CN105069419A - Traffic sign detection method based on edge color pair and characteristic filters - Google Patents
Traffic sign detection method based on edge color pair and characteristic filters Download PDFInfo
- Publication number
- CN105069419A CN105069419A CN201510447215.2A CN201510447215A CN105069419A CN 105069419 A CN105069419 A CN 105069419A CN 201510447215 A CN201510447215 A CN 201510447215A CN 105069419 A CN105069419 A CN 105069419A
- Authority
- CN
- China
- Prior art keywords
- traffic sign
- color
- edge
- beta
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention relates to a traffic sign detection method based on an edge color pair and characteristic filters. A collected image is pre-processed, noise interference is eliminated, and the color contrast of the image is improved; edge detection is carried out on the pre-processed image in a color edge detection method based on direction, area and distance measurement, and edge information is extracted; edge color pair detection is carried out on edge points, and edge points which do not satisfy color collocation of a traffic sign are removed; morphological close operation is carried out on residual edge points to form a communication area, and realize coarse positioning of the traffic sign; and a characteristic filter is designed to carry out characteristic filtering on ROI obtained by coarse positioning, ROI that satisfies the characteristic of the traffic sign is reserved, ROI that does not satisfy the characteristic of the traffic sign is removed, and a final detected traffic sign is obtained. The faded traffic sign can be detected effectively, the two levels of characteristic filters fully utilizes the area characteristic and the symmetry characteristic of the traffic sign, and the correct rate of detection is improved.
Description
Technical field
The present invention relates to a kind of computer image processing technology, particularly a kind of method for traffic sign detection based on edge color pair and Feature Selection device.
Background technology
Traffic sign transmits customizing messages with graphical symbol and word, in order to the direction of traffic that regulates the traffic, indicates to ensure the coast is clear and safe facility.Carry out identification to traffic sign to have important practical significance: be conducive to safe driving on the one hand, reduce the incidence of traffic hazard; Be conducive to the development promoting intelligent vehicle on the other hand.
Traffic Sign Recognition mainly comprises road traffic sign detection and traffic sign and to classify two modules, and can wherein road traffic sign detection be key modules in Traffic Sign Recognition System, accurately detect that traffic sign will directly affect the success or failure of Traffic Sign Recognition System.
Method at present for road traffic sign detection mainly contains: Color invariants and tower-type gradient direction histogram detection method; Based on radial symmetry detection method; Color filter method; The detection method of view-based access control model antagonistic pairs; The thresholding method etc. that HIS space tone is combined with saturation degree.There are two common shortcomings in these methods: flase drop can occur when existing and traffic sign has the chaff interference of similar shape with similar background color; Cannot be detected when traffic sign exists COLOR FASTNESS PROBLEM.
Summary of the invention
The present invention be directed to present method for traffic sign detection Problems existing, propose a kind of method for traffic sign detection based on edge color pair and Feature Selection device, effectively solve the problem that homologue interference and traffic sign fade.
Technical scheme of the present invention is: a kind of method for traffic sign detection based on edge color pair and Feature Selection device, specifically comprises the steps:
1) carry out pre-service to the image collected, stress release treatment disturbs, and increases color of image contrast;
2) based on the Color edge detection method of Distance Measure of Direction Region, rim detection is carried out to pretreated image employing, extract marginal information;
3) edge point carries out edge color pair detection, and the marginal point meeting traffic sign colour match remains, and incongruently removes;
4) then morphology closed operation is carried out to remaining marginal point, form UNICOM region, be defined as traffic sign suspicious region ROI, achieve traffic sign coarse positioning;
5) finally secondary characteristics screening washer is designed, Feature Selection is carried out to the ROI region of coarse positioning, the first order is area features screening, add up the pixel number of each ROI region, the number of pixel is denoted as the area of ROI, setting area max-thresholds and minimum threshold, if the area of ROI is in these two threshold ranges, then retain this ROI region, if not in this threshold range, then get rid of this ROI region; The second level is symmetrical feature screening washer, first, utilizes canny operator to carry out rim detection to ROI region, extracts the marginal information of ROI region; Then seek out gradient magnitude and the direction of marginal point, and be multiplied by a weights K at the gradient direction of marginal point, so just can potential symcenter be weighted; Arrange a threshold value M, if the weights that there is any in certain ROI region are greater than M, then this point is exactly symmetric points, and this ROI region is exactly symmetrical region, can be judged to be traffic sign region, otherwise just this ROI region is got rid of;
6) carrying out horizontal and vertical projection to left traffic sign region after screening, can by the Traffic Sign Segment in original image out, be just the traffic sign finally detected.
