JPH09326009A - Serial sepcific number segmenting device for license plate - Google Patents

Abstract

PROBLEM TO BE SOLVED: To stably segment a license plate serial specific number by binarizing an image according to a specific threshold value. SOLUTION: A profile generating means 15 generates a density profile along a horizontal line crossing a plate image being a license plate area obtained by a plate position detecting means 14. A maximal value detecting means 16 outputs the maximum and minimum value of the profile. A threshold setting means 17 determines a proper luminance value between the maximum and minimum values outputted by the maximal value detecting means 16 as the threshold value. On the basis of the threshold value which is thus set, a binarizing means 18 binarizes the plate image obtained by the plate position detecting means 14. The binarized image is segmented by a serial specific number segmenting means 19.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a serial designated number cutting device for a license plate of a vehicle.

[0002]

2. Description of the Related Art In a vehicle automatic recognition system, a license plate reading device for reading a license plate from an image including a license plate of the vehicle is used. In this license plate reading device, in order to cut out a series designated number of the license plate from the photographed image, the photographed original image is binarized, and the series designated number of the license plate is cut out from the binarized image. There is.
There is a min-max method in the binarization processing of this image, and according to this, the binarization processing that does not have a strong adverse effect due to weather conditions, plate stains, etc. can be performed by using the variable threshold processing. it can.

[0003]

In the conventional technique as described above, it is necessary to perform the filtering process on the original image a plurality of times, and the number of times is optimally about half the maximum pixel width of the character. There is. However, a vehicle traveling on a general road has a mixture of plates of different sizes such as a large format plate and a medium format plate, and the traveling speed is also different, so that the photographing distance and the like are not constant. For this reason, the size of the image on the plate changes, and it is not possible to always perform optimum filter processing, making it difficult to obtain a stable cutting operation.

[0004]

The present invention adopts the following constitution in order to solve the above points. <Structure> The license plate series designated number cutting device according to claim 1 is a profile creating means for creating a density profile along a straight line that crosses the license plate area of an image in which the license plate area is specified, and the obtained density. Detecting means for obtaining the maximum and minimum values of the profile, threshold setting means for setting a threshold value from the minimum and maximum values obtained by the detecting means, and binarization processing of the image based on the set threshold values. The present invention is characterized by including binarization processing means and cutting-out means for cutting out a series designated number of a license plate from a binarized image obtained by the binarization processing means.

Further, according to a fourth aspect of the present invention, there is provided a license plate serial designation number cutting device, wherein the license plate serial designation number cutting device binarizes the image in which the license plate area is specified, and the binarizing means. And a clipping interpolating means for estimating a character position that could not be cut out from the cutting result cut out by the cutting means and cutting out the character position. And

<Operation> In the license plate series designated number cutting device according to the first aspect, the threshold value in the binarization processing of the image is the density profile created along a straight line crossing the license plate area of the image. A stable binarized image can be obtained irrespective of the size of the license plate area of the image by the binarization processing set between the maximum value and the minimum value and using this threshold value. Therefore, it is possible to perform stable clipping without being strongly affected by external influences such as weather and stains on the plate, and regardless of the size of the license plate cut out from the image.

Further, in the license plate series designated number cutting-out device according to the fourth aspect, even if the character part cannot be cut out by the cutting-out means due to dirt or the like, the interpolation means is cut out by the cutting-out means. Since the character position that could not be cut out is estimated from the cutout result and the character position is cut out, it is possible to reliably cut out the license plate series designated number regardless of whether the plate is dirty or not.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail with reference to the illustrated embodiments. <Specific Example 1> FIG. 1 is a block diagram showing a license plate serially designated number cutting device according to a specific example 1. The license plate serial designation number cutting-out device 10 converts video images into a digital image by a video input device 11 such as a video camera for obtaining an image of a running vehicle and an analog image obtained by the video input device 11. A / D converter 12, an image memory 13 for storing a digital image obtained by the A / D converter 12, and a well-known plate position detection for detecting a license plate position from an image in the image memory 13. Means 14 and. As the plate position detecting means 14, for example, JP-A-63-25
As described in Japanese Patent No. 1898, it is possible to apply means for performing horizontal image differentiation processing and extracting the license plate position in response to the output.

