CN112418143A - Traffic sign identification method for unmanned vehicle - Google Patents

Traffic sign identification method for unmanned vehicle Download PDF

Info

Publication number
CN112418143A
CN112418143A CN202011405928.XA CN202011405928A CN112418143A CN 112418143 A CN112418143 A CN 112418143A CN 202011405928 A CN202011405928 A CN 202011405928A CN 112418143 A CN112418143 A CN 112418143A
Authority
CN
China
Prior art keywords
image
region
interest
traffic sign
contour
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
Application number
CN202011405928.XA
Other languages
Chinese (zh)
Inventor
王英丽
李晓静
马宏斌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Heilongjiang University
Original Assignee
Heilongjiang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Heilongjiang University filed Critical Heilongjiang University
Priority to CN202011405928.XA priority Critical patent/CN112418143A/en
Publication of CN112418143A publication Critical patent/CN112418143A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G06V20/582—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of traffic signs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration using local operators
    • G06T5/30—Erosion or dilatation, e.g. thinning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30248—Vehicle exterior or interior
    • G06T2207/30252—Vehicle exterior; Vicinity of vehicle

Landscapes

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

Abstract

A traffic sign recognition method for an unmanned vehicle. The invention belongs to the technical field of traffic sign information identification, and aims to solve the problems that a commonly used method for analyzing the shape of a traffic sign is Hough transformation and popularization, the method is complex in calculation, can only detect a round sign and is not strong in practicability, and the method comprises the following steps: preprocessing an original image acquired by a vehicle-mounted camera to obtain a primary region of interest; performing edge connection on the primary region of interest to obtain the outer contour of the primary region of interest, and filling a closed region in the outer contour; all the contours are obtained through a contour processing algorithm, the contours with the areas within the specified area range in all the contours are selected, and convex hull processing is sequentially carried out on the contours; and (3) corresponding each interested contour image to the original identification image in the step 1, thereby identifying the type of the traffic sign to which the interested area belongs. The invention is used for identifying the traffic sign of the unmanned vehicle.

