CN1598487A - Method for visual guiding by manual road sign - Google Patents

Abstract

The invention relates to an information technology automatic control field, the method is made up of road sign designing method, road sign position, road sign incriminating and navigating device, and road sign incriminating and control flow. The road sign designing method is a gray mode road sign, which is made up of two long black bar with the same width, four transverse long black bar with the same width and the middle numbers, the printed road sign is pasted on the plane vertically to the ground, such as the wall or side surface of computer desk, the height is similar to the camera height of robot, the incriminating and navigating device is only robot with a single camera, finally, it is accomplished road sign incrimination and control flow. The flow carries on binarization to the 256 gray images, then it is scanned and detected, finally the road sign number can be incriminated; the mode is simple, and the cost is low.

Description

Artificial road sign vision navigation method

Technical field

The infotech automation field is adapted at control use in the various merchandising machine people navigation.

Background technology

At present, the location of autonomous mobile robot and navigation problem are important research directions of field in intelligent robotics, also be a gordian technique of intelligent mobile robot simultaneously, for this reason, various location and navigation technologies and onboard sensor system emerge in large numbers in succession, and currently used main method has: rely on compass, odometer etc. to provide the relative positioning system of mobile robot's approximate location and direction; Utilize sonar to set up local scenery model by the distance of measuring between scenery and the robot; The electromagnetism guidance method is to utilize to be layered on the motion path that underground or ground magnetic stripe constitutes robot, comes the constrained robot to walk along magnetic stripe.This method is used for industry more, and the deficiency of existence is that the path can not be changed easily, and cost is also very expensive; To be generating laser combine with specific cooperative target laser scanning method, can calculate the exact position of generating laser in real time; Navigation along the line is that robot utilizes photo-sensitive cell to follow the tracks of a visible or invisible fluorescent pigment of being drawn in advance on the ground at the volley; The most widely used vision guided navigation is the image by scenery around the video camera picked-up, utilize the whole bag of tricks in the Digital Image Processing to analyze, identify some physical features (natural landmark) or artificial feature (artificial road sign) in the scenery then, thereby determine the position of robot, in scientific research and military affairs, widely use.

Visible sensation method is the result who develops rapidly along with the raising of computer speed and optical instrument precision in recent years, it also is a kind of advanced positioning navigation method, mobile robot based on vision guided navigation has better flexibility, has higher intelligence, could really be called intelligent mobile robot.Vision guided navigation based on natural landmark is compared with artificial landmark navigation, and goal-setting complexity, control difficulty are big, are difficult to realize the accurate control of robot, uses for merchandising machine people's control and makes troubles.

Summary of the invention

For solving the deficiency of above air navigation aid, the invention provides a kind of vision navigation method based on artificial road sign, in the method, the artificial position of road sign in world coordinate system is known in advance, after from the scene image of catching, extracting the image coordinate of road sign, by the position of road sign in image and their geometric relationships in world coordinate system, just can calculate the absolute position of video camera in world coordinate system, thereby reach that expense is cheap, noiselessness, no deleterious effect, robot navigation's purpose of containing much information.

Design proposal of the present invention is achieved in that

The artificial road sign vision navigation method of the present invention is made up of following method and apparatus:

1, the method for designing of road sign

2, road sign deposit position

3, landmark identification guider

4, landmark identification control flow

Wherein, the road sign method for designing is a kind of grayscale mode road sign, as shown in Figure 4, it is made of the long secret note of the long secret note of two endways same width, four same width sidewards and middle numeral three parts, is identical for six secret notes of different road signs, being used for road sign detects, and the variation of numeral is used for distinguishing road sign, because each numeral can from 0 to 9, therefore can form 100 kinds of (2 bit digital, 0～99) or 1000 kinds of (3 bit digital, 0～999) different patterns.

The road sign deposit position:

Printed digital road sign is bonded on the plane perpendicular to the ground, as the side of wall or computer desk; Height is roughly the same with the camera height of robot.

The landmark identification guider only is made of the robot that has single camera, and resolution of video camera can be provided with arbitrarily, and this device is 320 * 240.