Described step 2) Color edge detection method carries out rim detection, and extract marginal information, concrete steps are as follows:
A: each pixel in image is handled as follows: (the i in image, j) individual pixel C (i, j) represent, then C (i, j)=(R (i, j), G (i, j), B (i, j)), wherein R (i, j), G (i, j), B (i, j) be pixel C (i, j) red in RGB color space, green and blue color component, with C (i, j) window of (2s+1) (2s+1) is made centered by, wherein s be more than or equal to 1 constant, C (i is crossed in window, j) the line segment L becoming β (0≤β ￡ π) with vertical direction is made, window is divided into w
1and w
2two parts, statistics w
1and w
2two number of pixels N comprised, according to formula
Calculate w respectively
1and w
2the color component mean value of interior pixel R, G, B
with
wherein
w respectively
1and w
2red, green, blue three components of an interior kth pixel; B: according to formula
Calculate w
1and w
2between average color distance, when β changes between 0 and π, D
βa maximal value must be had, note
0≤β≤π, D
sas the tolerance whether C (i, j) is marginal point, a threshold value T is set, if D
sduring > T, this pixel is considered to a marginal point, and to the calculating that all pixels in image are carried out as above, finally obtain colour edging image, wherein threshold value T calculates according to formula T=μ+3 σ, and wherein μ is D
saverage, σ is D
sstandard variance.
Described step 3) edge color pair detection method is as follows: get edge a bit, one the 1 × linear window of (2m+1) is made in the vertical direction of this tangent line, wherein m be more than or equal to 1 integer, the pixel with traffic sign background color or symbol color is first searched in the side of linear window inward flange point, illustrate that it is not traffic sign marginal point if can not find, it is removed in edge image, if find, then there is in opposite side search the pixel of another color of matching with this color, if opposite side also can not find, illustrate that it neither traffic sign marginal point, this marginal point is removed equally in edge image, if find, illustrate that it may be the marginal point of traffic sign, it is retained in edge image, utilize support vector machines to judge color.
Beneficial effect of the present invention is: the method for traffic sign detection that the present invention is based on edge color pair and Feature Selection device, propose and adopt edge color pair to carry out coarse positioning to traffic sign, if containing when there is the chaff interference of similar shape and background color to traffic sign in the image collected, road traffic sign detection is gone out, can effectively detect the traffic sign faded; Devise a secondary characteristics screening washer in the present invention, make full use of area features and the symmetrical feature of traffic sign, the accuracy of detection can be improved.
Accompanying drawing explanation
Fig. 1 is the method for traffic sign detection process flow diagram that the present invention is based on edge color pair and Feature Selection device;
Fig. 2 is traffic sign coarse positioning process flow diagram in the present invention;
Fig. 3 is traffic sign fine positioning process flow diagram in the present invention;
Fig. 4 is the former figure of embodiment of the present invention road traffic sign detection;
Fig. 5 is the outline map after embodiment of the present invention Color edge detection;
Fig. 6 is the outline map after embodiment of the present invention edge color pair detects;
Fig. 7 is the ROI region figure obtained after embodiment of the present invention closing operation of mathematical morphology;
Fig. 8 is the traffic sign areal map after embodiment of the present invention Feature Selection;
Fig. 9 is the traffic indication map that the embodiment of the present invention detects.
Embodiment
Process flow diagram of the present invention as shown in Figure 1, first carries out pre-service to the image collected, and object is in order to stress release treatment interference, increases color of image contrast; Secondly pretreated image is adopted and carry out rim detection based on the Color edge detection method of Distance Measure of Direction Region, extract marginal information; Then edge point carries out edge color pair detection, and the marginal point meeting traffic sign colour match remains, and incongruently removes; Then morphology closed operation is carried out to remaining marginal point, forms UNICOM region, these regions we be called traffic sign suspicious region (ROI), so just achieve traffic sign coarse positioning; Finally design a Feature Selection device, carry out Feature Selection to the ROI region of coarse positioning, the ROI region meeting traffic sign feature remains, and incongruently removes, and the region of the ROI be left like this is exactly the traffic sign finally detected.