Further, the license plate series designated number cutting-out device 10 has a profile creating means 15 for creating an image of the license plate obtained by the plate position detecting means 14, that is, a density profile on the horizontal line of the plate image, and a profile creating means. An extreme value detecting means 16 for obtaining a maximum value and a minimum value from the concentration profile obtained by the means 15, and an extreme value detecting means 16
Threshold setting means 17 for selecting and setting an optimal threshold for binarizing the image stored in the image memory 13 from the maximum and minimum values obtained by the above, and the threshold setting means 17. And a binarization processing unit 18 for binarizing the image stored in the image memory 13 according to the threshold value.

The binarized image containing the license plate portion obtained by the binarization processing means 18 is obtained by obtaining the coordinates of the numeral portion by a series well-specified number cutting-out means 19 of the license plate which is well known in the prior art. The number part is cut out.

The profile creating means 15, the extreme value detecting means 16, the threshold value setting means 17 and the binarization processing means 18 are
It can be realized by a computer. 2 and 3 are explanatory views (No. 1 and No. 2) showing an example of the density profile obtained by the profile creating means 15.
It is. The profile creating means 15, as shown in FIG. 2A, is a plate image 20 which is a license plate area obtained by the plate position detecting means 14.
A density profile is created along a horizontal line 21 that traverses. The serial designation number of the license plate is the largest character among the characters on the license plate, and is located in the lower portion including the central portion of the cut-out plate image 20. Therefore, based on these pieces of information, the horizontal direction line 21 The position can be set appropriately, for example, can be set at a height position of about 1/3 of the license plate.

From the concentration profile obtained by the profile creating means 15, the extreme value detecting means 16 determines the maximum value and the minimum value of the profile by variables max [1 to 1].
It is output as m] [2] and variables min [1 to n] [2]. Here, m is the number of local maxima, n is the number of local minima respectively, and max [1 to n] [1] and min [1 to m] [1] are X coordinate values giving the respective extreme values. Show. max [1] [1] <max [2] [1] <
… <Max [m] [1], min [1] [1] <min [2] [1] <… <min
There is a relationship of [n] [1].

The threshold value setting means 17 is a minimum value max [L] [2] (1≤L of the maximum values output from the extreme value detecting means 16).
≤m) and the maximum value of the minimum values min [H] [2] (1≤H
≦ n), and if max [L] [2]> min [H] [2], max
A threshold value is set to an appropriate luminance value between [L] [2] and min [H] [2], that is, between the two straight lines 22 and 23 in the example shown in FIG.

By thus determining the threshold value,
As shown in the plate image 20 of FIG. 2 (b),
Even if the plate image 20 is partially shaded, max
As long as [L] [2]> min [H] [2], that is, unless the relationship between the minimum value of the maximum value and the maximum value of the minimum value is reversed, it is possible to reliably separate the character part and the background part. A threshold for binarization is provided.

On the other hand, as shown in FIG. 3 (c),
For example, as the brightness changes from the lower left corner to the upper right corner of the plate image 20, the minimum maximum value max [L]
[2] may be less than or equal to the minimum value min [H] [2]. In such a case, in order to obtain the x coordinate value x1 that divides the x-axis region so that the relationship of max [L] [2]> min [H] [2] is not reversed, max [L] [2] ＞ min [H *] [2] (where H
* Obtains x1 for which the maximum value of the minimum values in 0 <x ≦ x1 holds. As shown in FIG. 3D, 0 <x ≦ x1 region and x1 <divided by such x1
Thresholds are determined between the straight lines 22a and 23b and between the straight lines 22b and 23b in the area x. In FIG. 3D, ma
x [L *] [2] is the minimum value among the maximum values in the region of x1 <x.