The landmark identification control flow is made up of the following step, and its process flow diagram is by shown in Figure 10:

Step 1: at first with the 256 color shade image binaryzations that absorbed, method is a lot, only introduces a kind of straightforward procedure here.Behind the robot camera irradiation road sign, in entire image, search out maximum gradation value Graymax and minimum gradation value Graymin, thus the threshold value that will use when obtaining image binaryzation $T=\frac{\mathrm{Gray}\max +\mathrm{Gray}\min }{2},$ Utilize this threshold value image binaryzation, with gray-scale value greater than T become 255, with gray-scale value less than T become 0, mainly be the contrast that increases road sign and surrounding environment, give prominence to artificial landmark identification signal;

Step 2: whether earlier detect horizontal direction 1. have the compound ratio wanted exist, whether if exist, then detecting its left side 8. 2. has vertical compound ratio to exist with the right if lining by line scan to top from the image bottom earlier.In case both direction has all detected compound ratio, so just found an accurate road sign, and then can determine by straight line 3., 4., 5., the rectangular area that 6. surrounded.In this rectangular area, utilize the method for getting threshold value in the first step that the original image in the zone is carried out accurate binaryzation once more, be convenient to become the landmark identification under the illumination condition like this, at by straight line 3., 4., 5., 6., two rectangular areas 7. being surrounded, 3. and center line 4. wherein 7. be, make each limit all parallel mobile to the center of numeral gradually, thereby find the minimum rectangle position that comprises numeral;

(as shown in Figure 6)

Step 3: discriminating digit in two minimum rectangle that computing machine obtains in by previous step respectively, examine 10 numerals among Fig. 7, in the drawings according to three straight line a, b, the intersection location of c and 10 numerals different, computing machine just can identify different digital, for example only just can distinguish 0 with straight line c, 1,4,7; And remaining numeral 2,3,5,6,8,9 is identical with the situation that c intersects, and this moment, b just can further determine it specifically has been which numeral with the crossing situation of numeral again by straight line a.

Artificial road sign vision navigation method is a foundation based on geometric invariance and numeric structure characteristics:

Its neutral line video camera imaging process: (as shown in Figure 1)

O wherein _W-X _WY _WZ _WBe world coordinate system, O _c-X _cY _cZ _cBe camera coordinate system, O _I-Y _IZ _IBe the ideal image coordinate system.Video camera can be arranged in world coordinate system with any pose, O _cBe camera lens photocentre, X _cThe forward of axle is an optical axis direction, O _IBe positioned at X _cThe forward f place of axle, f is the focal length of camera lens, therefore, according to the pinhole imaging system principle, the arbitrary object P in the space is X just at the rear of camera lens _cNegative sense-f the place of axle as becoming an inverted image Q who has dwindled on the plane, in order to connect simultaneously also for the convenience on illustrating with human vision, we are at X _cThe forward f place of axle sets up an ideal image plane O _I-Y _IZ _I, object P will become a upright downscaled images I on this plane this moment, and its projected size is identical with Q;

At last, again the image on the ideal image plane is changed into pixel image, promptly be the relation between object P and the pixel image, for based on the spatial pose that is exactly to determine video camera on the self-align question essence of the mobile robot of artificial road sign by the relative position of artificial road sign on pixel image of several known spatial coordinates.

Geometric transformation in the image-forming principle and character:

If the position of video camera in world coordinate system is that the coordinate of photocentre is (x ₀, y ₀, z ₀) ^T, attitude matrix is R.Some coordinate in world coordinate system is (x in the space _w, y _w, z _w) ^T, the coordinate in camera coordinate system is (x _c, y _c, z _c) ^T, being transformed between them then

$\left[\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right]=R\·\left[\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right]+\left[\begin{array}{c}{x}_0\\ y_0\\ z_0\end{array}\right]$ If make P=(x ₀, y ₀, z ₀) ^T, then have $\left[\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right]{\left[\begin{array}{cc}R& P\end{array}\right]}_{3\×4}\left[\begin{array}{c}{x}_c\\ y_c\\ z_c\\ 1\end{array}\right],----\left(1\right)$ Be that world coordinates is tied to camera coordinate system The orthogonal space conversion(rigid body translation), three-dimensional camera coordinate system (x _c, y _c, z _c) ^TTo ideal image coordinate system (y _u, z _v) ^TBetween conversion The central projection conversion: $y_u=f\·\frac{y_c}{x_c},z_v=f\·\frac{z_c}{x_c}$ Another kind of form is $\left[\begin{array}{c}{y}_u\\ z_v\\ 1\end{array}\right]=\left[\begin{array}{ccc}f& 0& 0\\ 0& f& 0\\ 0& 0& f\end{array}\right]\·\begin{array}{c}\end{array}\left[\begin{array}{c}\underset{\‾}{y_c}\\ x_c\\ \underset{\‾}{z_c}\\ x_c\\ 1\end{array}\right]---\left(2\right)$

