CN109886168B - Ground traffic sign identification method based on hierarchy - Google Patents
Ground traffic sign identification method based on hierarchy Download PDFInfo
- Publication number
- CN109886168B CN109886168B CN201910102556.4A CN201910102556A CN109886168B CN 109886168 B CN109886168 B CN 109886168B CN 201910102556 A CN201910102556 A CN 201910102556A CN 109886168 B CN109886168 B CN 109886168B
- Authority
- CN
- China
- Prior art keywords
- image
- lum
- value
- lane line
- pixel values
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a ground traffic sign identification method based on hierarchy, which comprises the steps of firstly analyzing and processing a video image, screening the size and the angle of an extracted rectangular area through lane line detection, detecting and tracking lane lines in the video image to obtain stable and accurate lane lines, and intercepting an interested area; then, a self-adaptive binarization technology is adopted to define pixel values of the gray level image, all pixel points are divided into n levels, the pixel values of the uppermost layer and the lowermost layer are omitted to obtain effective pixel values, and target detection is respectively carried out on the pixel point clustering areas of each divided level to realize the identification of the traffic signs. The invention has the advantages of accuracy and high efficiency, and the threshold calculated based on the hierarchy can highlight the required lane line part, thereby being beneficial to the identification of the ground traffic sign.
Description
Technical Field
The invention relates to the field of computer vision and intelligent driving, in particular to a ground traffic sign identification method based on hierarchy.
Background
As an important research content in ITS (intelligent transportation system), traffic sign recognition technology is indispensable, and over the years, with the rapid development of related technologies such as computer vision, pattern recognition and artificial intelligence, the traffic sign recognition technology has attracted more and more attention and has become an important research content in the field of computer vision, wherein traffic sign detection based on video stream is a research hotspot in the fields of computer vision and pattern recognition.
However, at present, domestic research on intelligent traffic systems is still a preliminary stage, research on the aspects of traffic sign detection, recognition and 3D display is less, and the recognition difficulty is increased due to the fact that the current traffic road conditions are more complex, so that the theme of the ground traffic sign detection method selected at this time is to use the more effective recognition of traffic signs as the departure point and the footfall point.
Most of the current adaptive binarization methods are otsu (maximum inter-class difference) algorithm and well ner (fast adaptive image binarization) algorithm, which have good effect under the condition of uncomplicated environment, but have great limitation on urban roads, are easily interfered by environment, weather conditions and the like, and cannot effectively and prominently detect the target. The traditional target area distinguishing method has the delay influence on the detection result due to long time consumption, the problems are effectively solved by the target area distinguishing technology based on the hierarchy level, and the binarization effect is good under the condition of poor environment. And an approximate region of the target can be quickly found through the hierarchy, other interference factors in the region of interest are eliminated, and the requirement can be better met.
Disclosure of Invention
The purpose of the invention is as follows: the invention provides a floor traffic sign identification method based on hierarchy, which can be used for quickly detecting and identifying a floor traffic sign.
The technical scheme is as follows: the invention relates to a ground traffic sign identification method based on hierarchy, which comprises the following steps:
(1) preprocessing a video image to obtain an image subjected to noise reduction;
(2) taking the preprocessed image as input, obtaining a rectangular region in the image by using a feature extraction algorithm, and extracting and detecting lane lines by screening the rectangular region;
(3) taking an image obtained after lane line extraction as input, obtaining the edge of a lane line by using an edge detection algorithm, then obtaining a set of lane line points by using a Hough line detection algorithm, dividing the set of left and right lane line points by using the slope of the midpoint of the set, respectively fitting the sets of the two points into two optimal straight lines, realizing the detection of the lane line, continuously updating the sets of the points, and realizing the tracking of the lane line;
(4) according to the identified lane lines, carrying out inverse perspective transformation on the regions among the lane lines, and extracting the regions for identifying the ground traffic marks;
(5) layering the images through the standard deviation of the image pixel values, removing the pixel values of the uppermost layer and the lowermost layer, carrying out logarithm averaging on the rest pixel values, then obtaining the natural index value of the average value, eliminating the interference of the weather environment on the binarization effect, and highlighting the lane line part;
(6) and establishing a model for the collected lane line samples by a deep learning method, identifying the pixel point gathering area of each hierarchy by the model, and locking the target area.
The video image preprocessing method in the step (1) mainly comprises the following steps: graying an image, removing image noise, defining a region of interest, binarizing the image, performing morphological operation and Gaussian blurring.