Description

Traffic sign identification method for unmanned vehicle
Technical Field
The invention relates to the technical field of traffic sign information identification, in particular to a traffic sign identification method of an unmanned vehicle.
Background
Real-time perception and identification of driving environment are one of the key problems of unmanned driving of vehicles, and road traffic signs are important components of the driving environment. The invention researches a rapid positioning and identification method of traffic signs in a dynamic scene, detects and identifies the traffic identification information in the driving process of vehicles, and can provide timely road environment information for unmanned vehicles so as to abide by traffic rules. By using the latest research achievements of human visual cognition mechanism, computer vision and pattern recognition theory for reference, the fast, automatic and robust traffic sign recognition system is researched and designed, powerful support is provided for relevant theory and application development of the unmanned vehicle, and the traffic sign recognition system has important theoretical significance and practical value.
The detection and identification research of traffic signs has been carried out for more than 30 years at home and abroad, and a lot of progress is made. The threshold segmentation method is most widely applied to the color space in the detection stage, is simple, has low calculation complexity and can obtain a better segmentation effect. In order to obtain the complete outer contour of the traffic sign, the region of interest after threshold segmentation is expanded to obtain a color image which is equal to the original image in size and has a black background, the color image is grayed, the edges of the gray images are extracted and connected, and the complete outer contour of the traffic sign can be obtained and can adapt to the conditions that the traffic sign is shielded and deformed and the like. The method for analyzing the shape of the traffic sign is Hough transformation and popularization, the method is complex in calculation, only can be used for detecting the circular sign, and the practicability is not strong.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to provide a traffic sign identification method of an unmanned vehicle, so as to overcome the defects in the prior art.
In order to achieve the above object, the present invention provides a traffic sign recognition method for an unmanned vehicle, comprising the steps of:
step 1: preprocessing an original image acquired by a vehicle-mounted camera to obtain a primary region of interest;
the preprocessing comprises RGB color value transformation, threshold segmentation and noise point elimination by morphological filtering;
step 2: performing edge connection on the primary region of interest to obtain an outer contour of the primary region of interest, filling a closed region in the outer contour, and eliminating burrs and noise interference by corrosion and expansion operations to obtain the region of interest;
and step 3: extracting shape marking graphs from the region of interest after edge connection as classification features to realize coarse classification of the traffic signs and obtain color images of candidate traffic signs in different shapes;
step 4, obtaining all contours of the color image through a contour processing algorithm, sequentially calculating the areas of all the contours, selecting the contours with the areas within a specified area range in all the contours, sequentially performing convex hull processing on the contours, and connecting the contours after the convex hull processing into a new contour image, thereby obtaining new contour images of the red channel and the blue channel;
step 5, calculating the ratio of the area and the perimeter of each contour of each new contour image, comparing the ratio with the area perimeter of the traffic sign to screen out the non-conforming contours, then calculating the invariant moment characteristic values of the conforming contours, sequentially calculating Euclidean distances between the invariant moment characteristic value of each conforming contour and the circular invariant moment characteristic value and the square invariant moment characteristic value of the traffic sign, finally comparing the Euclidean distances with the specified distance range of the corresponding shape, and selecting the contour in the corresponding specified distance range as the interested contour image, thereby obtaining the interested contour image of the red channel and the blue channel;
and 6, corresponding each interested contour image to the original identification image in the step 1, obtaining an interested area on the original identification image according to the interested contour image, traversing the standard templates of various traffic signs in the template library, and selecting the standard template of the traffic sign matched with the interested area on the original identification image, thereby identifying the type of the traffic sign of the interested area.
As a further description of the traffic sign recognition method for an unmanned vehicle according to the present invention, preferably, the specific process of step 2 is: (1) marking T with the area larger than the threshold value for the binary image after threshold segmentationARegion Rgeioni，cI ∈ {1, 2, …, N }; let each region Rgeioni，cH of the circumscribed rectangleiHeight and width are each HiAnd WiThe coordinate of the upper left corner is (x)i1，yi1) Setting the height and width of the original image as H and W respectively; to Rgeioni，cEach extending from top to bottom and having a rectangular height 1/4, i.e. Hi/4, 1/4 extending to the left and right, i.e. Wi/4, the coordinates (Ex) of the upper left corner of the expanded circumscribed rectanglei1，Eyi1) And height and width EHsi，EWiRespectively as follows:
Figure BDA0002818513690000021
Figure BDA0002818513690000022