Concrete steps are as follows:
First traffic sign coarse positioning process flow diagram as shown in Figure 2,
Step1: carry out pre-service to the image collected, comprises the operation such as medium filtering, histogram equalization, as shown in Figure 4 the former figure of embodiment road traffic sign detection;
Step2: pretreated image is carried out Color edge detection, extracts colour edging information.Each pixel in image is handled as follows: (the i in image, j) individual pixel C (i, j) represent, then C (i, j)=(R (i, j), G (i, j), B (i, j)), wherein R (i, j), G (i, j), B (i, j) be pixel C (i, j) red in RGB color space, green and blue color component, with C (i, j) window of (2s+1) (2s+1) is made centered by, wherein s be more than or equal to 1 constant, C (i is crossed in window, j) the line segment L becoming β (0≤β ￡ π) with vertical direction is made, window is divided into w
1and w
2two parts, statistics w
1and w
2two number of pixels N comprised, according to formula
Calculate w respectively
1and w
2the color component mean value of interior pixel R, G, B
with
wherein
w respectively
1and w
2red, green, blue three components of an interior kth pixel.Again according to formula
Calculate w
1and w
2between average color distance, when β changes between 0 and π, D
βa maximal value must be had, note
0≤β≤π, D
sas the tolerance whether C (i, j) is marginal point, (threshold value T can calculate according to formula T=μ+3 σ, and wherein μ is D to arrange a threshold value T
saverage, σ is D
sstandard variance), if D
sduring > T, this pixel is considered to a marginal point.To the calculating that all pixels in image are carried out as above, finally colour edging image can be obtained.As the outline map that Fig. 5 is after embodiment of the present invention Color edge detection.
Step3: respectively Color pair check processing is carried out for each marginal point on next edge detected in Step2.Edge color pair detection method is as follows: get edge a bit, one the 1 × linear window of (2m+1) is made in the vertical direction of this tangent line, wherein m be more than or equal to 1 integer, the pixel with traffic sign background color or symbol color is first searched in the side of linear window inward flange point, illustrate that it is not traffic sign marginal point if can not find, it is removed in edge image, if find, then there is in opposite side search the pixel of another color of matching with this color, if opposite side also can not find, illustrate that it neither traffic sign marginal point, this marginal point is removed equally in edge image, if find, illustrate that it may be the marginal point of traffic sign, it is retained in edge image.Here utilize support vector machine (SVM) technology to judge color.As the outline map that Fig. 6 is after embodiment of the present invention edge color pair detects.
Step4: after the edge color pair of Step3 detects, the marginal point not meeting traffic sign colour match is removed, remaining marginal point is exactly likely the marginal point of traffic sign, in order to carry out next step process, we utilize the closed operation in morphology to obtain connected region, these connected regions, we are referred to as traffic sign suspicious region (ROI); If Fig. 7 is the ROI region figure obtained after embodiment of the present invention closing operation of mathematical morphology.
Traffic sign coarse positioning is achieved to this.Then traffic sign fine positioning process flow diagram as shown in Figure 3, enters fine positioning, uses the Feature Selection device of design to screen:
Step5: because traffic sign has fixing geometric properties, such as the area of traffic sign excessively shape that is too small, traffic sign all can not can not have symmetry etc., and the geometric properties of traffic sign can be utilized to screen ROI region.Design a secondary characteristics screening washer, the first order is area features screening, add up the pixel number of each ROI region, the number of pixel is denoted as the area of ROI, setting area max-thresholds and minimum threshold, if the area of ROI is in these two threshold ranges, then retain this ROI region, if not in this threshold range, then get rid of this ROI region; The second level is symmetrical feature screening washer, first, utilizes canny operator to carry out rim detection to ROI region, extracts the marginal information of ROI region; Then seek out gradient magnitude and the direction of marginal point, and be multiplied by a weights K at the gradient direction of marginal point, so just can potential symcenter be weighted; Arrange a threshold value M, if the weights that there is any in certain ROI region are greater than M, then this point is exactly symmetric points, and this ROI region is exactly symmetrical region, can be judged to be traffic sign region, otherwise just this ROI region is got rid of.As the traffic sign areal map that Fig. 8 is after embodiment of the present invention Feature Selection.
Step6: carry out horizontal and vertical projection to left traffic sign region after screening, just can by the Traffic Sign Segment in original image out.If Fig. 9 is the traffic indication map that the embodiment of the present invention detects.
