CN202771439U - Traffic sign automatic identification system based on MATLAB - Google Patents

Abstract

The utility model discloses a traffic sign automatic identification system based on MATLAB. The traffic sign automatic identification system includes a digital DV camera installed on a vehicle. The digital DV camera is connected with a computer through a data line; and the computer is connected with a display device or a warning device. According to the traffic sign automatic identification system of the utility model, the camera added on the vehicle can acquire traffic sign information appearing in a travelling process of the vehicle; through automatically extracting characteristic portions of a photographed image, a shipboard aircraft can detect a corresponding traffic sign and prompt or warn a driver in time, such that the driver can control the vehicle, thereby maintaining smooth traffic and preventing traffic accidents.

Description

Based on the MATLAB automatic traffic sign identification device

Technical field

The utility model relates to the traffic applied technical field, relates in particular to a kind of device based on MATLAB automatic recognition of traffic signs technology.

Background technology

Along with the continuous research and development of intelligent transportation system, the automatic recognition of traffic signs system is paid attention to gradually and is developed.In day by day flourishing traffic system, the automatic identification of traffic sign also will occupy more and more consequence, the automatic identification of traffic sign will become semi-automatic and important component part automotive vehicle, so the research of automatic recognition of traffic signs is just had important value.Road signs automatically identification are the research fields of utilizing computer vision technique to identify automatically.When vehicle front has traffic mark to occur, can give the alarm signal of sound of driver and image, thereby reminding driver is noted the sign in the place ahead, and vehicle is controlled, reduce the generation of traffic hazard.

Automatic traffic sign identification device will obtain to use more widely as an important servicing unit of intelligent transportation system.At the automobile in future automatically, the automatic identification of traffic sign is important one of assembly of wanting of its control system in the semi-automatic driving system.In the electronic communication map, automatic recognition of traffic signs submits necessary information to the location of vehicle in communication chart.Therefore, the automatic recognition of traffic signs systematic research has extremely important practical significance and value.

Summary of the invention

The purpose of this utility model is, provide a kind of based on the automatic will recognition device of MATLAB traffic sign, utilization is installed camera additional at automobile, in the Vehicle Driving Cycle process, the road signs information that occurs is gathered, carrier-borne aircraft is by automatically extracting the characteristics part in the captured image, thereby detects corresponding traffic sign, makes prompting or warning to the driver timely, vehicle is controlled, to keep the smooth generation with preventing traffic hazard of traffic process.

In order to realize above-mentioned task, the technical scheme that the utility model adopts is:

A kind ofly it is characterized in that based on the MATLAB automatic traffic sign identification device, comprise the digital DV video camera that is installed on the automobile, digital DV video camera links to each other with computing machine, and computing machine connects display or phonetic alarm.

Of the present utility model real-time is better based on the MATLAB automatic traffic sign identification device, the clear picture that obtains.Numeral DV video camera Real-time Obtaining is to the road signs image, and with the road signs image that gathers, in the input computing machine, image preprocessor in the computing machine is at first carried out necessary pre-service work to the image that gathers, and mainly is to carry out histogrammic equalization in the HIS space and carry out color segmentation at rgb space.Then the image after cutting apart is carried out necessary processing, remove unnecessary noise and small size interference region, then according to the circular characteristic feature that indicates, extract with circularity and may be the zone of target, exclude the interference region of those non-the round degree of characteristics, and the image that extracts is carried out rim detection with the Log operator.Carry out at last the coupling identification work of image object, with Hu not bending moment make feature templates as matching characteristic, effectively mate again identification, at last the sign of identifying is shown or by display or phonetic alarm the driver is warned.

Description of drawings

Fig. 1 is structural principle block diagram of the present utility model;

Fig. 2 is image preprocessor process flow diagram.

Fig. 3 is based on the figure image intensifying program flow diagram of histogram equalization.

Fig. 4 is based on the image recognition program process flow diagram of characteristic matching.

