically included in the step 1：Select 20 width standard picture construction feature picture libraries.
Step 2 specifically includes：
2-1 is to carry out denoising work to image first, using Gaussian smoothing filter.For the validity of verification method, marking
Be added man made noise in quasi- image, and by after denoising image and standard picture compare.
The Gaussian template that 2-2 this patents use is：
The Gaussian smoothing filter of 2-3 this patents uses two-dimensional Gaussian function, is divided into the progress of two steps：1) by image and one
It ties up Gaussian function and carries out convolution；2) by the convolution results identical one-dimensional Gaussian function convolution vertical with direction.
Step 3 carries out edge extracting to the imagery exploitation generalised Hough transform after denoising and specifically includes：
3-1 drawing template establishments first are selected reference point (generally barycenter) in image-region, are then calculated each marginal point
The relative position of its gradient direction angle φ, examination point and each boundary point can use their line direction angle alpha and wire length r tables
Show.R, α and φ determines a marginal point jointly, they are once arranged in R tables by the size of gradient direction φ and is stored, is deposited
It is marginal points information (r, α) to put content, forms R tables.
3-2 followed by edge detection carry out edge according to the principle of generalised Hough transform according to preset threshold value
Detection, records the marginal points information of image to be detected, in the statistics stage, according to different multi-party rotationally-varying, changes R tables again
Location reference point and statistical match number.For the multiple target object in image, the matching times of anchor point are more than the threshold of setting
When value, judgement current anchor point is image reference point.Image is then needed because rotation scaling will appear multiple reference points according to coordinate
Relationship restores effective image.
The imagery exploitation wavelet decomposition extraction image low-frequency information after extraction is specifically included in step 4：
4-1 is analyzed using Haar small echos, is defined scaling function and is
4-2 calculates corresponding wavelet function
4-3 carries out wavelet transformation to signal, and weighted sum is obtained with different scaling functions.
2 power reduction scale of 4-4 wavelet basis functions, the details of the small echo embodiment signal of small scale, and large scale
Small echo embodies the profile of signal.Wavelet decomposition can cover entire frequency domain, by choosing suitable filter, can greatly subtract
Correlation between small or the extracted different characteristic of removal.
4-5 is for two-dimensional image data, using quick Mallat wavelet decompositions algorithm.Each layer of wavelet transformation all can
By picture breakdown at four parts, low-frequency approximation, vertical high frequency, horizontal high-frequent and the diagonal high frequency section of image are respectively represented.
4-6 represent the scaling function of different information as：
Wherein HH, LH and HL then respectively represent diagonal, high-frequency information both vertically as well as horizontally, and LL represents low-frequency information,It is wavelet basis function, ψ (x) is scaling function, and g (x) is wavelet basis function weighted sum.
Corresponding characteristic image is established in step 5 according to low-frequency information obtained above to specifically include：
5-1 wavelet transformations obtain the low frequency and high-frequency information of image, and high-frequency information is usually the interference information in image,
Low-frequency information is characteristic information essential, that recognition effect is good.Therefore, we take several road sign image creation training set picture libraries,
Using all images carry out wavelet decomposition and preserve its low-frequency wavelet coefficients generation image as characteristic image, establish characteristic image
Library.
Match cognization is carried out in step 6 to the characteristic image of above-mentioned foundation and characteristic image library with Euclidean distance specifically to wrap
It includes：
The image that 6-1 takes is combined into test set, will be in each width low-frequency image and characteristic image library in test set
Low-frequency image matched, be then classified into the characteristic image of Euclidean distance minimum, if two graphical representations are same
One mark information, expression identifies successfully, otherwise indicates recognition failures.
The feature extraction and identification of the present invention uses the Feature extraction and recognition method based on wavelet transformation, overcomes traditional
The shortcomings that feature extracting method based on Fourier, guarantee do not lose high-frequency information and do not lose low-frequency information；Experimental result
Show compared with standard picture：The algorithm can overcome Gaussian white noise caused by having Changes in weather, illumination variation and acquisition
The influence of caused noise when mired rotation, ratio variation and translation occurs for road sign image.Compared with prior art, the present invention can
To identify road sign information effectively in real time, precision high reliability is strong, and error in judgement rate is small in identification process, can be extensive
For in the advanced driving assistance system of automatic driving vehicle.
The landmark identification system based on generalised Hough transform and wavelet transformation that the present invention also provides a kind of, including：
Module is established in characteristic image library, for selecting standard picture construction feature image library；
Denoising module, for carrying out denoising operation for the image collected；
Edge extracting module, for carrying out edge extracting to the imagery exploitation generalised Hough transform after denoising；
Image low-frequency information extraction module, for the imagery exploitation wavelet decomposition extraction image low-frequency information after extracting；
Characteristic image establishes module, for establishing corresponding characteristic image according to image low-frequency information；
Identification module, for carrying out match cognization to the characteristic image established and characteristic image library using Euclidean distance.