the traffic sign can completely appear in the processed image to obtain a color image Seg with the same size as the original imagex，y，c(ii) a Color image Segx，y，cGraying into Gray, using Canny operator to make edge detection for Gray image Gray, setting background as black, and making the extended interested Region boundary be detected as edge, so that the extended interested Region can be detectedi，cThe boundary is set as a non-edge, so that the influence of the expanded boundary on edge connection can be eliminated;
(2) sit at each marginal pointIn the 8 neighborhoods of the label (x, y), if there is only one point (x)1，y1) If the value of (b) is 1, then the point (x, y) is recorded as an edge point; let a be x-x1，b＝y-y1The following rules were used to grow the edges:
Figure BDA0002818513690000031
adding the growing points into the edge point set for the grown edge point degeneration; if the boundary of the original small region is reached or the growth frequency reaches a preset upper limit, stopping growth; judging whether the point in the edge point set is an edge end point, if so, setting the point as 0 to restore the edge of the area, and if not, setting the point as an error growing point;
(3) obtaining a closed area where the traffic sign is located after the edge connection processing; filling the closed region, and using the binary image erosion and expansion morphological operation with the same structural elements to eliminate the burr and noise interference.
As a further explanation of the traffic sign recognition method for an unmanned vehicle according to the present invention, preferably, the step 6 includes:
step 61, converting the region-of-interest image on the original identification image into a gray image, and performing binarization processing on the gray image after Gaussian filtering to obtain a binarization image corresponding to the region-of-interest;
step 62, the number of columns and the number of rows of the binarized image obtained in step 61 are n and m, the percentage of the element values 255 in each column is calculated to obtain n percentages which are recorded as an array x (n), the percentage of the element values 255 in each column is calculated to obtain m percentages which are recorded as an array y (m), and the array x (n) and the array y (m) are used as the characteristics of the region of interest;
step 63, matching the standard templates of various traffic signs in the template library with the characteristics of the region of interest in sequence to obtain the final matching values of the region of interest and the standard templates of various traffic signs;
and step 64, taking the template type corresponding to the minimum matching value in the final matching values of the standard templates of all the traffic signs as the traffic sign type of the region of interest.
As a further explanation of the traffic sign recognition method for an unmanned vehicle according to the present invention, preferably, the step 63 includes:
step 631, converting the standard templates of various traffic signs into 13/10 times, 12/10 times, 10/10 times, 12/13 times, 11/13 times and 10/13 times of the size of the region of interest in sequence by using a bilinear interpolation method and a fixed step length;
step 632, the number of columns and the number of rows of the binarized image obtained after each size conversion is a, b, and corresponding features are obtained according to the method of step 52, the features are expressed as an array a (a) and an array b (b), and step (m-a)/3 is selected:
when step is 0, x (n) matches a (a): calculating the average covariance of X (n) and A (a) and taking the minimum value as the horizontal histogram matching result by utilizing A (a) to start from X (1) to X (a) matching on X (n), and when step is not equal to 0, calculating the average covariance of each time by moving A (a) on X (n) according to step size and taking the minimum value as the horizontal histogram matching result;
when step is 0, y (m) matches b (b): calculating the average covariance of Y (m) and B (b) by using B (b) to start from Y (1) to Y (b) matching on Y (m) and taking the minimum value as the vertical histogram matching result, and when step ≠ 0, calculating the average covariance of each time by moving B (b) on Y (m) according to step size and taking the minimum value as the vertical histogram matching result;
step 633, taking the average value of the horizontal histogram matching result and the vertical histogram matching result as the final matching result of the size transformation;
step 634, sequentially obtaining the final matching results of each size transformation according to the methods of step 532 and step 533, and taking the minimum value in the final matching results of each size transformation as the final matching value of the region of interest and the standard template of the traffic sign;
and step 635, sequentially obtaining the final matching values of the region of interest and the standard templates of the traffic signs according to the method in the step 534.
As a further description of the traffic sign recognition method for an unmanned vehicle according to the present invention, it is preferable that the binarization processing in step 61 performs binarization by using the average value of the image gray levels plus one sixth of the maximum and minimum difference values of the image gray levels as a segmentation threshold.
The invention has the beneficial effects that:
1. the road traffic sign recognition method for the unmanned vehicle, disclosed by the invention, adopts the shape marking graph as the feature to perform similarity matching, can analyze various traffic sign shapes, and has high feature extraction and matching speed.
2. The method has high recognition rate because the image problem is converted into the contour problem through edge contour extraction, and then the screening is carried out through layer-by-layer screening and convex hull processing, so that the interference information is greatly reduced, the image recognition is carried out on a small part of areas, and the recognition rate is greatly improved. And aiming at the mark identification part, the identification speed is also improved by reducing the image information dimension.