In the region of x1 <x, max [L *] [2] and min
When [H] [2] is obtained, if the relationship of max [L *] [2]> min [H] [2] is not established, that is, (on the x-axis as shown in FIG. 3D) Max [L] [2]> min [H] [2]
In the case where the relationship of 1 is reversed, the threshold is determined in each area by dividing into a number of areas so that this relationship is not reversed.

FIG. 4 is a flowchart showing the procedure for determining the threshold value. The threshold setting means 17 is the extreme value detecting means 1
From the maximum and minimum values obtained in step 6 and the information about the x-axis coordinates that give them, xmin is set to 0 and xmax is set to the larger value a of m and n (step S101). ). Within this setting range of x, the minimum value max [L] [2] of the maximum values and the maximum value min of the minimum values.
[H] [2] is obtained (step S102). At this time, the relationship of max [L] [2]> min [H] [2] (step S10
If 3) is satisfied, the process proceeds to step S104, and if the relationship is not satisfied, the process proceeds to step S105.

In step S104, the range of x (xmin
, Xmax), minimum value max [L] [2] and maximum value min [H]
[2] is stored, and if xmax is the larger value a of m and n (step S108), a threshold value is set to the minimum value max [L] [2] and the maximum value min [H] [2]. Yes (step S10
9). On the other hand, when the predetermined relationship is not satisfied in step S103, the minimum value max [L] is determined in step S105.
The relationship between max [L] [1] giving [2] and min [H] [1] giving the maximum value min [H] [2] is compared. The relationship between the two is ma
If x [L] [1] <min [H] [1], xmax = H-1 is executed (step S106), and if this relationship is not satisfied,
xmin = L-1 is executed (step S107), and in any case, the process returns to step S102. If the relationship of xmax = a is not satisfied in step S108,
xmin = xmax (step S110) and xmax = a (step S111) are set, and the process returns to step S102.

The binarization processing means 18 performs the binarization processing on the plate image obtained by the plate position detection means 14 based on the threshold value thus set. The image subjected to the binarization processing is subjected to the cutting processing by the series designated number cutting means 19. The serial designation number cutting-out means 19 sets 0 of the binary image as a black pixel, and labels this black pixel. The serially designated number cutting-out means 19 detects the coordinates of a region in which the features of the circumscribed rectangle of each labeled region satisfy the following conditions as a serially designated number region. The conditions are (1) the area S, height H, and width W of the circumscribed rectangle are within predetermined threshold ranges, (2) the center position of the circumscribed rectangle is located in the lower half of the license plate, ( 3) Aspect ratio of circumscribed rectangle (H
/ W) is within a predetermined threshold range.

The serial designation number cut-out means 19 sets a circumscribed rectangle that satisfies all of these conditions as a serial designation number.
Determine and detect the coordinates. On the other hand, if there is no circumscribing rectangle that satisfies these conditions, the series designated number cutout unit 19 performs the same processing on the image obtained by inverting the black and white of the plate image.

In the above description, the density plot by the density profile creating means 15 is converted into a single horizontal line 21.
As shown above, a plurality of horizontal lines 21 are drawn to obtain the maximum and minimum values of the concentration profile on each line, the minimum value thereof is calculated from all the maximum values, and the maximum value thereof is calculated from all the minimum values. A value can be obtained, and the threshold value can be set based on the minimum value and the maximum value in the same manner as described above.

As described above, according to the license plate serial designated number cutting device 10 shown in FIG. 1, the change in horizontal luminance on the serial designated numbers is examined to set the threshold value for the binarization processing of the screen. Therefore, it is difficult to be affected by external influences such as weather and stains on the plate, and since the character width does not contribute to the determination of the threshold value, the size of the plate image to be cut out can be determined. Regardless of, stable cutting can be performed.

<Specific Example 2> In the second specific example, a part of the serially designated number of the license plate cannot be cut out by the serially designated number cutting means due to external influences such as weather, stains on the plate, influence of the plate frame and the like. If it is, estimate the remaining number part from the cut out number part,
Cutout interpolation means for cutting out this portion is provided.