Claims (3)
1., based on a method for traffic sign detection for edge color pair and Feature Selection device, it is characterized in that, specifically comprise the steps:
1) carry out pre-service to the image collected, stress release treatment disturbs, and increases color of image contrast;
2) based on the Color edge detection method of Distance Measure of Direction Region, rim detection is carried out to pretreated image employing, extract marginal information;
3) edge point carries out edge color pair detection, and the marginal point meeting traffic sign colour match remains, and incongruently removes;
4) then morphology closed operation is carried out to remaining marginal point, form UNICOM region, be defined as traffic sign suspicious region ROI, achieve traffic sign coarse positioning;
5) finally secondary characteristics screening washer is designed, Feature Selection is carried out to the ROI region of coarse positioning, the first order is area features screening, add up the pixel number of each ROI region, the number of pixel is denoted as the area of ROI, setting area max-thresholds and minimum threshold, if the area of ROI is in these two threshold ranges, then retain this ROI region, if not in this threshold range, then get rid of this ROI region; The second level is symmetrical feature screening washer, first, utilizes canny operator to carry out rim detection to ROI region, extracts the marginal information of ROI region; Then seek out gradient magnitude and the direction of marginal point, and be multiplied by a weights K at the gradient direction of marginal point, so just can potential symcenter be weighted; Arrange a threshold value M, if the weights that there is any in certain ROI region are greater than M, then this point is exactly symmetric points, and this ROI region is exactly symmetrical region, can be judged to be traffic sign region, otherwise just this ROI region is got rid of;
6) carrying out horizontal and vertical projection to left traffic sign region after screening, can by the Traffic Sign Segment in original image out, be just the traffic sign finally detected.
2., according to claim 1 based on the method for traffic sign detection of edge color pair and Feature Selection device, it is characterized in that, described step 2) Color edge detection method carries out rim detection, and extract marginal information, concrete steps are as follows:
A: each pixel in image is handled as follows: (the i in image, j) individual pixel C (i, j) represent, then C (i, j)=(R (i, j), G (i, j), B (i, j)), wherein R (i, j), G (i, j), B (i, j) be pixel C (i, j) red in RGB color space, green and blue color component, with C (i, j) window of (2s+1) (2s+1) is made centered by, wherein s be more than or equal to 1 constant, C (i is crossed in window, j) the line segment L becoming β (0≤β≤π) with vertical direction is made, window is divided into w
1and w
2two parts, statistics w
1and w
2two number of pixels N comprised, according to formula
Calculate w respectively
1and w
2the color component mean value of interior pixel R, G, B
with
wherein
w respectively
1and w
2red, green, blue three components of an interior kth pixel;
B: according to formula
Calculate w
1and w
2between average color distance, when β changes between 0 and π, D
βa maximal value must be had, note
0≤β≤π, D
sas the tolerance whether C (i, j) is marginal point, a threshold value T is set, if D
sduring >T, this pixel is considered to a marginal point, and to the calculating that all pixels in image are carried out as above, finally obtain colour edging image, wherein threshold value T calculates according to formula T=μ+3 σ, and wherein μ is D
saverage, σ is D
sstandard variance.
3. according to claim 1 or 2 based on the method for traffic sign detection of edge color pair and Feature Selection device, it is characterized in that, described step 3) edge color pair detection method is as follows: get edge a bit, one the 1 × linear window of (2m+1) is made in the vertical direction of this tangent line, wherein m be more than or equal to 1 integer, the pixel with traffic sign background color or symbol color is first searched in the side of linear window inward flange point, illustrate that it is not traffic sign marginal point if can not find, it is removed in edge image, if find, then there is in opposite side search the pixel of another color of matching with this color, if opposite side also can not find, illustrate that it neither traffic sign marginal point, this marginal point is removed equally in edge image, if find, illustrate that it may be the marginal point of traffic sign, it is retained in edge image, utilize support vector machines to judge color.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510447215.2A CN105069419A (en) | 2015-07-27 | 2015-07-27 | Traffic sign detection method based on edge color pair and characteristic filters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510447215.2A CN105069419A (en) | 2015-07-27 | 2015-07-27 | Traffic sign detection method based on edge color pair and characteristic filters |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105069419A true CN105069419A (en) | 2015-11-18 |
Family
ID=54498781
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510447215.2A Pending CN105069419A (en) | 2015-07-27 | 2015-07-27 | Traffic sign detection method based on edge color pair and characteristic filters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105069419A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105787475A (en) * | 2016-03-29 | 2016-07-20 | 西南交通大学 | Traffic sign detection and identification method under complex environment |
