JPH03296181A - Character array extracting method - Google Patents

Abstract

PURPOSE:To extract an array of serial characters even when the characters differ in image pickup level by extracting a character array which matches a read object number from all rectangles which are extracted from a binary image having plural levels. CONSTITUTION:A character extraction part 6 which extracts characters from an image binarized by a binarization processing part 5 extracts the characters of the number of a moving vehicle in the character-extracted image and the matching character array as to all the rectangles extracted from binarization picture by plural level. Therefore, even if a series of characters to be extracted differ in image pickup level individually, binarization is carried out with binarization levels matching the respective characters to extract the rectangles and when the rectangles are extracted to some extent, the character array extraction is carried out as to all the rectangles. Consequently, even when the series of characters are not binarized with the same level, the character array can be extracted.

Description

[Detailed Description of the Invention] [Field of Industrial Application] This invention is a vehicle number system that automatically recognizes the displayed vehicle number or train number by imaging the 11i side or rear side of a moving car, for example, a train of cars. The present invention relates to a character arrangement extraction method for a reading device.

[Prior Art 1] FIG. 4 is a flowchart of a conventional character array extraction method. In the figure, Sl is an image capture input step that inputs a multivalued image from a camera, and S2 is a step in which a specific binary level is determined from the captured image input in step S1, and pixels that are whiter than that level are set to white, and pixels that are darker than that level are set to black. Convert to 2
Value conversion processing step S3 is a rectangle extraction processing step for extracting a rectangle that seems to be one character according to the size of the character of the double number corresponding to character extraction from the image binarized in step S2, S6 is extraction in step S3 S7 is a character array determination processing step of determining whether a valid character array has been found in step S6. is step S7
A binarization count processing step in which the number of binarizations is counted in order to determine whether to perform binarization again if the result is NG (No Good).

S9 is a binarization level set processing step that sets another binarization level when it is determined in step S8 that binarization should be continued, and SIO is a character recognition processing step that is performed when it is OK in step S7.

FIG. 5 shows an example of character string extraction using the flow described above. Figure 5 fal fb) is a multivalue level 0 to 40
This is an example of a captured image when the image is captured.

FIG. 5(a) is an example of imaging when the imaging level is constant.

Fig. 5 fb) is an example of imaging when the imaging level is not constant on the left and right sides of the image, and in both cases the average density of all pixels is approximately 2.
It is 2.5. Figure 5 fcl (dl is Figure 5 fat (
FIG. 5 is an example of a binarized image when each captured image of bl is binarized at a binarization level of 20.
Figure (cl) shows a rectangle extracted from the binarized image of (d).

When the imaging level is constant as shown in Figure 5 (a), the two characters l 5111141+ are cut out as shown in Figure 5 tel, but on the left and right side of the image as shown in Figure 5 (b). When the imaging level is not constant, only one character "2°" out of the characters "3°l I+ 21" is extracted. Even if binarization is performed at a different binarization level, Fig. 5fb
In the case of l, there is no binarization level in which the two characters 3°'°"2" appear in both degrees, so only one of them can be extracted.

[Problems to be Solved by the Invention] Conventional character array extraction methods do not take into consideration the case where each character to be extracted has a different imaging level. The character array was extracted only for . However, in reality, even in a series of characters, the imaging level of each character is often different.
There was a problem in that character sequences could not be extracted frequently.

This invention was made to solve the problem,
The object of the present invention is to obtain a character array extraction method that can extract a character array of a series of characters even when each character has a different imaging level.

[Means for Solving the Problems] A character array extraction method according to the present invention targets all rectangles extracted from a binarized image at multiple levels, and extracts the number of the moving vehicle on the character extraction side from among the rectangles. It extracts a character sequence that matches a character sequence.

[Operation] In this invention, 9 characters are extracted from all rectangles extracted from binarized images at multiple levels in order to extract the characters of the vehicle number and the matching character sequence. Even if the imaging level of each character of the corresponding vehicle number is different, each character is extracted at a binarization level suitable for each character and becomes a target when extracting a character array.

[Embodiment 1] FIG. 1 is a block diagram of a vehicle number reading device to which the character arrangement extraction method according to the present invention is applied. In the figure, (1) is a position sensor that detects when a running automobile or train has reached or passed a specified position and generates a trigger signal.

(2) is a strobe that emits light in response to a trigger signal from the position sensor (1), and (3) is a camera with a shutter that captures an image of the front or rear of a car or train in response to a trigger signal from the position sensor (1). , (4) is this camera (
3) a frame memory that stores the input image from (
5) is a binarization processing unit that binarizes the image stored in this frame memory (4), and (6) is a binarization processing unit that binarizes the image stored in this frame memory (4). A character extraction unit that extracts characters from.