The screening of the rectangular area in the step (2) is realized by the following steps: defining a maximum range and a minimum range of a rectangular area, acquiring sine function values of the size of the rectangular area and the angle of the rectangular area, defining a maximum value and a minimum value of a positive line function, and selecting the rectangular area which retains the following values as 1:
wherein size is the size of the rectangular region, MaxSize is the maximum area, MinSize is the minimum area, angle is the sine function value of the angle of the rectangular region, maxngle is the maximum value, and MinAngle is the minimum value.
The step (5) comprises the following steps:
(51) defining pixel values Lum (x, y) of a gray image, wherein the standard deviation of all the pixel values is Sd (x, y), dividing all pixel points into N levels, and N is the number of the pixel points;
(52) the pixel values of the uppermost layer and the lowermost layer are cut off to obtain the effective pixel value Lum (x, y)effThe method is realized by the following formula:
Lum(x,y)eff＝Lum(x,y),Lum(x,y)min+Sd(x,y)＜Lum(x,y)＜Lum(x,y)max-Sd(x,y)；
(53) find the Lum (x, y)effNatural logarithm of (d), for all Lum (x, y)effThe logarithm of the binary threshold is calculated to obtain an average value, and then a natural index value of the average value, namely the searched binary threshold is calculated;
(54) and (4) substituting the threshold value obtained in the step (53) into a threshold function of the opencv self-carrying region, so as to obtain a region-of-interest binary image with a good effect.
And (4) the model in the step (6) is a trained support vector machine model.
Has the advantages that: compared with the prior art, the invention has the beneficial effects that: the method has the advantages of accuracy and high efficiency, the threshold calculated based on the hierarchy can highlight the required lane line part, the condition of large-area whitening or large-area blackening of other algorithms cannot occur, and the identification of ground traffic signs is facilitated; and the approximate region can be found more quickly by identifying the pixel point gathering region of each layer through a deep learning model compared with a detectMultiScale (thresholding) function of the opencv self-band.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a graph of the effect of lane lines extracted using an adaptive threshold segmentation method based on a Gaussian distribution model;
FIG. 3 is a diagram showing the effect of lane line tracking after screening straight lines using perspective transformation;
FIG. 4 is a diagram of adaptive binarization effects based on hierarchy;
fig. 5 is a diagram showing the effect of discriminating the hierarchical ground traffic sign area.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
As shown in fig. 1, the invention discloses a method for identifying a ground traffic sign based on a hierarchy, which comprises the following steps:
1. and preprocessing the acquired video image to provide support for subsequent image processing.
The video data used in the embodiment is the video data which is acquired by a vehicle-mounted camera when a motor coach drives on a highway and is shot at a first visual angle of a driver, and an original video image contains much information and noise which are irrelevant to subsequent image processing operation, so that image preprocessing is firstly carried out, and the processing steps mainly comprise image graying and image noise removal. Since the RGB three-channel image does not help lane line segmentation, feature extraction and related operation, firstly, graying processing is carried out, the RGB three-channel image is converted into a single-channel grayscale image, and then, a morphological filtering method is used for removing salt and pepper noise in the image to obtain a smoother image, which is beneficial to extracting accurate and reliable lane lines.
2. Taking a preprocessed image as an input, firstly, separating the lane line from a complex background by using an adaptive threshold segmentation method based on a Gaussian distribution model. Fig. 2 shows the effect of the lane lines separated. Then Canny (the optimal edge detection algorithm) is selected to detect the edge of the lane line, and a proper rectangular area is selected through feature extraction to screen out the correct lane line. Defining a maximum range (MaxSize) and a minimum range (MinSize) of a rectangular area, acquiring sine function values (angle) of the size (size) of the rectangular area and the angle of the rectangular area, defining a maximum value (MaxAngle) and a minimum value (MinAngle) of a positive line function, selecting the rectangular area which retains the following formula and is 1, and screening by the following formula:
in the formula, size is the size of the rectangular region, MaxSize is the maximum area, MinSize is the minimum area, angle is the sine function value of the angle of the rectangular region, maxngle is the maximum value, and MinAngle is the minimum value.
3. And taking the screened lane line image as input, extracting the feature points of the lane by using a random Hough transform method, then constructing a mathematical model to perform straight line fitting on the feature points of the lane line, and then extracting a target region by using inverse perspective transform. And finally, tracking the lane line and extracting a target area. And finally, the optimization algorithm ensures the stability and effectiveness of the lane line detection.