Ideal image coordinate system (y _u, z _v) ^TTo pixel coordinate system (u, v) ^TBetween transformation relation be The plane affined transformation: $\left[\begin{array}{c}u\\ v\end{array}\right]=\left[\begin{array}{ccc}{k}_y& 0& u_0\\ 0& k_z& v_0\end{array}\right]\left[\begin{array}{c}{y}_u\\ z_v\\ 1\end{array}\right]---\left(3\right)$

K wherein _y, k _zBe the pixel dimension coefficient, i.e. resolution on Y and the Z direction, (u ₀, v ₀) ^TIt is the picture centre coordinate.

We claim the space position parameter (x of video camera ₀, y ₀, z ₀) ^TAnd direction parameter α, beta, gamma is that rotation matrix R is the video camera external parameter; And the title focal distance f, the pixel dimension coefficient k _y, k _zBe intrinsic parameters of the camera.

In a word, three kinds of geometric transformations from the space object to the pixel image, have been experienced altogether: orthogonal space conversion, central projection conversion, plane affined transformation, in the ordinary course of things, these three kinds of geometric transformations have two common unchangeability, promptly become straight line into straight line with the constant and compound ratio unchangeability of the sequence of positions of time point on straight line.So-called compound ratio is meant establishes A, and B, C, D are four dissimilaritys of conllinear, if they press A on straight line, and C, D, the series arrangement of B, then following this relation just is called this compound ratio of 4, is designated as $\left(\mathrm{ABCD}\right)=\frac{\frac{\mathrm{AC}}{\mathrm{CB}}}{\frac{\mathrm{AD}}{\mathrm{DB}}}.$ So-called compound ratio unchangeability as shown in Figure 2, is thrown 4 ACDB on the straight line l by any central point Ω

Shadow just obtains A ' C ' D ' B ' to any straight line Δ, have this moment $\left(\mathrm{ACDB}\right)=\frac{\frac{\mathrm{AC}}{\mathrm{CB}}}{\frac{\mathrm{AD}}{\mathrm{DB}}}=\frac{\frac{A^{\′}{C}^{\′}}{C^{\′}{B}^{\′}}}{\frac{A^{\′}{D}^{\′}}{D^{\′}{B}^{\′}}}=\left(A^{\′}{C}^{\′}{D}^{\′}{B}^{\′}\right).$ Because the orthogonal space conversion does not change line segment length, so obviously satisfy this character; The example that the central projection conversion is shown in Figure 2 just; Simple for the plane affined transformation than the invariant that is its conversion, promptly $\frac{\mathrm{AC}}{\mathrm{CB}}=\frac{A^{\′}{C}^{\′}}{C^{\′}{B}^{\′}},\frac{\mathrm{AD}}{\mathrm{DB}}=\frac{A^{\′}{D}^{\′}}{D^{\′}{B}^{\′}},$ So also satisfy the compound ratio unchangeability.Because the arbitrariness of central point Ω and straight line l, can derive as drawing a conclusion, face, look side ways, under the different conditions such as myopia, long sight, the compound ratio in the pixel image all equals the compound ratio in the space, as shown in Figure 3.

Advantage of the present invention: in conjunction with the navigation experience of " looking for the number of looking for behind the door earlier ", according to compound ratio unchangeability and numeric structure characteristics under three geometric transformations (orthogonal space conversion, central projection conversion, plane affined transformation) in the video camera imaging principle, we have proposed a kind of artificial road sign that can be applicable to actual vision guided navigation.The characteristics of this road sign are: also be easy to be detected nearly 1000 kinds of schema categories under complex background; Under myopia, long sight, large angle strabismus and change illumination condition, has very high identification stability; Pattern is simple, be easy to make, and expense is cheap; The contained information of road sign is very directly perceived, easy to install.

In the detection and identifying of road sign, the detection of road sign has made full use of the characteristics of compound ratio unchangeability, and the identification of different road signs has then utilized structural difference between numeral.Based on above-mentioned 2 points, we also can change the numeral in the road sign pattern into any symbol easy to identify, and effect is the same.