The present invention also provides a kind of computer readable storage mediums, including the computer program that can be executed by processor, should
Computer program specifically executes the landmark identification system based on generalised Hough transform and wavelet transformation of above-mentioned technical proposal.
In the specific embodiment of the present invention, the landmark identification method based on generalised Hough transform and wavelet transformation, packet
Include following steps：
Step 1：Training set picture library is built, if Fig. 2 is the training set picture library that the present invention is built；
Step 2：Denoising operation is carried out for the image collected, if Fig. 3 is to the comparison after a certain noise image denoising
Figure；
Step 3：Edge extracting is carried out to the imagery exploitation generalised Hough transform after denoising, as Fig. 4 present invention obtains
Edge image；
Step 4：Image low-frequency information is extracted using wavelet decomposition, if Fig. 5 is the figure that the present invention is obtained using wavelet decomposition
Picture；
Step 5：Corresponding characteristic image library is established according to low-frequency information, if Fig. 6 is that the present invention is established using low-frequency information
Characteristic image library；
Step 6：It is identified using Euclidean distance matching image, if Fig. 7 and Fig. 8 is the test result that the present invention obtains
Figure；
For the present invention when carrying out field test, algorithm is average to take 61ms per frame, and accuracy rate possesses very well 90% or more
Real-time and accuracy.
To sum up, the present invention is had the advantage that using two-dimensional Gaussian function progress smothing filtering：
1), two-dimensional Gaussian function have rotational symmetry, it means that Gaussian smoothing filter subsequent edges detection in not
It can be biased to either direction.
2), Gaussian function is monotropic function, this shows the weighted mean of Gaussian filter neighborhood of pixels to replace the point
Pixel value, and each neighborhood territory pixel point weights are that dullness increases and decreases at a distance from central point with the point.
3), the Fourier transformation frequency spectrum of Gaussian function is single-lobe, it means that smoothed image will not be by unwanted height
Frequency signal is polluted, while remaining most of desired signal.
4), Gaussian filter width has parameter σ to characterize, and σ and smoothness are very simple, and σ is bigger, Gauss
Filter bandwidth is wider, and smoothness is better.
5), due to the separability of Gaussian function, big Gaussian filter can be effectively realized.
The present invention can detect the curve of arbitrary shape using generalised Hough transform, be detected compared to traditional Hough transformation
With broader practice.
Invention obtains image low-frequency information using wavelet decomposition herein, to construction feature image library, remains in image
Characteristic information most essential, recognition effect is best compares traditional recognition methods, accuracy rate higher.
It should be understood that for those of ordinary skills, it can be modified or changed according to the above description,
And all these modifications and variations should all belong to the protection domain of appended claims of the present invention.
Claims (9)
1. a kind of landmark identification method based on generalised Hough transform and wavelet transformation, which is characterized in that include the following steps：
Step 1：Select standard picture construction feature image library；
Step 2：Denoising operation is carried out for the image collected；
Step 3：Edge extracting is carried out to the imagery exploitation generalised Hough transform after denoising；
Step 4：Imagery exploitation wavelet decomposition after extraction is extracted into image low-frequency information；
Step 5：Corresponding characteristic image is established according to image low-frequency information；
Step 6：Match cognization is carried out to the characteristic image established and characteristic image library using Euclidean distance.
2. the landmark identification method based on generalised Hough transform and wavelet transformation as described in claim 1, which is characterized in that step
It is specially in rapid 1：Select characteristic image library of at least 20 treated standard pictures as road sign.
3. the landmark identification method based on generalised Hough transform and wavelet transformation as described in claim 1, which is characterized in that step
Rapid 2 are specially：Denoising work is carried out to the image collected, using Gaussian smoothing filter image is gone dry, is verification method
Validity, is added man made noise in standard picture, and by after denoising image and standard picture compare；
Wherein, Gaussian smoothing filter uses two-dimensional Gaussian function, is divided into the progress of two steps：1）By image and one-dimensional Gaussian function into
Row convolution；2）By the convolution results identical one-dimensional Gaussian function convolution vertical with direction.
