JPH0827818B2 - Rotation angle detection method for character strings - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial application] The present invention extracts a target pattern such as a character string with an image sensor, and rotates the target pattern in an apparatus for recognizing and inspecting the image by subjecting it to image processing. The present invention relates to a corner detection method.

[Conventional technology]

Conventionally, as this type of detection method, the width of the pattern is detected by projecting the binarized image of the target pattern from many directions, and the direction in which the detected width is the minimum is the direction of the arrangement of the character strings. It is known that the rotation angle is detected.

[Problems to be solved by the invention]

However, since this detection method requires projection from many directions, it takes much time to improve the detection accuracy, and cannot be applied when high-speed and high-precision processing is required. There was a point.

Therefore, an object of the present invention is to provide a rotation angle detecting method capable of detecting a rotation angle of a target pattern such as a character string at high speed and with high accuracy.

[Means for solving problems]

In order to detect the rotation angle of the character string from the binarized image information obtained by picking up the image of the character string and binarizing it for each pixel, scanning is performed from a direction substantially at right angles to the character string, and a change point forming a boundary with the background. When the last (first) change point is sequentially extracted on each scanning line, the maximum point (minimum point) at which the difference between adjacent change points changes from increase (decrease) to decrease (increase) At least one is extracted for each character, and only those that satisfy a predetermined condition (those expected to be aligned on one straight line) are extracted from these maximum points (minimum points), and the direction of the point group is extracted. The rotation angle is detected by obtaining

[Action]

The final (first) change point on each scan line is extracted from the change points that form the boundary with the background by scanning the character string from the direction substantially perpendicular to it, and the difference between adjacent change points increases (decreases). ) To decrease (increase), the maximum point (minimum point) is found. Since there is always one maximum point (minimum point) for each character, there is a maximum point (minimum point) for each character.
Therefore, the rotation angle can be detected at high speed and with high accuracy by determining only the direction of those that are presumed to be aligned on a straight line.

〔Example〕

FIG. 1 is a flow chart showing an embodiment of the present invention, FIG. 2 is a block diagram showing an image processing apparatus to which the present invention is applied, and FIG. 3 is an explanatory diagram for concretely explaining the embodiment.

First, FIG. 2 will be described. In the figure, 1 is an image sensor such as a television camera for inputting an image of a target pattern, 2 is a binarization circuit, 3 is a binary image memory, 4
Is an address generation circuit for generating an address when scanning the binary image memory 3, 5 is a change point extraction circuit, 6 is a data memory for storing change point data obtained by the change point extraction circuit 5, and 7 is a processor ( CPU) and 8 are buses of CPU7.

The image of the target pattern is input by the image sensor 1, the obtained image information is binarized by the binarization circuit 2, and the obtained binarized image is stored in the image memory 3.
Next, the binarized image stored in the image memory 3 is supplied with necessary addresses from the address generating means 4 and read out, and the change point data obtained by inputting to the change point extraction circuit 5 is stored in the data memory 6. To be done. The change point data in the data memory 6 is read by the CPU 7 through the CPU bus 8 and processed to calculate the rotation angle.

Next, the embodiment will be described in detail with reference to FIG. 2A and 2B are binarized images of the target pattern, 10 is the target pattern (character string), 11 is the scanning direction (y-axis direction) when extracting the change point, and 12 is each scanning line. The final change point of is shown (this is what looks like a beard in the figure). Also,
P ₁ , P ₂ , ..., P _i , P _{i + 1} , P _{i + 2} , ......, P _n-1 , P _n are points where the scanning direction coordinates are maximum, L ₁ , L ₂ , ...... , L _i , L _{i + 1} , L _{i + 2} , ......, L _n-1
Are line segments , Θ ₁ , θ ₂ , ..., θ _i , θ _{i + 1} , θ _{i + 2} , ..., θ _n-1 are L ₁ , L ₂ , ..., L _i , L _{i + 1} , _Let L _{i + 2} , ......, L _{n-1 be} the angle
_{i + 1} ′ is a line segment Indicates. Note that θ ₁ , θ ₂ , ... Are angles measured from the x-axis of the line segments L ₁ , L ₂ ,.