3. A convex hull algorithm can be adopted to process a certain shielding problem, details are ignored, the outline is fuzzified, and the robustness of identification can be improved;
due to the shielding problem, the edge of the traffic sign is lack, for example, some leaves are shielded, the edge of a round traffic sign is not a complete round, and the missing part can be made up through convex shell processing, so that a certain shielding problem is solved. Meanwhile, because we use invariant moment to process later, the first few parameters of the invariant moment mainly focus on the overall shape of the outline, and we can omit the detail part of the edge through convex hull processing, thereby increasing the stability of the characteristic value.
Drawings
FIG. 1 is a hierarchical structure of a traffic sign;
FIG. 2 color thresholding and pre-processed image; (graph a is the original graph, graph b is the binary image (red and blue are superposed) of the traffic sign color segmentation process, graph c is the preprocessed binary image, and graph d is the color image after the traffic sign interesting region image is expanded)
FIG. 3 is an image of a traffic sign region of interest edge junction; (graph a is the region of interest edge extraction and end point marking, graph b is the image after primary edge growth, and graph c is the image after edge degradation operation (d) edge filling and morphological processing);
FIG. 4 is a schematic diagram of a region of interest and a corresponding template in the unmanned vehicle-oriented traffic sign identification method according to the present invention;
FIG. 5 is a schematic diagram of a horizontal and vertical histogram scaling optimal matching algorithm in the unmanned vehicle-oriented traffic sign recognition method of the present invention;
fig. 6 is a schematic diagram of the recognition result of the unmanned vehicle-oriented traffic sign recognition method of the present invention.
Detailed Description
To further understand the structure, characteristics and other objects of the present invention, the following detailed description is given with reference to the accompanying preferred embodiments, which are only used to illustrate the technical solutions of the present invention and are not to limit the present invention.
In a first specific embodiment, a method for identifying a traffic sign of an unmanned vehicle according to this embodiment includes the steps of:
step 1: preprocessing an original image acquired by a vehicle-mounted camera to obtain a primary region of interest;
the preprocessing comprises RGB color value transformation, threshold segmentation and noise point elimination by morphological filtering;
step 2: performing edge connection on the primary region of interest to obtain an outer contour of the primary region of interest, filling a closed region in the outer contour, and eliminating burrs and noise interference by corrosion and expansion operations to obtain the region of interest;
and step 3: extracting shape marking graphs from the region of interest after edge connection as classification features to realize coarse classification of the traffic signs and obtain color images of candidate traffic signs in different shapes;
step 4, obtaining all contours of the color image through a contour processing algorithm, sequentially calculating the areas of all the contours, selecting the contours with the areas within a specified area range in all the contours, sequentially performing convex hull processing on the contours, and connecting the contours after the convex hull processing into a new contour image, thereby obtaining new contour images of the red channel and the blue channel;
step 5, calculating the ratio of the area and the perimeter of each contour of each new contour image, comparing the ratio with the area perimeter of the traffic sign to screen out the non-conforming contours, then calculating the invariant moment characteristic values of the conforming contours, sequentially calculating Euclidean distances between the invariant moment characteristic value of each conforming contour and the circular invariant moment characteristic value and the square invariant moment characteristic value of the traffic sign, finally comparing the Euclidean distances with the specified distance range of the corresponding shape, and selecting the contour in the corresponding specified distance range as the interested contour image, thereby obtaining the interested contour image of the red channel and the blue channel;
and 6, corresponding each interested contour image to the original identification image in the step 1, obtaining an interested area on the original identification image according to the interested contour image, traversing the standard templates of various traffic signs in the template library, and selecting the standard template of the traffic sign matched with the interested area on the original identification image, thereby identifying the type of the traffic sign of the interested area.
In a second embodiment, the present embodiment is a further description of the method for identifying a traffic sign of an unmanned vehicle according to the first embodiment, and the specific process of step 2 is as follows: (1) marking T with the area larger than the threshold value for the binary image after threshold segmentationAreaRegion Rgeioni，cI ∈ {1, 2, …, N }; let each region Rgeioni，cH of the circumscribed rectangleiHeight and width are each HiAnd WiThe coordinate of the upper left corner is (x)i1，yi1) Setting the height and width of the original image as H and W respectively; to Rgeioni，cEach extending from top to bottom and having a rectangular height 1/4, i.e. Hi/4, 1/4 extending to the left and right, i.e. Wi/4, the coordinates (Ex) of the upper left corner of the expanded circumscribed rectanglei1，Eyi1) And height and width EHsi，EWiRespectively as follows:
Figure BDA0002818513690000071
Figure BDA0002818513690000072
the traffic sign can completely appear in the processed image to obtain a color image Seg with the same size as the original imagex，y，c(ii) a Color image Segx，y，cGraying into Gray, using Canny operator to make edge detection for Gray image Gray, setting background as black, and making the extended interested Region boundary be detected as edge, so that the extended interested Region can be detectedi，cThe boundary is set as a non-edge, so that the influence of the expanded boundary on edge connection can be eliminated;
(2) in an 8 neighborhood of each edge point coordinate (x, y), if there is only one point (x)1，y1) If the value of (b) is 1, then the point (x, y) is recorded as an edge point; let a be x-x1，b＝y-y1The following rules were used to grow the edges:
Figure BDA0002818513690000073
adding the growing points into the edge point set for the grown edge point degeneration; if the boundary of the original small region is reached or the growth frequency reaches a preset upper limit, stopping growth; judging whether the point in the edge point set is an edge end point, if so, setting the point as 0 to restore the edge of the area, and if not, setting the point as an error growing point;
(3) obtaining a closed area where the traffic sign is located after the edge connection processing; filling the closed region, and using the binary image erosion and expansion morphological operation with the same structural elements to eliminate the burr and noise interference.
In a third embodiment, the present embodiment is further directed to the method for identifying a traffic sign of an unmanned vehicle according to the first embodiment, and the step 6 includes:
step 61, converting the region-of-interest image on the original identification image into a gray image, and performing binarization processing on the gray image after Gaussian filtering to obtain a binarization image corresponding to the region-of-interest;
step 62, the number of columns and the number of rows of the binarized image obtained in step 61 are n and m, the percentage of the element values 255 in each column is calculated to obtain n percentages which are recorded as an array x (n), the percentage of the element values 255 in each column is calculated to obtain m percentages which are recorded as an array y (m), and the array x (n) and the array y (m) are used as the characteristics of the region of interest;
step 63, matching the standard templates of various traffic signs in the template library with the characteristics of the region of interest in sequence to obtain the final matching values of the region of interest and the standard templates of various traffic signs;
and step 64, taking the template type corresponding to the minimum matching value in the final matching values of the standard templates of all the traffic signs as the traffic sign type of the region of interest.
In a fourth embodiment, the present embodiment is further directed to the method for identifying a traffic sign of an unmanned vehicle according to the first embodiment, and the step 63 includes:
step 631, converting the standard templates of various traffic signs into 13/10 times, 12/10 times, 10/10 times, 12/13 times, 11/13 times and 10/13 times of the size of the region of interest in sequence by using a bilinear interpolation method and a fixed step length;
step 632, the number of columns and the number of rows of the binarized image obtained after each size conversion is a, b, and corresponding features are obtained according to the method of step 52, the features are expressed as an array a (a) and an array b (b), and step (m-a)/3 is selected:
when step is 0, x (n) matches a (a): calculating the average covariance of X (n) and A (a) and taking the minimum value as the horizontal histogram matching result by utilizing A (a) to start from X (1) to X (a) matching on X (n), and when step is not equal to 0, calculating the average covariance of each time by moving A (a) on X (n) according to step size and taking the minimum value as the horizontal histogram matching result;
when step is 0, y (m) matches b (b): calculating the average covariance of Y (m) and B (b) by using B (b) to start from Y (1) to Y (b) matching on Y (m) and taking the minimum value as the vertical histogram matching result, and when step ≠ 0, calculating the average covariance of each time by moving B (b) on Y (m) according to step size and taking the minimum value as the vertical histogram matching result;
step 633, taking the average value of the horizontal histogram matching result and the vertical histogram matching result as the final matching result of the size transformation;
step 634, sequentially obtaining the final matching results of each size transformation according to the methods of step 532 and step 533, and taking the minimum value in the final matching results of each size transformation as the final matching value of the region of interest and the standard template of the traffic sign;
and step 635, sequentially obtaining the final matching values of the region of interest and the standard templates of the traffic signs according to the method in the step 534.
In a fifth embodiment, the present embodiment is further directed to the method for identifying a traffic sign of an unmanned vehicle according to the first embodiment, wherein the binarization processing in the step 61 performs binarization by using a sixth of the average value of the image gray scale plus the maximum and minimum difference value of the image gray scale as a segmentation threshold.
It should be noted that the above summary and the detailed description are intended to demonstrate the practical application of the technical solutions provided by the present invention, and should not be construed as limiting the scope of the present invention. Various modifications, equivalent substitutions, or improvements may be made by those skilled in the art within the spirit and principles of the invention. The scope of the invention is to be determined by the appended claims.