4. The regions between the lane lines are subjected to inverse perspective transformation according to the recognized lane lines, and regions for ground traffic sign recognition are extracted, as shown in fig. 3.
5. Defining an adaptive binary method: defining pixel values Lum (x, y) of the gray image, wherein the standard deviation of all the pixel values is Sd (x, y), dividing all the pixel points into N levels, and N is the number of the pixel points. The pixel values of the uppermost layer and the lowermost layer are cut off to obtain the effective pixel value Lum (x, y)effThe formula is as follows:
Lum(x,y)eff＝Lum(x,y),Lum(x,y)min+Sd(x,y)＜Lum(x,y)＜Lum(x,y)max-Sd (x, y); then, the Lum (x, y) is determinedeffThe natural logarithm of (d); then all Lum (x, y)effThe logarithm of the level is averaged, then the natural index value of the average is calculated, the obtained result is the binary threshold value required to be searched, then the threshold value is substituted into the threshold function of the opencv, and the adaptive binary effect diagram based on the hierarchy is shown in fig. 4.
6. And (3) respectively carrying out target detection on each layer-level pixel point clustering region classified in the step (5) by utilizing a method for locking a layer-level target through a trained SVM model, wherein the effect of target region discrimination based on the layers is shown in FIG. 5, when the target is detected, carrying out voice prompt on a driver, and if the target is not detected, continuing to detect the next layer.
Claims (5)
1. A ground traffic sign identification method based on hierarchy is characterized by comprising the following steps:
(1) preprocessing a video image to obtain an image subjected to noise reduction;
(2) taking the preprocessed image as input, obtaining a rectangular region in the image by using a feature extraction algorithm, and extracting and detecting lane lines by screening the rectangular region;
(3) taking an image obtained after lane line extraction as input, obtaining the edge of a lane line by using an edge detection algorithm, then obtaining a set of lane line points by using a Hough line detection algorithm, dividing the set of left and right lane line points by using the slope of the midpoint of the set, respectively fitting the sets of the two points into two optimal straight lines, realizing the detection of the lane line, continuously updating the sets of the points, and realizing the tracking of the lane line;
(4) according to the identified lane lines, carrying out inverse perspective transformation on the regions among the lane lines, and extracting the regions for identifying the ground traffic signs;
(5) layering the images through the standard deviation of the image pixel values, removing the pixel values of the uppermost layer and the lowermost layer, carrying out logarithm averaging on the rest pixel values, then obtaining the natural index value of the average value, eliminating the interference of the weather environment on the binarization effect, and highlighting the lane line part;
(6) and establishing a model for the collected lane line samples by a deep learning method, identifying the pixel point gathering area of each hierarchy by the model, and locking the target area.
2. The method for recognizing a ground traffic sign based on hierarchy according to claim 1, wherein the preprocessing of the video image in step (1) mainly comprises: graying an image, removing image noise, defining a region of interest, binarizing the image, performing morphological operation and Gaussian blurring.
3. The method for recognizing ground traffic signs based on hierarchy according to claim 1, wherein the step (2) of filtering the rectangular areas is implemented by: defining a maximum range and a minimum range of a rectangular area, acquiring a sine function value of the size of the rectangular area and the angle of the rectangular area, defining a maximum value and a minimum value of the sine function, and selecting the rectangular area which retains the following values as 1:
wherein size is the size of the rectangular region, MaxSize is the maximum area, MinSize is the minimum area, angle is the sine function value of the angle of the rectangular region, maxngle is the maximum value, and MinAngle is the minimum value.
4. The method for identifying a level-based ground traffic sign according to claim 1, wherein the step (5) comprises the steps of:
(51) defining pixel values Lum (x, y) of a gray image, wherein the standard deviation of all the pixel values is Sd (x, y), and dividing all pixel points into n levels;
(52) the pixel values of the uppermost layer and the lowermost layer are cut off to obtain the effective pixel value Lum (x, y)effThe method is realized by the following formula:
Lum(x,y)eff＝Lum(x,y),Lum(x,y)min+Sd(x,y)＜Lum(x,y)＜Lum(x,y)max-Sd(x,y)
wherein, Lum (x, y)minIs the minimum pixel value of the gray image, Lum (x, y)maxIs the maximum pixel value of the gray scale image;
(53) find the Lum (x, y)effNatural logarithm of (d), for all Lum (x, y)effThe logarithm of the binary threshold is calculated to obtain an average value, and then a natural index value of the average value, namely the searched binary threshold is calculated;
(54) and (4) substituting the threshold value obtained in the step (53) into a threshold function of the opencv self-carrying region, so as to obtain a region-of-interest binary image with a good effect.