First, the binary image of the target pattern is sequentially scanned in the scanning direction 11, and the final change point 12 of each scanning line is extracted. From the point sequence of the extracted change points, all the points whose coordinates in the scanning direction are maximum are extracted (P ₁ , P ₂ , ..., P _i , ...,
P _n ). At this time, the maximum points are sequentially obtained as the differences between the final change points adjacent to each other, and are extracted as the points at which the differences change from increase to decrease. Accordingly, the character "7" in FIG. 3 (a), for example P _1, P ₂ points are extracted as the local maxima. Next, these points P ₁ , P ₂ , ..., P _i , ..., P _n are connected in this order by line segments L ₁ , L ₂ , ..., L _i , ..., L _n-1 , and L ₁ , L ₂ ,…,
The angles θ ₁ , θ ₂ , ..., θ _i , θ _n-1 of L _i , ..., L _n-1 are obtained. Then, the following operation is performed for each line segment.

(1) For i = 1,2, ..., n-2, if θ _i −θ _{i + 1} <α for a small constant angle α, P _{i + 1} is excluded from the maximum point, and P _The line segment L _{i + 1} ′ connecting _i and P _{i + 2} is set as a new L _{i + 1} , and the direction θ _{i + 1} ′ is obtained. The operation (1) is repeated until there is no maximum point to be excluded, and then the next operation is performed.

(2) If θ ₁ −θ ₂ > −α, P ₁ is excluded from the maximum point.

(3) If θ _n-2 −θ _n-1 > −α, P _n is excluded from the maximum point. Regarding the maximum points remaining at the time when the operations (1) to (3) are completed, the direction of the arrangement of the maximum points is obtained by the least squares method, and this is taken as the rotation angle.

The above processing is summarized in a flowchart in FIG. That is, the final change point is extracted (), the maximum point is obtained (), and the direction of the line segment connecting the maximum points in sequence is obtained (). Next, the process (1) is performed (-), then the process (2) (-), the process (3) (,
). For the last remaining maximum points, the direction of the arrangement of the maximum points is obtained by the least squares method ().

Next, a second embodiment will be described with reference to FIG. (A) and (b) of the figure are binarized images of the target pattern, 10 is the target pattern (character string), 11 is the scanning direction when extracting the change points, and 13 is the first change point of each scan line. Indicates. Also,
Q ₁ , Q ₂ , ..., Q _i , Q _{i + 1} , Q _{i + 2} , ......, Q _m-1 , Q _m are points where the coordinates in the scanning direction are minimal, l ₁ , l ₂ , ... …, L _i , l _{i + 1} , l _{i + 2} , ……, l
_m-1 is a line segment , Η ₁ , η ₂ , ..., η _i , η _{i + 1} , η _{i + 2} , ..., η _m-1 are l ₁ , l ₂ , ......, l _i , l _{i + 1} , l _{i + 2} , ......, l _m-1 angle, l
_{i + 1} ′ is a line segment Indicates. Further, how to take the angles η ₁ , η ₂ , ... Is the same as in the case of FIG.

First, the binarized image of the target pattern is sequentially scanned in the scanning direction 11, and the first changing point 13 of each scanning line is extracted. From the extracted point sequence of change points, all points at which the coordinates in the scanning direction are the minimum are extracted (Q ₁ , Q ₂ , ..., Q _i , ...,
Q _m ). This extraction method is the same as in FIG. 3, but here the point at which the difference changes from decreasing to increasing is extracted as the minimum point. Next, Q ₁ , Q ₂ , ..., Q _i , ..., Q _m are line segments l ₁ , l ₂ ,
..., l _i, ..., connected by _{l m-1, l 1,} l 2, ..., l i, ..., each angle eta ₁ of _{l m-1, η 2,} ..., η i, η m-1 Ask for. Then, the following operation is performed for each line segment.

(4) For i = 1, 2, ..., M-2, η _i −η _{i + 1} > −
If α, Q _{i + 1} is excluded, and a line segment l _{i + 1} ′ connecting Q _i and Q _{i + 2} is newly set as l _{i + 1} to obtain the direction η _{i + 1} ′. The operation (4) is repeated until there is no minimum point to be excluded, and then the next operation is performed.

(5) If η ₁ −η ₂ <α, exclude Q ₁ from the minimum point.

(6) If η _n-2 −η _n-1 <α, exclude Q _n from the minimum point. With respect to the minimum points remaining at the time when the operations (4) to (6) are completed, the direction of the arrangement of the minimum points is obtained by the least square method, and this is set as the rotation angle.