4. the landmark identification method based on generalised Hough transform and wavelet transformation as described in claim 1, which is characterized in that step
Edge detection is carried out in rapid 3 using generalised Hough transform to the image after denoising to specifically include：
3-1 drawing template establishments select reference point in image-region, its gradient direction angle, examination point and Ge Bian are calculated to each marginal point
The relative position of boundary's point is indicated with their line deflection and wire length, and determines a marginal point jointly, they are pressed
The size of gradient direction is once arranged in table and stores, and storage content is marginal points information, forms table；
3-2 edge detections carry out edge detection according to the principle of generalised Hough transform according to preset threshold value, and record waits for
The marginal points information of detection image according to different multi-party rotationally-varying, changed table and repositions reference point simultaneously in the statistics stage
Statistical match number, for the multiple target object in image, when the matching times of anchor point are more than the threshold value of setting, judgement is current
Anchor point is image reference point；Image then needs effective according to the reduction of coordinate relationship because rotation scaling will appear multiple reference points
Image.
5. the landmark identification method based on generalised Hough transform and wavelet transformation as described in claim 1, which is characterized in that step
Rapid 4 specifically include：
4-1 is analyzed using Haar small echos, is defined scaling function, is calculated corresponding wavelet function；
4-2 carries out wavelet transformation to signal, and weighted sum is obtained with different scaling functions；
2 power reduction scale of 4-3 wavelet basis functions, the details of the small echo embodiment signal of small scale, and the small echo of large scale
Embody the profile of signal；Wavelet decomposition covers entire frequency domain, by choosing suitable filter, is reduced or removed and extracts difference
Correlation between feature；
4-4 is for two-dimensional image data, and using quick Mallat wavelet decompositions algorithm, each layer of wavelet transformation can all incite somebody to action
Picture breakdown respectively represents low-frequency approximation, vertical high frequency, horizontal high-frequent and the diagonal high frequency section of image at four parts.
6. the landmark identification method based on generalised Hough transform and wavelet transformation as described in claim 1, which is characterized in that step
Rapid 5 are specially：Wavelet transformation obtains the low frequency and high-frequency information of image, several road sign image creation training set picture libraries is taken, by institute
There is image to carry out wavelet decomposition and preserves the image of its low-frequency wavelet coefficients generation as characteristic image.
7. the landmark identification method based on generalised Hough transform and wavelet transformation as described in claim 1, which is characterized in that step
Rapid 6 specifically include：The image collected forms test set, by each width characteristics of low-frequency image and the characteristic image library in test set
In low-frequency image matched, be then classified into the characteristic image of Euclidean distance minimum, if two graphical representations
Same mark information, expression identifies successfully, otherwise indicates recognition failures.
8. a kind of landmark identification system based on generalised Hough transform and wavelet transformation, which is characterized in that including：
Module is established in characteristic image library, for selecting standard picture construction feature image library；
Denoising module, for carrying out denoising operation for the image collected；
Edge extracting module, for carrying out edge extracting to the imagery exploitation generalised Hough transform after denoising；
Image low-frequency information extraction module, for the imagery exploitation wavelet decomposition extraction image low-frequency information after extracting；
Characteristic image establishes module, for establishing corresponding characteristic image according to image low-frequency information；
Identification module, for carrying out match cognization to the characteristic image established and characteristic image library using Euclidean distance.
9. a kind of computer readable storage medium, which is characterized in that including the computer program that can be executed by processor, the calculating
Machine program specifically executes the landmark identification based on generalised Hough transform and wavelet transformation as described in any one of claim 1-7
System.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810361896.4A CN108520252B (en) | 2018-04-20 | 2018-04-20 | Road sign identification method based on generalized Hough transform and wavelet transform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810361896.4A CN108520252B (en) | 2018-04-20 | 2018-04-20 | Road sign identification method based on generalized Hough transform and wavelet transform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108520252A true CN108520252A (en) | 2018-09-11 |
| CN108520252B CN108520252B (en) | 2022-03-01 |
Family
ID=63428942
