JP3080097B2 - Parallel line figure extraction method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION The present invention relates to a method for extracting a parallel line figure included in image data such as a map and a machine drawing read by a computer.

[Conventional technology]

In the process of recognizing and understanding images such as maps and mechanical drawings with a computer, especially in the process of recognizing map figures and extracting road figures, parallel lines are extracted from images containing various types of line figures. Extraction processing is required.

As described above, the method of extracting parallel lines is being studied in, for example, extracting a road figure or the like during map figure processing. As a method of tracking a parallel line, a method of taking the inner product of both vectors while tracking the inside of the parallel line in parallel with two vectors and extracting the parallel line (IEICE Transactions, vol.J67D No12, pp1
1429-11426, 1984.12), by taking the contour of a pixel so that the pixel is always on the right side, the direction of the vector (pair vector) which is a candidate for the parallel line is reversed by using the opposite direction. There is a way to ask (AUTOCARTO JAPA
N Papers, 62).

[Problems to be solved by the invention]

Extraction of parallel lines requires high-speed processing because of the accuracy and the enormous amount of data to be handled. However, in any of the above-described methods, generally, parallel line detection processing is difficult, processing is complicated, and high-speed processing is difficult.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a parallel line graphic extraction method that can accurately extract parallel lines by combining basic processes and can perform high-speed processing by hardware or parallel calculation.

[Means for solving the problem]

According to the parallel line graphic extraction method of the present invention, a load matrix is prepared in which a central portion is a positive value and a peripheral portion is a negative value in a straight line, and convolution integration is performed by a product-sum operation of the load matrix and image data. A first step of extracting a line figure in a specific direction from the image data by performing a threshold process on the result obtained in the first step to set a pixel value smaller than a predetermined threshold to 0; In the image data obtained in the second step and the second step, three inspection points are set at positions equidistant from each other on the same line in an azimuth perpendicular to the azimuth of the load matrix, and pixels at the inspection points at both ends are taken. By performing a logical OR operation between the values and performing a logical AND operation between the result of the logical OR operation and the pixel value at the central inspection point, a third method for extracting only a line figure at a specific interval
And a step.

Another feature of the parallel line graphic extraction method of the present invention is that a load matrix is prepared in which a central portion has a positive value and a peripheral portion has a negative value in a straight line. A first step of performing convolution integration by a multiply-accumulate operation, and performing a threshold process on the result obtained in the first step to set a pixel value smaller than a predetermined threshold to 0, thereby obtaining image data in the image data. A second step of extracting a line figure having a specific azimuth; and in the image data obtained in the second step, three inspection points at azimuths perpendicular to the azimuth of the load matrix and at equal distances from each other on the same line. By performing an exclusive OR operation between the pixel values at the inspection points at both ends, and performing an AND operation between the result of the exclusive OR operation and the pixel value at the central inspection point to obtain a line at a specific interval. And having a third step of extracting shape only.

Another feature of the parallel line graphic extraction method of the present invention is that a plurality of the load matrices are prepared, and convolution integration is performed on the image data with each of the load matrices, and a threshold process is performed. It is characterized in that a line figure in a plurality of directions is extracted from the data, and a logical sum operation or an arithmetic sum operation is performed between pixels at the same position with respect to each image data from which the line figure in a specific direction is extracted.

[Action]

The parallel line extraction method of the present invention extracts a parallel line figure by convolution integration of image data and a load matrix and threshold processing. Furthermore, only parallel line figures at specific intervals are extracted by a predetermined logical operation.

That is, a convolution integral using a load matrix for straight line detection and threshold processing are executed to extract a line figure in a specific direction. Next, with respect to the image data obtained by extracting the line figure in the specific direction, three inspection points are set at positions equidistant from each other on the same line in a direction perpendicular to the direction of the load matrix, and pixel values of these inspection points are determined. By performing a logical operation between them, only parallel line figures at specific intervals in specific directions in the image data are extracted.

If the above processing is performed for a plurality of directions and the sum or the logical sum of the results is calculated by operation between the image data, parallel lines in all directions can be extracted.

〔Example〕

FIG. 1 is a diagram for explaining an embodiment of the present invention, and FIG. 2 is a diagram showing the width (interval between parallel lines) of the present embodiment.
Shows the result of performing a process of extracting a parallel line having a width of 12 pixels from a concentric graphic (a) having 6, 12, and 24 pixels, respectively.