Description of drawings

Fig. 1 is the linear video camera imaging schematic diagram of the present invention;

Fig. 2 is a compound ratio unchangeability synoptic diagram of the present invention;

Fig. 3 is the compound ratio synoptic diagram under the various environment of observation of the present invention;

Fig. 4 is a road sign mode chart of the present invention;

Fig. 5 is that secret note of the present invention is at compound ratio unchangeability application principle figure;

Fig. 6 detects the road sign synoptic diagram for the present invention;

Fig. 7 is road sign number recognition principle figure of the present invention;

Fig. 8 is a road sign optimal design synoptic diagram of the present invention;

Fig. 9 is an Xc geometric meaning synoptic diagram of the present invention;

Figure 10 is a landmark identification process flow diagram of the present invention;

Figure 11 is road sign of the present invention location synoptic diagram.

Embodiment

The artificial road sign vision navigation method of the present invention is described in detail in conjunction with the accompanying drawings.

The explanation of road sign Design Pattern:

Utilize secret note to construct another advantage of compound ratio, can prove that under heeling condition, the compound ratio value is still constant, promptly $\frac{\frac{\mathrm{AC}}{\mathrm{CB}}}{\frac{\mathrm{AD}}{\mathrm{DB}}}=\frac{\frac{\mathrm{MP}}{\mathrm{PN}}}{\frac{\mathrm{MQ}}{\mathrm{QN}}},$ Thereby strengthened the stability that road sign detects.

Though this road sign can not solve the partial occlusion problem, under a lot of situation of road sign pattern, it is inessential that occlusion issue just seems, this has been blocked, and can also come self-align by other road sign from image.

In actual applications, image resolution ratio and road sign can be more greatly, and decipherment distance is farther like this, and stability is higher.

The design of road sign each several part optimized dimensions is also stable influential to what discern, and as shown in Figure 8, a partly wants long enough, to guarantee the size of numeral; The b part will have certain distance, is convenient to determine to comprise under the stravismus situation rectangular area of numeral; The numeral size is determined to guarantee c, d, e long enough, be convenient under long sight and stravismus situation, determine to comprise the minimum rectangle of numeral, f will have certain altitude, so just can detect every advance walking along the street mark of 5 row or 10, save working time greatly, g will have certain-length, can guarantee so still can detect compound ratio on the vertical direction in large-angle inclined apparent time.

Robot localization method based on a road sign:

In video camera imaging principle one joint in front, know, from space object to having experienced three kinds of geometric transformations altogether by video camera imaging sketch map picture: orthogonal space conversion, central projection conversion, plane affined transformation.As shown in Figure 1, suppose O _w-X _wY _wZ _wFor getting fixed world coordinate system arbitrarily; O _c-X _cY _cZ _cBe camera coordinate system, the photocentre of camera lens is an initial point, and its position in world coordinate system is (x ₀, y ₀, z ₀) ^T, X _cAxle is an optical axis direction, if the Eulerian angle α of known camera coordinate system in world coordinate system, beta, gamma, we are easy to obtain the attitude matrix R of video camera, so just can write out the rigid body translation formula between world coordinate system and the camera coordinate system $\left[\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right]=R\·\begin{array}{}\end{array}\left[\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right]+\left[\begin{array}{c}{x}_0\\ y_0\\ z_0\end{array}\right]----\left(4\right)$

(x wherein _w, y _w, z _w) ^TRepresent some coordinate in world coordinate system in the space, (x _c, y _c, z _c) ^TRepresent this coordinate in camera coordinate system, obviously, can obtain following relation by geometric knowledge $\left[\begin{array}{c}1\\ y_I\\ z_I\end{array}\right]=\left[\begin{array}{ccc}1& 0& 0\\ 0& f& 0\\ 0& 0& f\end{array}\right]\·\left[\begin{array}{c}1\\ y_c/x_c\\ z_c/x_c\end{array}\right],$ (y wherein _I, z _I) ^TBe the ideal image coordinate system.Here it is is tied to ideal image coordinate system (y from the three-dimensional camera coordinate _I, z _I) ^TBetween the central projection conversion.

By ideal image coordinate system basis $\left[\begin{array}{c}{y}_p\\ z_p\end{array}\right]=\left[\begin{array}{cc}{k}_y& 0\\ 0& k_z\end{array}\right]\·\left[\begin{array}{c}{y}_I\\ z_I\end{array}\right]$ Just can obtain pixel coordinate system, wherein k _y, k _zBe the pixel dimension coefficient, i.e. resolution on Y and the Z direction.

The above analysis, we can obtain being tied to transformation relation between the image data coordinate system by world coordinates.Space position parameter (the x of video camera in the formula ₀, y ₀, z ₀) ^TAnd direction parameter α, beta, gamma is that rotation matrix R is the video camera external parameter; And the title focal distance f, the pixel dimension coefficient k _y, k _zBe intrinsic parameters of the camera.