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810361896.4A Active CN108520252B (en) | 2018-04-20 | 2018-04-20 | Road sign identification method based on generalized Hough transform and wavelet transform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108520252B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109858409A (en) * | 2019-01-18 | 2019-06-07 | 深圳壹账通智能科技有限公司 | Manual figure conversion method, device, equipment and medium |
| CN113688828A (en) * | 2021-07-23 | 2021-11-23 | 山东云海国创云计算装备产业创新中心有限公司 | Bad element identification method and related device |
| CN117197072A (en) * | 2023-09-07 | 2023-12-08 | 石家庄铁道大学 | Automatic object counting method based on machine vision |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1897015A (en) * | 2006-05-18 | 2007-01-17 | 王海燕 | Method and system for inspecting and tracting vehicle based on machine vision |
| US20120143037A1 (en) * | 2009-04-07 | 2012-06-07 | Kayvan Najarian | Accurate Pelvic Fracture Detection for X-Ray and CT Images |
| CN106022218A (en) * | 2016-05-06 | 2016-10-12 | 浙江工业大学 | Palm print palm vein image layer fusion method based on wavelet transformation and Gabor filter |
| CN106324618A (en) * | 2015-06-17 | 2017-01-11 | 百利得汽车主动安全系统（苏州）有限公司 | System for detecting lane line based on laser radar and realization method thereof |
| CN107292263A (en) * | 2017-06-19 | 2017-10-24 | 深圳市创艺工业技术有限公司 | A kind of motor-driven, electric vehicle automatic driving recognition system |
-
2018
- 2018-04-20 CN CN201810361896.4A patent/CN108520252B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1897015A (en) * | 2006-05-18 | 2007-01-17 | 王海燕 | Method and system for inspecting and tracting vehicle based on machine vision |
| US20120143037A1 (en) * | 2009-04-07 | 2012-06-07 | Kayvan Najarian | Accurate Pelvic Fracture Detection for X-Ray and CT Images |
| CN106324618A (en) * | 2015-06-17 | 2017-01-11 | 百利得汽车主动安全系统（苏州）有限公司 | System for detecting lane line based on laser radar and realization method thereof |
| CN106022218A (en) * | 2016-05-06 | 2016-10-12 | 浙江工业大学 | Palm print palm vein image layer fusion method based on wavelet transformation and Gabor filter |
| CN107292263A (en) * | 2017-06-19 | 2017-10-24 | 深圳市创艺工业技术有限公司 | A kind of motor-driven, electric vehicle automatic driving recognition system |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109858409A (en) * | 2019-01-18 | 2019-06-07 | 深圳壹账通智能科技有限公司 | Manual figure conversion method, device, equipment and medium |
| CN113688828A (en) * | 2021-07-23 | 2021-11-23 | 山东云海国创云计算装备产业创新中心有限公司 | Bad element identification method and related device |
| CN113688828B (en) * | 2021-07-23 | 2023-09-29 | 山东云海国创云计算装备产业创新中心有限公司 | Bad element identification method and related device |
| CN117197072A (en) * | 2023-09-07 | 2023-12-08 | 石家庄铁道大学 | Automatic object counting method based on machine vision |
| CN117197072B (en) * | 2023-09-07 | 2024-04-05 | 石家庄铁道大学 | Automatic object counting method based on machine vision |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108520252B (en) | 2022-03-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107610114B (en) | optical satellite remote sensing image cloud and snow fog detection method based on support vector machine | |
| WO2019196130A1 (en) | Classifier training method and device for vehicle-mounted thermal imaging pedestrian detection | |
| CN108805904B (en) | Moving ship detection and tracking method based on satellite sequence image | |
| CN109918971B (en) | Method and device for detecting number of people in monitoring video | |
| CN104463877B (en) | A kind of water front method for registering based on radar image Yu electronic chart information | |
| CN104361343B (en) | Vehicle type recognition method and its device | |
| CN102750540B (en) | Morphological filtering enhancement-based maximally stable extremal region (MSER) video text detection method | |
| CN104598933B (en) | A kind of image reproduction detection method based on multi-feature fusion | |
| CN111340824B (en) | Image feature segmentation method based on data mining | |
| Tajbakhsh et al. | Automatic polyp detection using global geometric constraints and local intensity variation patterns | |
| CN108694393A (en) | A kind of certificate image text area extraction method based on depth convolution | |
| CN104299009B (en) | License plate character recognition method based on multi-feature fusion | |
| CN104217221A (en) | Method for detecting calligraphy and paintings based on textural features | |
| CN107066972B (en) | Natural scene Method for text detection based on multichannel extremal region | |
| CN103093274B (en) | Method based on the people counting of video | |
| CN103996198A (en) | Method for detecting region of interest in complicated natural environment | |
| CN108520252A (en) | Landmark identification method based on generalised Hough transform and wavelet transformation | |
| CN106845328A (en) | A kind of Intelligent human-face recognition methods and system based on dual camera | |
| CN106096491B (en) | Automatic identification method for microaneurysms in fundus color photographic image | |
| CN109726681A (en) | It is a kind of that location algorithm is identified based on the blind way of machine learning identification and image segmentation | |
| CN108154147A (en) | The region of interest area detecting method of view-based access control model attention model | |
| CN109886168B (en) | Ground traffic sign identification method based on hierarchy | |
| CN109784216B (en) | Vehicle-mounted thermal imaging pedestrian detection Rois extraction method based on probability map | |
| Sultana et al. | Detection of pectoral muscle in mammograms using a mean-shift segmentation approach | |
| Feng et al. | Iris localization with dual coarse-to-fine strategy |
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 |