The license plate series designated number cutting-out device 110 shown in FIG. 5 includes the same video input means 11, A / D converter 12, image memory 13 and plate position detecting means 14 as shown in FIG.

As the binarization processing means 18 for obtaining the binarized image from the image stored in the image memory 13, the profile creating means 15, the extreme value detecting means 16 and the threshold setting means 17 shown in FIG. It is possible to use a binarization processing unit that performs binarization processing based on the obtained threshold value. Instead of this, as the binarization processing unit 18, various binarization methods such as a discriminant analysis method for the multivalued plate image obtained by the plate position detection unit 14 can be adopted.

The binarized plate image obtained by the binarization processing means 18 cuts out a series of designated numbers of the license plate by a method similar to that described with reference to FIG. 1, for example. Interpolation means 24 is provided for cutting out the portion when it is determined that the cutting is not performed.

FIG. 6 is an explanatory diagram of the license plate standard. The license plate 25 has four numerical areas 25 for the serial number assigned to the license plate.
a to 25d are specified. Numerical areas 25a to 25
The center-to-center distance of d is indicated by a to d, and the vertical dimension is indicated by h. The center-to-center distance b between the number regions 25b and 25c in the central portion is larger than the center distance a between the number regions adjacent to the outside.

FIG. 7 shows an example in which a defective portion can be inferred based on such an arrangement relationship. 7A to 7
FIG. 7C shows a case where three characters can be cut out by the series designated number cutting-out means 19 and one character in the blank frame cannot be cut out, and FIG. 7D can similarly cut out two characters. , When two characters in the blank frame cannot be cut out, and FIG. 7E shows a case where two characters can be similarly cut out and one character in the blank frame cannot be cut out. In the relationship shown in FIG. 7, the existence of the blank frame portion can be estimated, and the interpolating means 24 cuts out the blank frame portion along this estimation.

The center-to-center distance and vertical dimension of each numeral area are
Since it varies depending on the size of the plate image, each distance is set to correspond to this variation in size.
Handled in proportion to the vertical dimension. The center-to-center distance and the vertical dimension ratio are indicated by a / h, b / h, c / h, and d / h, respectively, and these values are a / h> b / h> c / h> d / h. Have a relationship. Therefore, the threshold for discriminating between a / h and b / h is Tab, and similarly, the threshold for discriminating between b / h and c / h is Tbc, and between c / h and d / h. The threshold for this is Tcd. Further, the lower limit threshold of the numerical range that can be regarded as a / h is Taa, and the upper threshold of the numerical range that can be regarded as d / h is Tdd. At that time, Taa ≦ a / h <
Tab ≦ b / h <Tbc ≦ c / h <Tcd ≦ d / h <T
The relationship of dd is established.

From the above-mentioned relationship, an example in which the interpolation means 24 cuts out a numeral portion which cannot be cut out by the series designated number cutting means 19 will be described with reference to the case shown in FIG. 7A. The example of FIG. 7A is a case where three out of four characters can be cut out, and among the characters that can be cut out, for example, the interval between the first character and the second character from the left is X.
If the distance between the second and third characters is expressed by X23, the distance X12 and the distance X23 are given by the following conditions: Tbc ≦ X12 / h <Tcd and Taa ≦ X23 / h.
When <Tab is satisfied, it can be estimated that there is a character between the first character and the second character, and in this case, the interpolating means 24 cuts out the corresponding part according to the position information obtained in advance.

Also in the case of FIGS. 7B to 7E, FIG.
It is possible to estimate that there is a defective portion by the same method as in the case of (a), and this defective portion can be cut out by the interpolating means 24. Also, the candidate character is 5
When there are more than one character, a threshold value (Taa to Tdd) can be used to select a candidate that matches the sequence of the designated sequence numbers.

The size of the designated number and the number interval are
It depends on whether the license plate is medium format or large format. For example, when each value of a / h to d / h is represented by (medium format, large format), a / h (0.7, 0.7), b / h (1.1, 0.9), c / h
(1.8, 1.6) and d / h (2.4, 2.3) can be adopted. In this case, Taa = 0.
6, Tab = 0.8, Tbc = 1.4, Tcd = 2.0
And the thresholds of Tdd = 2.7 can be used together.