Claims (5)

1. A traffic sign recognition method for an unmanned vehicle is characterized by comprising the following steps:
step 1: preprocessing an original image acquired by a vehicle-mounted camera to obtain a primary region of interest;
the preprocessing comprises RGB color value transformation, threshold segmentation and noise point elimination by morphological filtering;
step 2: performing edge connection on the primary region of interest to obtain an outer contour of the primary region of interest, filling a closed region in the outer contour, and eliminating burrs and noise interference by corrosion and expansion operations to obtain the region of interest;
and step 3: extracting shape marking graphs from the region of interest after edge connection as classification features to realize coarse classification of the traffic signs and obtain color images of candidate traffic signs in different shapes;
step 4, obtaining all contours of the color image through a contour processing algorithm, sequentially calculating the areas of all the contours, selecting the contours with the areas within a specified area range in all the contours, sequentially performing convex hull processing on the contours, and connecting the contours after the convex hull processing into a new contour image, thereby obtaining new contour images of the red channel and the blue channel;
step 5, calculating the ratio of the area and the perimeter of each contour of each new contour image, comparing the ratio with the area perimeter of the traffic sign to screen out the non-conforming contours, then calculating the invariant moment characteristic values of the conforming contours, sequentially calculating Euclidean distances between the invariant moment characteristic value of each conforming contour and the circular invariant moment characteristic value and the square invariant moment characteristic value of the traffic sign, finally comparing the Euclidean distances with the specified distance range of the corresponding shape, and selecting the contour in the corresponding specified distance range as the interested contour image, thereby obtaining the interested contour image of the red channel and the blue channel;
and 6, corresponding each interested contour image to the original identification image in the step 1, obtaining an interested area on the original identification image according to the interested contour image, traversing the standard templates of various traffic signs in the template library, and selecting the standard template of the traffic sign matched with the interested area on the original identification image, thereby identifying the type of the traffic sign of the interested area.
2. The method for recognizing the traffic sign of the unmanned vehicle as claimed in claim 1, wherein the specific process of the step 2 is as follows: (1) marking T with the area larger than the threshold value for the binary image after threshold segmentationARegion Rgeioni，cI ∈ {1, 2, …, N }; let each region Rgeioni，cH of the circumscribed rectangleiHeight and width are each HiAnd WiThe coordinate of the upper left corner is (x)i1，yi1) Setting the height and width of the original image as H and W respectively; to Rgeioni，cEach extending from top to bottom and having a rectangular height 1/4, i.e. Hi/4, 1/4 extending to the left and right, i.e. Wi/4, the coordinates (Ex) of the upper left corner of the expanded circumscribed rectanglei1，Eyi1) And height and width EHsi，EWiRespectively as follows:
Figure FDA0002818513680000021
Figure FDA0002818513680000022
the traffic sign can completely appear in the processed image to obtain a color image Seg with the same size as the original imagex，y，c(ii) a Color image Segx，y，cGraying into Gray, using Canny operator to make edge detection for Gray image Gray, setting background as black, and making the extended interested Region boundary be detected as edge, so that the extended interested Region can be detectedi，cThe boundary is set as a non-edge, so that the influence of the expanded boundary on edge connection can be eliminated;
(2) in an 8 neighborhood of each edge point coordinate (x, y), if there is only one point (x)1，y1) If the value of (b) is 1, then the point (x, y) is recorded as an edge point; let a be x-x1，b＝y-y1The following rules were used to grow the edges:
Figure FDA0002818513680000023