Because we artificial road sign be bonded at perpendicular plane, ground on, and height is roughly the same with the height of video camera, if the work road surface of autonomous mobile robot is smooth so, then robot just has only around Z with respect to the attitude of world coordinate system _wThe rotation γ of axle promptly is that the outer ginseng of video camera only is x ₀, y ₀, γ, this also is the pose of robot in world coordinate system.So formula (4) can be reduced to $\left[\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right]=\left[\begin{array}{ccc}\cos \mathrm{\γ}& -\sin \mathrm{\γ}& 0\\ \sin \mathrm{\γ}& \cos \mathrm{\γ}& 0\\ 0& 0& 1\end{array}\right]\·\left[\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right]+\left[\begin{array}{c}{x}_0\\ y_0\\ 0\end{array}\right].$ Thereby the transformation for mula that is tied between the pixel coordinate system by world coordinates can specifically be written as

$x_c\·\left[\begin{array}{c}1\\ y_p\\ z_p\end{array}\right]=\left[\begin{array}{ccc}1& 0& 0\\ 0& k_y& 0\\ 0& 0& k_z\end{array}\right]\·\left[\begin{array}{ccc}1& 0& 0\\ 0& f& 0\\ 0& 0& f\end{array}\right]\·\left[\begin{array}{ccc}\cos \mathrm{\γ}& \sin \mathrm{\γ}& 0\\ -\sin \mathrm{\γ}& \cos \mathrm{\γ}& 0\\ 0& 0& 1\end{array}\right]\·[\left[\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right]-\left[\begin{array}{c}{x}_0\\ y_0\\ 0\end{array}\right]],$ Further launch

Before the self-align algorithm of deriving, explain x earlier _cGeometric meaning, as shown in Figure 9, can prove x _cPut F exactly at X _cProjection on the axle and the distance of photocentre, i.e. AE.PROOF: in Δ ABE, because BA ⊥ AE, so AE=DEsin ∠ ADE, again owing to BE ⊥ GH, thus ∠ ADE=π-∠ AGH, and ∠ AGH=γ, so AE=DEsin γ.

BE=y _w-y ₀, FB=x _w-x ₀So, BD=FBctg (so π-γ). AE=(BE-BD) sin γ=(y _w-y ₀) sin γ+(x _w-x ₀) cos γ.

Suppose that video camera is desirable pin-hole model, the Q point is the central point of image so.Be easy to obtain the length PQ=y of PQ in ideal image according to the position P of object point F in pixel image _I=y _p/ k _y, because EQ=f, so ∠ PEQ=arctan (PQ/EQ)=arctan (y _p/ (k _yF)), according to formula EF=x _c/ cos (∠ PEQ)=x _c/ cos (arctan (y _p/ (k _yF))), just be easy to try to achieve the distance between object point F and the video camera photocentre, promptly be,, just can obtain the distance of object point and camera lens photocentre if known the projection of object point on primary optical axis and the distance of camera lens photocentre.

Illustrate below and how to carry out self-align based on an artificial road sign, as shown in figure 11, contain road sign in the image in case identify, utilize certain bar sweep trace from left to right to be easy to just can obtain four borders Y in pixel coordinate system of two endways secret notes so _pAxial coordinate y _p ^A, y _p ^B, y _p ^C, y _p ^D, along y _p ^AAnd y _p ^DFind from bottom to top respectively sidewards two secret notes bottom with Z in pixel coordinate system topmost _pAxial coordinate z _p ^E, z _p ^F, z _p ^G, z _p ^HBecause road sign is attached on the wall perpendicular to the ground, and video camera adopts and looks squarely mode, thus EF and GH all perpendicular to the primary optical axis of video camera, i.e. X _CAxle.So it be easy to show that, some E and some F are at X _CProjection on the axle is a same point, might as well be made as x _c, according to the 3rd equation in the formula (5), can obtain two equations as follows so, Subtract each other and to obtain $x_c=f\·k_z\·(z_w^E-z_w^F)/(z_p^E-z_p^F),$ Z in the formula _w ^E-z _w ^FBe the fixed value that when road sign designs, just can decide, z _p ^E, z _p ^FBe an E and some F Z in pixel coordinate system _pAxial coordinate so according to the analysis of front, can obtain the distance of video camera photocentre to straight line EF, is made as L _Left, in like manner, also can obtain the distance of video camera photocentre to straight line GH, be made as L _RightBecause A limit, the leftmost side and D limit, the rightmost side coordinate in world coordinate system of two endways secret notes is known, is made as (x respectively _w ^A, y _w ^A) ^T(x _w ^D, y _w ^D) ^TSo just can construct the equation of two circles according to the photocentre of video camera to the distance on these two limits, (x wherein ₀, y ₀) ^TBe the coordinate in the alive boundary of the video camera photocentre coordinate system, also can regard the position of robot in world coordinate system as, top equation is that two circles intersect, and can obtain two groups generally speaking and separate, be easy to irrational one group separated by check and delete, utilize the robot location (x that has obtained ₀, y ₀) ^T, in conjunction with preceding two equations in (5), just can obtain the attitude angle γ of robot in world coordinate system again, detailed process is as follows: second equation arrangement of first equation substitution in (5) can be got $\tan \mathrm{\γ}=\frac{f\·k_y\·(y_w-y_0)-y_p\·(x_w-x_0)}{y_p\·(y_w-y_0)+f\·k_y\·(x_w-x_0)},$ Obtain the value of γ then according to actual conditions.

