Traffic sign detection method based on classification template matching

A traffic sign, template matching technology, applied in character and pattern recognition, instruments, computer parts and other directions, can solve the problems of high light and image quality requirements, complex operation process, poor real-time performance, etc., to achieve high real-time performance and accuracy , reduce interference information, good real-time effect

Inactive Publication Date: 2012-12-12
NO 8357 RES INST OF THE THIRD ACADEMY OF CHINA AEROSPACE SCI & IND
5 Cites 29 Cited by

AI-Extracted Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is that the existing traffic sign recognition method has complex operation process, poor r...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Abstract

The invention belongs to the technical field of machine vision and image processing and in particular relates to a traffic sign detection method based on classification template matching. The problems that the conventional traffic sign identification method is complex in operation process, low in real-time property and high in light and image quality requirements are solved. The method comprises the following steps of: segmenting areas containing traffic signs in a photographed image according to different color areas, namely classifying the traffic signs according to colors; (2) screening communicating areas in the image through the shape and area characteristics after the color classification, and positioning the sign area; and (3) identifying through a template matching method. The method has high operability and is insensitive to the light change, and when a source image is blurred, a high identification accuracy rate can be guaranteed, and the method is high in real-time property and can be used for an automatic traffic sign detection device on a vehicle.

Application Domain

Technology Topic

Template matchingTraffic sign detection +8

Image

  • Traffic sign detection method based on classification template matching
  • Traffic sign detection method based on classification template matching
  • Traffic sign detection method based on classification template matching

Examples

  • Experimental program(1)