adding the growing points into the edge point set for the grown edge point degeneration; if the boundary of the original small region is reached or the growth frequency reaches a preset upper limit, stopping growth; judging whether the point in the edge point set is an edge end point, if so, setting the point as 0 to restore the edge of the area, and if not, setting the point as an error growing point;
(3) obtaining a closed area where the traffic sign is located after the edge connection processing; filling the closed region, and using the binary image erosion and expansion morphological operation with the same structural elements to eliminate the burr and noise interference.
3. The method for recognizing a traffic sign of an unmanned vehicle as claimed in claim 2, wherein the step 6 comprises:
step 61, converting the region-of-interest image on the original identification image into a gray image, and performing binarization processing on the gray image after Gaussian filtering to obtain a binarization image corresponding to the region-of-interest;
step 62, the number of columns and the number of rows of the binarized image obtained in step 61 are n and m, the percentage of the element values 255 in each column is calculated to obtain n percentages which are recorded as an array x (n), the percentage of the element values 255 in each column is calculated to obtain m percentages which are recorded as an array y (m), and the array x (n) and the array y (m) are used as the characteristics of the region of interest;
step 63, matching the standard templates of various traffic signs in the template library with the characteristics of the region of interest in sequence to obtain the final matching values of the region of interest and the standard templates of various traffic signs;
and step 64, taking the template type corresponding to the minimum matching value in the final matching values of the standard templates of all the traffic signs as the traffic sign type of the region of interest.
4. A method as claimed in claim 3, wherein said step 63 comprises:
step 631, converting the standard templates of various traffic signs into 13/10 times, 12/10 times, 10/10 times, 12/13 times, 11/13 times and 10/13 times of the size of the region of interest in sequence by using a bilinear interpolation method and a fixed step length;
step 632, the number of columns and the number of rows of the binarized image obtained after each size conversion is a, b, and corresponding features are obtained according to the method of step 52, the features are expressed as an array a (a) and an array b (b), and step (m-a)/3 is selected:
when step is 0, x (n) matches a (a): calculating the average covariance of X (n) and A (a) and taking the minimum value as the horizontal histogram matching result by utilizing A (a) to start from X (1) to X (a) matching on X (n), and when step is not equal to 0, calculating the average covariance of each time by moving A (a) on X (n) according to step size and taking the minimum value as the horizontal histogram matching result;
when step is 0, y (m) matches b (b): calculating the average covariance of Y (m) and B (b) by using B (b) to start from Y (1) to Y (b) matching on Y (m) and taking the minimum value as the vertical histogram matching result, and when step ≠ 0, calculating the average covariance of each time by moving B (b) on Y (m) according to step size and taking the minimum value as the vertical histogram matching result;
step 633, taking the average value of the horizontal histogram matching result and the vertical histogram matching result as the final matching result of the size transformation;
step 634, sequentially obtaining the final matching results of each size transformation according to the methods of step 532 and step 533, and taking the minimum value in the final matching results of each size transformation as the final matching value of the region of interest and the standard template of the traffic sign;
and step 635, sequentially obtaining the final matching values of the region of interest and the standard templates of the traffic signs according to the method in the step 534.
5. The method as claimed in claim 4, wherein the binarization processing in the step 61 binarizes with the average value of the image gray scale plus one sixth of the maximum and minimum difference value of the image gray scale as a segmentation threshold.
CN202011405928.XA 2020-12-04 2020-12-04 Traffic sign identification method for unmanned vehicle Pending CN112418143A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011405928.XA CN112418143A (en) 2020-12-04 2020-12-04 Traffic sign identification method for unmanned vehicle