So far, solved and utilize an artificial road sign to carry out the self-align problem of robot.Certainly, if when having a plurality of road sign in the image, can adopt multichannel to demarcate method for position.

The artificial road sign vision navigation method of the present invention through the field conduct stable and reliable operation, is a kind of merchandising machine people air navigation aid of comparatively science.

Claims (4)

1, a kind of artificial road sign vision navigation method, this method is by the method for designing of road sign, the road sign deposit position, the landmark identification guider, landmark identification control flow four parts are formed, it is characterized in that the road sign method for designing is a kind of grayscale mode road sign, it is by the long secret note of two endways same width, the long secret note of four same width sidewards and middle numeral three parts constitute, six secret notes for different road signs are identical, being used for road sign detects, and the variation of numeral is used for distinguishing road sign, because each numeral can from 0 to 9, therefore can form 100 kinds of (2 bit digital, 0～99) or 1000 kinds of (3 bit digital, 0～999) different patterns.

2, by the described a kind of artificial road sign vision navigation method of claim 1, it is characterized in that the road sign deposit position: printed digital road sign is bonded on the plane perpendicular to the ground, as the side of wall or computer desk; Height is roughly the same with the camera height of robot.

3, by the described a kind of artificial road sign vision navigation method of claim 1, it is characterized in that the landmark identification guider only is made of the robot that has single camera.

4,, it is characterized in that the landmark identification control flow is made up of the following step by the described a kind of artificial road sign vision navigation method of claim 1:

Step 1: at first with the 256 color shade image binaryzations that absorbed, behind the robot camera irradiation road sign, in entire image, search out maximum gradation value Graymax and minimum gradation value Graymin, thus the threshold value that will use when obtaining image binaryzation $T=\frac{\mathrm{Gray}\max +\mathrm{Gray}\min }{2},$ Utilize this threshold value image binaryzation, with gray-scale value greater than T become 255, with gray-scale value less than T become 0, mainly be the contrast that increases road sign and surrounding environment, give prominence to artificial landmark identification signal;

Whether step 2: lining by line scan to top from the image bottom earlier detects horizontal direction earlier and 1. has the compound ratio of wanting to exist, if exist, 8. 2. whether then detect its left side has vertical compound ratio to exist with the right, in case both direction has all detected compound ratio, so just found an accurate road sign, and then can determine 3. by straight line, 4., 5., 6. the rectangular area that is surrounded, in this rectangular area, utilize the method for getting threshold value in the first step that the original image in the zone is carried out accurate binaryzation once more, be convenient to become the landmark identification under the illumination condition like this, at 3. by straight line, 4., 5., 6., 7. two rectangular areas that surrounded, 3. and center line 4. wherein 7. be, make each limit all parallel mobile to the center of numeral gradually, thereby find the minimum rectangle position that comprises numeral;

Step 3: discriminating digit in two minimum rectangle that computing machine obtains in by previous step respectively, examine 10 numerals, according to three straight line a, b, the intersection location of c and 10 numerals different, computing machine just can identify different digital, for example only just can distinguish 0 with straight line c, 1,4,7; And remaining numeral 2,3,5,6,8,9 is identical with the situation that c intersects, and this moment, b just can further determine it specifically has been which numeral with the crossing situation of numeral again by straight line a.

Images