(7) is a character recognition unit that recognizes the characters extracted by this character extraction unit (6), and (8) is this character recognition unit (7).
(9) is a monitor that displays the recognition results output from the result output unit (8); (101 is a CPU that controls each of the above units);

FIG. 2 is a flowchart of a character sequence extraction method according to the present invention. In the figure, 5l-33, S6 to 510 are equivalent to the same numbers in FIG. S4 is a binarization count processing step in which the number of binarizations is counted in order to determine whether to perform binarization again after the completion of the rectangle extraction processing step S3, and S5 is a binarization count processing step in which characters are imaged at various levels. Taking into account that
This is a valorization level set.

When character array extraction is performed according to the above flow, character array extraction processing step S6 is performed for all rectangles extracted by binarization the number of times until it is determined that the binarization is completed in step S4 of counting the number of times of binarization. become.

FIG. 3 shows an example of character string extraction using the flow described above. In the figure, Figure 3 fal is a multilevel level 0 to 4.
This is an example of a captured image when the image is captured at 0.
)) This is an example of imaging when the imaging level is not constant between the left and right sides of the same image. FIG. 3 shows two binarization levels set in the binarization level set S5 for bl (cl is each captured image fa in FIG. 3).

This is an example of a binarized image when binarized at 22 and 20. In the binarization level set S5, the average density of all pixels is 2.
2.5 to 1 For example, it is assumed that the binarization level 22 is set to the average value as is (the decimal point is rounded down), and the binarization level 20 is set to the average -2 (the decimal point is rounded down). Fig. 3 fd) (e) is the binarized image (Fig. 3 fbl (
cl ) is shown.

From the binarized image in Figure 3 fb), only one character '3' is extracted, and from the binarized image in Figure 3 fc), only one character '2' is extracted, but the single character sequence extraction process Step S6
Now, we will extract the character array for both of these characters, so
The two-character string 3゛゛°2°゛ is correctly extracted.

The character array extraction method according to the present invention is performed as described above, and character array extraction is performed for characters that are binarized at various levels.

[Motion of the invention 1 As described above, according to the present invention, even if each character in a series of characters to be extracted has a different imaging level, each character can be binarized at the binarization level appropriate for each character. Then, when the extracted rectangles have accumulated, character array extraction is performed for all the rectangles. Therefore, a character arrangement can be extracted even if a series of characters are not necessarily binarized at the same binarization level, which is highly effective.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a vehicle number reading device to which the character sequence extraction method of the present invention is applied, Fig. 2 is a diagram showing the flow of the character sequence extraction method of the present invention, and Fig. 3 is a diagram showing the character sequence extraction method of the present invention. FIG. 4 is a diagram showing the flow of a conventional character array extraction method, and FIG. 5 is a diagram showing an example of character array extraction using the conventional character array extraction method. In the figure, (11 is a position sensor, (2) is a strobe, (
3) is a camera, (4) is a frame memory, (5) is a binarization processing unit, (6) is a character extraction unit, (7) is a character recognition unit,
(8) is the result output section, (9) is the monitor, (101 is [
:PU, Sl is the imaging input step, S2 is the binarization processing step, S3 is the rectangle extraction processing step, S
4 is a step for counting the number of times of binary conversion, S5 is a step for setting the level of binary conversion/ring step, S6 is a step for character sequence extraction processing, S7 is a step for character sequence extraction judgment processing, S8 is a step for counting the number of times of binary conversion, S9 is a two-step refresh step, and S10 is a character recognition processing subroutine. Note that the same line numbers in Figure 1 indicate the same or equivalent parts.

Claims (1)

[Claims] A position sensor that generates a trigger signal when a running vehicle reaches a specified position, a strobe that emits light upon receiving the output of the position sensor, and a camera with a shutter that captures an image upon receiving the output of the position sensor. A frame memory that stores an input image from this camera, a binarization processing unit that binarizes the image stored in this frame memory, and a binary image output from this binarization processing unit. A character extraction unit that extracts characters from the character extraction unit, a character recognition unit that recognizes the characters extracted by this character extraction unit, a result output unit that outputs the results recognized by this character recognition unit, and an output of this result output unit. In a vehicle number reading device that is equipped with a monitor that receives and outputs the results, and a CPU that controls each of the above parts, and that reads the vehicle number or train number on the front or rear of the vehicle every time a vehicle passes, it is especially useful for extracting character rectangles. When performing this, characters that match the number to be read are extracted from all rectangles extracted from binary images at multiple levels, taking into consideration cases where the imaging level of each character of the number to be read is not constant. A character array extraction method characterized by extracting an array.