| CN105825203A (en) * | 2016-03-30 | 2016-08-03 | 大连理工大学 | Ground arrowhead sign detection and identification method based on dotted pair matching and geometric structure matching |
| CN108734131A (en) * | 2018-05-22 | 2018-11-02 | 杭州电子科技大学 | A kind of traffic sign symmetry detection methods in image |
| CN109559536A (en) * | 2018-12-10 | 2019-04-02 | 百度在线网络技术（北京）有限公司 | Traffic lights, traffic light recognition method, device, equipment and storage medium |
| CN111192250A (en) * | 2019-12-30 | 2020-05-22 | 合肥联宝信息技术有限公司 | Data processing method and device, computer storage medium and computer |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101408942A (en) * | 2008-04-17 | 2009-04-15 | 浙江师范大学 | Method for locating license plate under a complicated background |
| CN101504717A (en) * | 2008-07-28 | 2009-08-12 | 上海高德威智能交通系统有限公司 | Characteristic area positioning method, car body color depth and color recognition method |
| CN101599125A (en) * | 2009-06-11 | 2009-12-09 | 上海交通大学 | The binarization method that the complex background hypograph is handled |
-
2015
- 2015-07-27 CN CN201510447215.2A patent/CN105069419A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101408942A (en) * | 2008-04-17 | 2009-04-15 | 浙江师范大学 | Method for locating license plate under a complicated background |
| CN101504717A (en) * | 2008-07-28 | 2009-08-12 | 上海高德威智能交通系统有限公司 | Characteristic area positioning method, car body color depth and color recognition method |
| CN101599125A (en) * | 2009-06-11 | 2009-12-09 | 上海交通大学 | The binarization method that the complex background hypograph is handled |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105787475A (en) * | 2016-03-29 | 2016-07-20 | 西南交通大学 | Traffic sign detection and identification method under complex environment |
| CN105825203A (en) * | 2016-03-30 | 2016-08-03 | 大连理工大学 | Ground arrowhead sign detection and identification method based on dotted pair matching and geometric structure matching |
| CN105825203B (en) * | 2016-03-30 | 2018-12-18 | 大连理工大学 | Based on point to matching and the matched ground arrow mark detection of geometry and recognition methods |
| CN108734131A (en) * | 2018-05-22 | 2018-11-02 | 杭州电子科技大学 | A kind of traffic sign symmetry detection methods in image |
| CN108734131B (en) * | 2018-05-22 | 2021-08-17 | 杭州电子科技大学 | Method for detecting symmetry of traffic sign in image |
| CN109559536A (en) * | 2018-12-10 | 2019-04-02 | 百度在线网络技术（北京）有限公司 | Traffic lights, traffic light recognition method, device, equipment and storage medium |
| CN111192250A (en) * | 2019-12-30 | 2020-05-22 | 合肥联宝信息技术有限公司 | Data processing method and device, computer storage medium and computer |
| CN111192250B (en) * | 2019-12-30 | 2022-02-08 | 合肥联宝信息技术有限公司 | Computer B-side frame detection method and device, computer storage medium and computer |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105069419A (en) | Traffic sign detection method based on edge color pair and characteristic filters | |
| CN103235938B (en) | The method and system of car plate detection and indentification | |
| Omachi et al. | Detection of traffic light using structural information | |
| CN109657632B (en) | Lane line detection and identification method | |
| CN102043950B (en) | Vehicle outline recognition method based on canny operator and marginal point statistic | |
| CN102855622B (en) | A kind of infrared remote sensing image sea ship detection method based on significance analysis | |
| CN103605977B (en) | Extracting method of lane line and device thereof | |
| CN102867188B (en) | Method for detecting seat state in meeting place based on cascade structure | |
| CN103473571A (en) | Human detection method | |
| CN104732227A (en) | Rapid license-plate positioning method based on definition and luminance evaluation | |
| CN102968646B (en) | A kind of detection method of license plate based on machine learning | |
| CN102999749B (en) | Based on the securing band violation event intelligent detecting method of Face datection | |
| CN102880863B (en) | Method for positioning license number and face of driver on basis of deformable part model | |
| CN103679677B (en) | A kind of bimodulus image decision level fusion tracking updating mutually based on model | |
| CN109670515A (en) | A kind of detection method and system changed for building in unmanned plane image | |
| CN103824091B (en) | A kind of licence plate recognition method for intelligent transportation system | |
| CN103679168A (en) | Detection method and detection device for character region | |
| CN107016362B (en) | Vehicle weight recognition method and system based on vehicle front windshield pasted mark | |
| CN103390167A (en) | Multi-characteristic layered traffic sign identification method | |
| CN105205489A (en) | License plate detection method based on color texture analyzer and machine learning | |
| CN102902957A (en) | Video-stream-based automatic license plate recognition method | |
| CN104463204A (en) | Target quantity statistical method | |
| CN105893960A (en) | Road traffic sign detecting method based on phase symmetry | |
| CN105787470A (en) | Method for detecting power transmission line tower in image based on polymerization multichannel characteristic | |
| CN103902985A (en) | High-robustness real-time lane detection algorithm based on ROI |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20151118 |