The utility model is described in further detail below in conjunction with drawings and Examples.

Embodiment

Referring to Fig. 1, the utility model provides a kind of based on the MATLAB automatic traffic sign identification device, comprises the digital DV video camera that is installed on the automobile, and digital DV video camera links to each other with computing machine by data line, and computing machine connects display device or warning device.

Photographing continuous picture by digital DV video camera is sent in the computing machine in the automobile, the figure image intensifying program of image preprocessor, histogram equalization is arranged and based on the image recognition program of characteristic matching in the computing machine, by the finish dealing with automatic identification of road sign of a series of image.

With reference to Fig. 2, of the present utility model as follows based on the MATLAB automatic traffic sign identification device course of work:

(1) image extracts

Obtaining of road signs image is to gather by a digital DV video camera that is placed on automotive interior, what numeral DV video camera photographed is traffic sign, in the process of driving, in the picture that digital DV video camera photographs, the size of the sign in the picture also can closely and gradually become large along with distance becomes, and these pictures are sent in the computing machine, after identification is finished, just no longer the same sign in other picture is identified the time of processing to reduce computer data.

The image processing links is exactly to select road signs for identifying processing from the road signs Computer of input, it is most important program in the device, in order to reduce the data volume of processing, all leach some frames that do not comprise useful information at every one-phase as far as possible and do not participate in follow-up processing, the image that obtains by video camera extracts to obtain image sequence for identification by image first; Then, to every two field picture, need to eliminate because the ghost image that causes picture frame is taken in staggered scanning by sub sampling; Then, just image segmentation and feature extraction and acquisition Characteristic of Image figure compare with the characteristic pattern with standard sign and realize the sign location, and finally sort out with final and confirm.

(2) sub sampling

In the present embodiment, the digital DV video camera of taking road signs adopts common home digital camera, and what it adopted is interlace mode, because be staggered scanning, so two fields that photograph are actually discrepant, this species diversity has formed ghost image.The object of taking relative motion in order to eliminate this staggered scanning causes ghost image, adopts sub-sampling methods to solve this problem here, and sub sampling is exactly to take interval delegation to extract the pixel of image, is actually and obtains the half frame images that a staggered scanning is obtained.

(3) figure image intensifying, Threshold segmentation, feature extraction

With reference to Fig. 3, figure image intensifying program based on histogram equalization, in the HIS color model, carry out histogram equalization, algorithm steps is as follows: at first the original path traffic sign is transformed into the HIS model from the RGB model, then with the H(tone), I(brightness), the S(saturation degree) carry out respectively histogram equalization, and then road signs are returned the RGB model from the HIS model conversion again, then will obtain the RGB road signs image of the obvious blast of a width of cloth.

The fundamental purpose of figure image intensifying is in order to improve the visual effect of final recognition image, improve image sharpness and, so that the analysis of computing machine and processing.RGB can not reflect the morphological feature of image, and Chang Yaoyong rgb2gray function becomes image transitions 8 gray-value image directly to process, and can pass through histogram, grey scale change; Again image is carried out histogram equalization; With the im2bw function gray level image is converted into bianry image.Bianry image is the image that each pixel only has two kinds of possible numerical value or gray shade scale state on the image.

Then, image is carried out Threshold segmentation, the target area is extracted, but the similarity of background color and color of object and R, the three-component correlativity of G, B, so cause noise spot a lot, so with function imerode and imdilate image is corroded expansion process, image is thoroughly extracted, do not stay any noise.

(4) rim detection, refinement

After adopting the corrosion plavini that bianry image is processed, to reach the effect of edge-smoothing.Then adopt the Log operator to carry out rim detection, satisfied the processing requirements to object.The Log algorithm is a kind of second order edge detection method.The zero crossing in the second-order differential comes detected edge points in the gradation of image value by seeking for it.The Log principle is: the edge that gray level becomes forms a unimodal function, peak corresponding edge point through differentiating operator; Unimodal function is carried out differential, and then the differential value at peak value place is 0, and peak value both sides opposite in sign, and the zero crossing in the corresponding second-order differential of original extreme point can be with the edge extracting of image out by detecting zero crossing.