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011405928.XA CN112418143A (en) 2020-12-04 2020-12-04 Traffic sign identification method for unmanned vehicle

Publications (1)

Publication Number Publication Date
CN112418143A true CN112418143A (en) 2021-02-26

Family

ID=74830195

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011405928.XA Pending CN112418143A (en) 2020-12-04 2020-12-04 Traffic sign identification method for unmanned vehicle

Country Status (1)

Country Link
CN (1) CN112418143A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113343976A (en) * 2021-05-13 2021-09-03 武汉大学 Anti-highlight interference engineering measurement mark extraction method based on color-edge fusion feature growth

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007183846A (en) * 2006-01-10 2007-07-19 Honda Motor Co Ltd Sign recognition apparatus for vehicle, and vehicle controller
CN102542260A (en) * 2011-12-30 2012-07-04 中南大学 Method for recognizing road traffic sign for unmanned vehicle
CN103971128A (en) * 2014-05-23 2014-08-06 北京理工大学 Traffic sign recognition method for driverless car
CN104732211A (en) * 2015-03-19 2015-06-24 杭州电子科技大学 Traffic sign detecting method based on self-adaptation threshold value

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007183846A (en) * 2006-01-10 2007-07-19 Honda Motor Co Ltd Sign recognition apparatus for vehicle, and vehicle controller
CN102542260A (en) * 2011-12-30 2012-07-04 中南大学 Method for recognizing road traffic sign for unmanned vehicle
CN103971128A (en) * 2014-05-23 2014-08-06 北京理工大学 Traffic sign recognition method for driverless car
CN104732211A (en) * 2015-03-19 2015-06-24 杭州电子科技大学 Traffic sign detecting method based on self-adaptation threshold value

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
邬思奇: "层次轮廓计算机视角的交通道路标识别", 《无线互联科技》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113343976A (en) * 2021-05-13 2021-09-03 武汉大学 Anti-highlight interference engineering measurement mark extraction method based on color-edge fusion feature growth

Similar Documents

Publication Publication Date Title
CN107316031B (en) Image feature extraction method for pedestrian re-identification
Patel et al. Automatic number plate recognition system (anpr): A survey
CN109816652B (en) Complex casting defect identification method based on gray level significance
CN102629322B (en) Character feature extraction method based on stroke shape of boundary point and application thereof
Paunwala et al. A novel multiple license plate extraction technique for complex background in Indian traffic conditions
CN107563380A (en) A kind of vehicle license plate detection recognition method being combined based on MSER and SWT
Peng et al. Recognition of low-resolution logos in vehicle images based on statistical random sparse distribution
CN112819094A (en) Target detection and identification method based on structural similarity measurement
CN107256398B (en) Feature fusion based individual milk cow identification method
CN112131963B (en) Road identification line extraction method based on driving direction structural feature constraint
CN115797813B (en) Water environment pollution detection method based on aerial image
CN108399644A (en) A kind of wall images recognition methods and its device
CN105701490A (en) Container number adaptive positioning method based on image entropy
CN112861654A (en) Famous tea picking point position information acquisition method based on machine vision
CN110516666B (en) License plate positioning method based on combination of MSER and ISODATA
CN109271882B (en) Method for extracting color-distinguished handwritten Chinese characters
CN116052152A (en) License plate recognition system based on contour detection and deep neural network
CN112418143A (en) Traffic sign identification method for unmanned vehicle
Fang et al. Towards real-time crack detection using a deep neural network with a Bayesian fusion algorithm
CN112418210B (en) Intelligent classification method for tower inspection information
Lin et al. License plate location method based on edge detection and mathematical morphology
CN109753880B (en) Detection and identification method for natural scene vehicle-mounted video road speed limit sign
CN112435272A (en) High-voltage transmission line connected domain removing method based on image contour analysis
Rastegar et al. An intelligent control system using an efficient License Plate Location and Recognition Approach
Huang Research on license plate image segmentation and intelligent character recognition

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210226