According to the second specific example, even if there is a defect in the cut-out by the series designated number cut-out means 19 due to external influences such as weather, stains or shadows on the plate, the interpolation means 24 cuts out the number of the defective portion. It is possible to increase the reliability of the work for cutting out the designated number.

<Third Concrete Example> A third concrete example shown in FIG.
Basically, an example in which the character candidate extracting means is incorporated in the second specific example shown in FIG. 5 is shown. The license plate series designated number cutting-out device 110 shown in FIG. 5 is a character candidate extracting means 26 for extracting a character candidate area from a multi-valued digital image stored in the image memory 13, and a character candidate extracted by the character candidate extracting means 26. With respect to the area, a judging means 28 for judging whether or not it is a character using the dictionary 27, and a character position verifying means for comparing and verifying the result of this judging means 28 with the result cut out by the serial designated number cutout means 19. And 29.

The plate position detecting means 14 has a flag of 0 unless the character position verifying means 29 receives a flag = 1 for specifying the detection range from the character position verifying means 29 for the digital image from the A / D converter 12. As in the first and second specific examples, the differential processing is applied to the entire surface of the image, and the plate image position is detected from the entire image area. However, when the plate position detecting means 14 receives the flag = 1 signal from the character position verifying means 29, the plate position detecting means 14 performs a plate detecting operation on a predetermined point centered on the character candidate position designated by the determining means 28.

The character candidate extracting means 26 is provided in the image memory 13
The adjacent pixels within a predetermined density difference are connected from the multi-valued digital image stored in to generate blocks of equal density areas. When the area S, the height H and the width W of the circumscribed rectangle of each block are within the predetermined threshold range, the block is regarded as a character candidate area.

The judging means 28 determines the character candidate area extracted by the character candidate extracting means 26 by the well-known pattern matching method using the character patterns stored in the dictionary 27, that is, the numeral patterns in the illustrated example. , Matching the number pattern of the dictionary 27,
Non-characters are excluded from the character candidate area based on a certain threshold.

The character position verification means 29 determines, as a character candidate area, one of the circumscribed rectangles of the block that has passed through the determination means 28.
A block located within the plate image obtained by the plate position detecting means 14 and having a block center position in the lower half of the plate image is selected as a series designated number candidate. Further, the character position verification means 29 determines whether the center position of each block exists within the circumscribed rectangle of each character cut out by the series designated number cutout means 19. Based on this determination, the character position verification means 29
Of the blocks obtained by, the center of the block is located outside the circumscribed rectangle of the sequence number candidate obtained by the sequence designated number cutout unit 19, and does not exist within the circumscribed rectangle of the sequence designated number cutout unit 19. The block is added to the serial number candidates obtained by the serial designated number cutout unit 19.

Further, when the plate position detecting means 14 detects a flag = 0 and the sequence designation number does not cut out a single character, and the determining means 28 detects a block in the character candidate area, the character is detected. Position verification means 2
9 returns the flag = 1 to the plate position detecting means 14.
As a result, the plate position detecting means 14 causes the determining means 2
The character position verification means 29 obtains and receives the information about the character candidate area obtained by the step 8, and performs the detection operation in the vicinity of this area.

As described above, according to the third specific example, the character detecting function of the plate position detecting means 14, the character candidate extracting means 26, the determining means 28 and the character position verifying means 29 are used.
When used in combination with the character detection function by, it is possible to further improve the reliability of the work of cutting out the serially designated number of the license plate.

The license plate series designated number cutting device according to the present invention can be incorporated therein as a part of a license plate recognition device necessary for travel time estimation, automatic charge collection, and parking lot management on traffic control.