Example Embodiment

[0039] The following describes a traffic sign detection method based on classification template matching provided by the present invention with reference to the accompanying drawings and embodiments:
[0040] Such as figure 1 As shown, a traffic sign detection method based on classification template matching includes the following steps:
[0041] (1) Segment the different color areas of the region of interest, and classify the traffic signs by color.
[0042] Since the captured image usually contains not only the traffic signs but also the road scenery and other information, but the effective and recognizable traffic signs always appear in a certain area of ​​the captured image, so set the area of ​​interest in the image; and the common signs can be pressed The colors are divided into red prohibition signs, blue indicator signs, yellow warning signs, and black signs (such as speed limit removal signs). The first step of classification can be carried out by color in turn.
[0043] When performing color classification, the following existing models are often used: YUV model, YIQ model, Lab model, etc. For better classification, the present invention preferably uses a rapid conversion method. The color of traffic signs is clearly distinguished in HSV space, so to convert rgb image to hsv image, let (r, g, b) be the red, green and blue coordinates of a color, and their values ​​are between 0 and 1. Real number between. Let max be equivalent to the largest of r, g, and b. Let min be equal to the smallest of these values. To find the (h, s, v) value in the HSV space, where h ∈ [0, 360) is the hue angle of the angle, and s, v ∈ [0, 1] is the saturation and brightness, the calculation formula is :
[0044]
[0045] s = 0 , if max = 0 max - min max otherwise
[0046] v=max
[0047] h is hue, each color corresponds to an h value; s is saturation, bright color saturation is high, white saturation is close to zero; v is brightness, this value has a large relationship with light intensity. In order to preserve as many areas of each color as possible, the preprocessing sets wide h, s, and v intervals for each color segmentation.
[0048] (2) Positioning of traffic signs
[0049] After color classification, the connected areas in the image are screened by features such as shape and area to locate traffic signs. Such as figure 1 As shown, the red and black segmented images search for circular areas, the blue segmented images search for circular filled areas, and the yellow image search for triangular areas.
[0050] The invention uses the pixel marking method to search for connected areas, traverses all the obtained areas, judges whether it is a traffic sign according to the area characteristics, and if so, draws the circumscribed rectangle of the connected area to locate the sign area.
[0051] One or more of the following features can be used when determining the shape:
[0052] a. Area S: Area is one of the most basic characteristics of an image. The area S of an image can be expressed by the number of pixels contained in the same marked area.
[0053] b. Circumference L
[0054] The perimeter L of the image is represented by the sum of the distances between two adjacent pixels on the outer boundary of the image.
[0055] c. Roundness R o , Inscribed circle radius r and morphological complexity e
[0056] Roundness R o It is used to describe how close the shape of a scene is to a circle, and its calculation formula is
[0057] R o = 4 πS L 2
[0058] Where S is the area of ​​the figure, L is the perimeter of the figure, R o The value range is 0≤R o ≤1, R o The larger the figure, the closer the figure is to a circle. Taking continuous circles, squares, and regular triangles as examples, calculate their circularity R o For: round, R o =1; square, R o =0.79; equilateral triangle, R o = 0.6.
[0059] The inscribed circle radius r is expressed by the following formula:
[0060] r = 2 S L
[0061] In the formula, S and L have the same meaning as above. Similarly, taking three typical continuous graphics as an example, the calculation results are: circle, r=R (R is the radius of the circle); square, (a is the side length of a square); equilateral triangle, (a is the side length of a regular triangle).
[0062] The shape complexity index is usually expressed by the discrete index e, and its calculation formula is:
[0063] e = L 2 S
[0064] This formula describes the size of the perimeter of the graph per unit area. A large e value indicates that the perimeter of the unit area is large, that is, if the graph is discrete, it is a complex graph; otherwise, it is a simple graph. The graph with the smallest e value is a circle. The calculation result of a typical continuous figure is: circle, e=12.6; regular triangle, e=20.8; square, e=16.0.
[0065] When judging a circle, e is usually 11.6~13.6 and R o 0.9~1.1; e is 18.8~22.8 and R when judging triangle o It is 0.5~0.7.
[0066] (3) Identification through template matching
[0067] The first step is to use the dynamic normalization method to normalize the logo image to an m×n target image.
[0068] Take the blue right turn indicator as an example. The previous step has determined the circumscribed rectangle of the traffic sign area and named it area 1, such as figure 2 As shown by the dotted line, this rectangle is reduced proportionally, so that the rectangle becomes the inscribed rectangle of the mark circle, called area 2, to remove the background interference outside the mark, such as figure 2 Shown by the solid line.
[0069] Binarize the image in area 2 through the bimodal method (other existing binarization methods can also be used, such as: clustering method, iteration method, P-quantile method, Otsu method, etc.), because at this time The contrast of the image in area 2 is relatively strong, and the double-peak method binarization can well segment the arrow and background in the traffic sign, and the method is not sensitive to light changes. Further determine the circumscribed rectangle of the symbol through the binary image, such as figure 2 The area within the dash-dotted line 3.
[0070] Normalize the image in area 3 to an image of the same size as the template. The existing normalization method can be used, but in order to express the image information more accurately, the present invention proposes a method for dynamically setting the normalization unit. Assuming that the width and height of the source image are x and y respectively, and the resolution of the feature image to be extracted is m×n, the source image is divided into m×n regions, and the row division calculation formula is as follows:
[0071] a. If x%m=0, each line of the source image should contain x/m pixels
[0072] b. If x%m≠0, each row of the former x%m area contains x/m+1 dots; each row of the latter m-x%m area contains x/m dots.
[0073] Column segmentation calculation is similar to row segmentation calculation. Finally, the original image is divided into m×n areas approximately evenly, and the number of black pixels in each area is detected. If the number of black points is greater than a fixed empirical threshold, it is considered The area is black, and the corresponding point of the target image is set as a black point. After the above processing, the source image can be normalized to the target image of m×n.
[0074] The second step is to perform template matching calculation.
[0075] Use all templates in the corresponding color category to perform matching operations with the normalized image, set the size of the image to be matched to J×K, w(s, t) and f(s, t) are the template image and the image to be matched, respectively For a certain pixel value of the image, the expression of the response result c(x, y) of the final correlation operation is
[0076] c ( x , y ) = X s = 0 K X t = 0 J w ( s , t ) f ( s , t )
[0077] A threshold of c(x, y) is set according to the actual situation. If it is higher than the threshold, the image is considered to be a corresponding type of traffic sign.
[0078] (4) Extract the template based on the actual traffic sign image.
[0079] The main steps are: align the camera with the traffic sign to be extracted from the template, run steps (1) to (3) to capture the traffic sign in the image, observe and confirm that the capture is correct and output the normalized image obtained in (3). Value array, the binary array file is the template. The template array is displayed as an image image 3 , 4 Shown.
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

no PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Similar technology patents

Centrifugal Blower

InactiveUS20080279681A1Reduce noiseSmooth bootPropellersPump componentsImpellerBell mouth
Owner:MITSUBISHI HEAVY IND LTD

Classification and recommendation of technical efficacy words

  • Reduce noise
  • Improve timeliness and accuracy

Method for accurately controlling rotating speed of fan

Owner:LANGCHAO ELECTRONIC INFORMATION IND CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products