FIG. 4 summarizes the above processing in a flowchart. The first change point is extracted (), the minimum point is obtained (), and the direction of the line segment connecting the minimum points in sequence is obtained ().
Next, the process (4) is performed (-), then the processes (5) (-) and the process (6) (,) are performed. For the last remaining minimum point, the direction of the arrangement of the minimum points is calculated by the method of least squares ().

EFFECTS OF THE INVENTION According to the present invention, the rotation angle can be detected with high accuracy by one image scan for obtaining the change point of the binary image of the target pattern and processing of the obtained one-dimensional data. Because you can
The rotation angle of a target pattern such as a character string can be detected at high speed and with high accuracy, and there is an advantage that a highly flexible inspection can be realized.

[Brief description of drawings]

FIG. 1 is a flow chart showing a first embodiment of the present invention,
2 is a block diagram showing an image processing apparatus to which the present invention is applied, FIG. 3 is an explanatory diagram for specifically explaining the first embodiment of the present invention, and FIG. 4 is a second embodiment of the present invention. FIG. 5 is an explanatory diagram for specifically explaining the second embodiment of the present invention. Explanation of symbols 1 ... Image sensor, 2 ... Binarization circuit, 3 ... Image memory, 4 ... Address generation circuit, 5 ... Change point extraction circuit,
6 ... Data memory, 7 ... Processor (CPU), 8 ...
… Bus, 10… Target pattern (character string), 11… Scanning direction, 12… Final change point, 13… First change point.

Claims (2)

[Claims] 1. A detection method for detecting a rotation angle of a character string from binarized image information obtained by imaging a character string and binarizing it for each pixel. Scanning from a predetermined direction, and sequentially extracting the final change point on each scan line among the change points forming a boundary with the background; (b) the scanning direction coordinates of the change points adjacent to each other. A step of obtaining a local maximum point group by extracting all the points at which the difference changes from increasing to decreasing as local maximum points (P _{1 to} P _n ); and (c) scanning these local maximum points (P _{1 to} P _n ). Line segments (L ₁ ~
L _n-1 ) multiple line segments obtained by connecting them Steps for calculating the angles (θ _{1 to} θ _n ) for all of them, and (d) when i = 1, 2, ...
If the angle difference between the two line segments is a preset constant angle α: If θ _i −θ _{i + 1} <α, then P _{i + 1} is excluded from the maximum point,
The operation of setting the line segment L _{i + 1} ′ connecting P _i and P _{i + 2} as a new L _{i + 1} is repeated until there is no maximum point to be excluded, and if θ ₁ −θ ₂ > −α For example, P ₁ is excluded from the maximal point, and if θ _n-2 −θ _n-1 > −α, P _n is excluded from the maximal point. A step of extracting only local maxima that are presumed to be aligned on a straight line; and (e) determining the direction of the local maxima remaining by the least squares method at the time when the above series of processing operations are completed. Is detected as the rotation angle of the character string, and the rotation angle detection method of the character string is provided. 2. A detection method for detecting a rotation angle of a character string from binarized image information obtained by imaging a character string and binarizing it for each pixel. Scanning from a predetermined direction and sequentially extracting the first change point on each scanning line among the change points forming a boundary with the background; (b) the scanning direction coordinates of the change points adjacent to each other. A step of obtaining a minimal point group by extracting all the points where the difference changes from decreasing to increasing as minimal points (Q _{1 to} Q _m ), and (c) scanning these minimal points (Q _{1 to} Q _m ). Line segments (l ₁ ~
multiple line segments obtained by connecting with l _m-1 ) For all of them, the step of calculating the angles (η _{1 to} η _m ) and (d) when i = 1, 2, ...
If the difference between the angles of the two line segments is η _i −η _{i + 1} <−α with respect to a preset constant angle α, then Q _{i + 1} is excluded from the minimum point and Q _i , The operation of setting the line segment l _{i + 1} ′ connecting Q _{i + 2} as a new l _{i + 1} is repeated until there are no minimum points to be excluded, and if η _i −η ₂ <α, then Q ₁ Is excluded from the minimum point, and if η _m-2 −η _m-1 <α, then Q _m is excluded from the minimum point. A step of extracting only the inferred minimum points, and (e) determining the direction of the arrangement of the minimum points remaining at the time when the series of processing operations described above is completed by the method of least squares, and rotating this to rotate the character string. A method for detecting a rotation angle of a character string, the method comprising: detecting as a corner.