According to the license plate series designated number cutting-out device of the present invention, the threshold value is set between the maximum value and the minimum value of the density profile created along a straight line crossing the license plate area of the image. Using 2
Since the binarization process makes it possible to obtain a stable binarized image regardless of the size of the license plate area of the image, it is not strongly affected by external influences such as weather or stains on the plate, and the image Stable cutting can be performed regardless of the size of the license plate cut from. Further, according to the license plate series designated number cutout device, even if the character portion cannot be cut out by the cutout means due to dirt or the like, the character which the interpolation means cannot cut out from the cutout result cut out by the cutout means. Since the position is estimated and the character position is cut out, it is possible to reliably cut out the license plate serial designation number regardless of whether the plate is dirty or not.

[Brief description of drawings]

FIG. 1 is a block diagram of a license plate series designated number cutting device according to a first specific example.

FIG. 2 is an explanatory diagram (No. 1) showing an example of a density profile by a profile creating means.

FIG. 3 is an explanatory view (No. 2) showing an example of the density profile by the profile creating means.

FIG. 4 is an operation flow chart of a threshold setting means.

FIG. 5 is a block diagram of a license plate series designated number cutting device according to a second specific example.

FIG. 6 is an explanatory diagram of a license plate standard.

FIG. 7 is an explanatory diagram showing an example of a number array supplemented by an interpolating action.

FIG. 8 is a block diagram of a license plate series designated number cutting device according to a third specific example.

[Explanation of symbols]

10, 110 Number plate serial designation number cutting device 11 Video input means 12 A / D converter 13 Image memory 14 Plate position detecting means 15 Profile creating means 16 Extreme value detecting means 17 Threshold setting means 18 Binarization processing means 19 Series designation Number cut-out means 21 Horizontal direction line 24 Interpolation means 26 Character candidate extraction means 27 Dictionary 28 Judgment means 29 Character position verification means

Claims (6)

[Claims] 1. A density profile creating means for creating a density profile along a straight line crossing a license plate area of an image in which a license plate area is specified, and an extreme value detection for finding a maximum value and a minimum value of the obtained density profile. Means, threshold value setting means for setting a threshold value from the minimum value and the maximum value obtained by the detecting means, a binarization processing means for performing a binarization process on the image based on the set threshold value, A license plate serial designated number cutting-out device, which includes a cutting-out unit that cuts out a serial designated number of a license plate from a binarized image obtained by the binarizing processing unit. 2. The maximum value and the minimum value when the maximum minimum value is smaller than the minimum maximum value among the plurality of minimum values and maximum values obtained by the detection means. The license plate serial designation number cutting device according to claim 1, wherein the threshold value is set in a range between the maximum value and the maximum value. 3. The threshold setting means is a region where the maximum value exceeds the minimum value when the maximum minimum value is greater than or equal to the minimum maximum value among the plurality of minimum values and maximum values obtained by the detection means. , And for each divided area, in the range between the maximum minimum value and the minimum maximum value for each area,
The license plate series designated number cutting-out device according to claim 1, wherein the threshold value is set. 4. A binarizing unit for binarizing an image in which a license plate area is specified, and 2 by the binarizing unit.
It includes a cutting-out unit that cuts out a series of designated numbers of the license plate from the binarized image, and a cutting-out interpolation unit that estimates a character position that could not be cut out from the cutting-out result cut out by the cutting-out unit and cuts out the character position. , License plate series specified number cutting device. 5. The cutout of the estimated character position by the interpolating means, the ratio between the center distance of each character of the license plate obtained in advance and the vertical dimension thereof, and a threshold value based on the ratio, and the character cut out by the cutout means. 5. The license plate series designated number cutting-out device according to claim 4, wherein the presence / absence of a character that cannot be cut out by the cutting-out means is judged by comparing the ratio between the character spacing and the vertical dimension of the character. 6. A character candidate extracting means for extracting character candidates from the image before the binarization processing, and a judging means for rejecting non-character candidates from the character candidates extracted by the character candidate extracting means. 5. The license plate series designated number cutting-out device according to claim 4, further comprising: a character position verification means for comparing the judgment result by the judging means with the character position information obtained by the cutting means.