5. The method for recognizing a level-based ground traffic sign according to claim 1, wherein the model in step (6) is a trained support vector machine model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910102556.4A CN109886168B (en) | 2019-02-01 | 2019-02-01 | Ground traffic sign identification method based on hierarchy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910102556.4A CN109886168B (en) | 2019-02-01 | 2019-02-01 | Ground traffic sign identification method based on hierarchy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109886168A CN109886168A (en) | 2019-06-14 |
| CN109886168B true CN109886168B (en) | 2020-10-16 |
Family
ID=66927878
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910102556.4A Active CN109886168B (en) | 2019-02-01 | 2019-02-01 | Ground traffic sign identification method based on hierarchy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109886168B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111476157B (en) * | 2020-04-07 | 2020-11-03 | 南京慧视领航信息技术有限公司 | Lane guide arrow recognition method under intersection monitoring environment |
| CN112115778B (en) * | 2020-08-11 | 2023-07-21 | 华南理工大学 | Intelligent lane line identification method under ring simulation condition |
| CN113269800B (en) * | 2021-05-25 | 2023-05-12 | 华能海南发电股份有限公司澄迈光伏电站 | Infrared image segmentation method for solar photovoltaic module |
| CN115294767B (en) * | 2022-08-02 | 2023-06-27 | 山西省智慧交通研究院有限公司 | Real-time detection and traffic safety early warning method and device for expressway lane line |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103425981B (en) * | 2012-05-22 | 2017-09-15 | 深圳中兴力维技术有限公司 | A kind of agglomerate extraction element and method based on cluster |
| CN104021553B (en) * | 2014-05-30 | 2016-12-07 | 哈尔滨工程大学 | A kind of sonar image object detection method based on pixel layering |
| CN105893949B (en) * | 2016-03-29 | 2019-07-12 | 西南交通大学 | A kind of method for detecting lane lines under complex road condition scene |
| CN106407924A (en) * | 2016-09-10 | 2017-02-15 | 上海大学 | Binocular road identifying and detecting method based on pavement characteristics |
| CN108573242A (en) * | 2018-04-26 | 2018-09-25 | 南京行车宝智能科技有限公司 | A kind of method for detecting lane lines and device |
-
2019
- 2019-02-01 CN CN201910102556.4A patent/CN109886168B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN109886168A (en) | 2019-06-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105373794B (en) | A kind of licence plate recognition method | |
| Marzougui et al. | A lane tracking method based on progressive probabilistic Hough transform | |
| CN109784344B (en) | Image non-target filtering method for ground plane identification recognition | |
| CN101334836B (en) | License plate positioning method incorporating color, size and texture characteristic | |
| Mammeri et al. | A real-time lane marking localization, tracking and communication system | |
| CN109657632B (en) | Lane line detection and identification method | |
| CN109886168B (en) | Ground traffic sign identification method based on hierarchy | |
| CN102509098B (en) | Fisheye image vehicle identification method | |
| CN107301405A (en) | Method for traffic sign detection under natural scene | |
| Alonso et al. | Robust vehicle detection through multidimensional classification for on board video based systems | |
| CN105528609A (en) | Vehicle license plate location method based on character position | |
| CN107895151A (en) | Method for detecting lane lines based on machine vision under a kind of high light conditions | |
| WO2023279966A1 (en) | Multi-lane-line detection method and apparatus, and detection device | |
| CN102842037A (en) | Method for removing vehicle shadow based on multi-feature fusion | |
| Qin et al. | A general framework for road marking detection and analysis | |
| Abedin et al. | Traffic sign recognition using surf: Speeded up robust feature descriptor and artificial neural network classifier | |
| CN114511770A (en) | Road sign plate identification method | |
| CN113239733B (en) | Multi-lane line detection method | |
| CN107133588A (en) | Vehicle identification method based on vehicle window feature extraction | |
| CN111027544A (en) | MSER license plate positioning method and system based on visual saliency detection | |
| Avery et al. | Investigation into shadow removal from traffic images | |
| CN112101108A (en) | Left-right-to-pass sign identification method based on pole position characteristics of graph | |
| CN107977608B (en) | Method for extracting road area of highway video image | |
| Salahshoor et al. | Application of intelligent systems for iranian license plate recognition | |
| Lu | A lane detection, tracking and recognition system for smart vehicles |
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 |