The computing formula of Log operator:

$\▿G(x,y)=\frac{{\∂}^{2}G}{{\∂}^{2}x}+\frac{{\∂}^{2}G}{{\∂}^{2}y}$

$=\frac{1}{{\mathrm{\π\σ}}^4}(\frac{x^2+y^2}{{\mathrm{\σ}}^2}-1)\exp (-\frac{x^2+y^2}{{2\mathrm{\σ}}^2})$

In the formula: G(x, y) be the function of selecting when image is processed; X, y are rounded coordinate; σ is the mean square deviation of Gaussian distribution.

In the present embodiment, what adopt the Log operator extraction is the two-wire edge, so need to do further processing to object edge, it is processed into the edge of single pixel, so that follow-up identification work.

Be the edge of removing target by the iterative scans image for the processing of refinement, all can remove one deck edge pixel each time, until obtain Single pixel edge during the pixel that target does not have can to delete again.

(5) characteristic matching and pattern-recognition

The method of characteristic matching is normally utilized the Traffic Sign Images of standard, then utilizes based on matching characteristic and sets up feature templates.Read with imread that image and image to be identified mate in the java standard library, the image in the storehouse is carried out binaryzation and carries out rim detection with the Log operator, seek the traffic sign characteristic area similar to feature templates of identification.Pattern-recognition is exactly the feature according to research object, assert its classification with certain analytical algorithm.Set up the matching characteristic sample, utilize image in the java standard library and be identified similarity between the image, namely their similarity coefficient is weighed the similarity of two width of cloth images.Two width of cloth images are more similar, and then similarity coefficient is more close to 1; Image is more dissimilar, and similarity coefficient is more near 0.Similarity coefficient is defined as follows:

$R_{\mathrm{xy}}=-\frac{x\·y}{x\·x-\mathrm{xy}+y\·y}$

x＝(x ₁,x ₂,…x _n)，y＝(y ₁,y ₂,…y _n)，

$x\·x=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_i\·x_i,$

$x\·y=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_i\·y_i,$ $y\·y=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i\·y_i$

X is the eigenwert of sample in the formula, and y is Characteristic of Image value to be identified, and i is the symbol of i feature of identification, and n is the number of matching characteristic.Formulate feature templates according to eigenvector, calculate similarity coefficient and threshold comparison, identify the classification of target.

Identification based on template matching method: images match is widely used in robot vision, industrial automation and unattended system.In the process of machine recognition things, often need different sensors or same sensor are spatially aimed at two width of cloth or multiple image that same scenery obtains under different time, different image-forming condition, or seek corresponding pattern in another width of cloth figure according to known mode, this just is called coupling.The template matching method of image is exactly whether to have certain known template image in the research piece image.

With reference to Fig. 4, the present embodiment has adopted not bending moment of HU, according to the different characteristics of identifying object, has selected effectively to identify the feature of different objects as the member of feature templates.The mark sheet that is identified the standard in the template base of traffic sign is compared one by one, and two width of cloth images are more similar, and then similarity coefficient is more close to 1; Image is more dissimilar, and similarity coefficient is more near 0.The similarity coefficient that a certain traffic sign arranged in finding template base judges that then being identified image is the standard traffic sign close to 1 the time.

After standard traffic mark is identified, computing machine will judge whether the driver follows the indication of sign, also can link to each other with speed counter, whether detect automobile travels by the speed of speed(-)limit sign, and note the sign in the place ahead by display or phonetic alarm driver, vehicle is controlled, reduced the generation of traffic hazard.

Claims (1)

1. one kind based on the MATLAB automatic traffic sign identification device, it is characterized in that, comprise the digital DV video camera that is installed on the automobile, digital DV video camera links to each other with computing machine by data line, and computing machine connects display device or warning device.

Images