In FIG. 1, input image data (FIG. 2 (a)) to be processed and read as a grayscale image by an image reading device such as an image scanner (not shown) is input to each of eight parallel line extracting units. These parallel line extraction units 1 to 3 perform convolution integration and threshold processing, which will be described later, on the input image data, thereby extracting line figures in specific directions. FIG. 2 (b)
For example, this is a result of extracting a line figure in the vertical direction by the parallel line extraction unit. Here, depending on how the threshold value is determined, a certain margin is given to the direction to be extracted, so that not only a strictly vertical direction but also a line figure having a substantially vertical direction is extracted.

Next, each of the parallel line extraction units to performs a logical operation, which will be described later, on the image data after the threshold processing, thereby extracting only parallel lines at a specific interval. FIG. 2 (c)
This is the result of performing a logical operation on the line figure in the vertical orientation extracted as shown in FIG. (B) and extracting only a parallel line having a width of 12 pixels.

As described above, one parallel line extracting unit can extract a parallel line figure in one azimuth. In this embodiment, in order to extract parallel line figures in all directions, the above convolution integration, threshold processing, and logical operation processing are performed for a plurality of directions by the parallel line extraction unit to and the sum or logical sum of the results is imaged. By processing in the interval calculation unit, a concentric figure (FIG. 2 (d)) consisting of parallel lines having a width of 12 pixels is extracted.

FIG. 3 is a diagram for explaining a processing procedure of each unit of the parallel line extracting unit to FIG. Parallel line segments are detected by the following convolution integral 1, threshold processing 2, and logical operation 3.

First, a line figure in a specific direction is extracted by convolution integral 1 and threshold processing 2. In the convolution integral 1, a product-sum operation with the input image data is executed using a load matrix for detecting a straight line, which is sufficiently smaller than the image size as shown in FIG. At this time, the load-sum matrix is scanned from the end to the end on the screen to execute a product-sum operation with the input image data.

Each element of the load matrix has a positive value at the center and a negative value at the periphery in a straight line, and these values are in a direction perpendicular to the azimuth of the matrix (vertical azimuth in the example of FIG. 4). And the difference between the two Gauss functions (DOG function: (Equation 1)).

Where σ ₁ , σ ₂ :
Standard deviation of the Gauss function (σ ₂ / σ ₁ = 1.6), σ _y : standard deviation of the Gaussian function of the orientation of the matrix, x: position on the matrix in the orientation perpendicular to the orientation of the matrix (x = 0 at the center of the matrix) ), Y: position on the matrix in the orientation of the matrix (y = 0 at the center of the matrix). It is preferable that the positive value area width and the negative value area width of the load matrix correspond to the line width of the figure included in the image data to be processed.

In the convolution integral, a weight matrix of size M × N is created by (Equation 1) (N / 2 = 3σ), and the product sum is calculated between the weight matrix and the image data. When the orientation of the load matrix is the vertical orientation shown in FIG.
If the image data to be processed is f (i, j), the load matrix is w (i, j), and the resulting image data is g (i, j), Becomes

In the threshold processing 2, a value equal to or less than a predetermined threshold is set to 0 for the image data after convolution integration. This allows
A line segment having the same or substantially the same direction as the direction of the load matrix is detected. That is, in the case of a line segment having substantially the same direction as the direction of the matrix, the line segment greatly overlaps with a region of a positive value in the matrix, and the value obtained by the convolution integral 1 is larger than the determined threshold. And remains as it is. On the other hand, in the case of a line segment having an azimuth different from the azimuth of the matrix, since both the positive value area and the negative value area in the matrix greatly overlap, the value obtained by the convolution integral 1 becomes smaller than the determined threshold. , The value is set to 0 by threshold processing 2. Also in the case of a point figure, the value obtained by the convolution integral 1 becomes small, so that it is deleted by the threshold processing 2. Therefore, at this point, a plurality of line figures for the specific direction are extracted as parallel lines.

Next, in the logical operation 3, with respect to the image data after the threshold processing, it is determined whether or not the extracted parallel lines are at a specific interval by the following method. First, as shown in FIG. 5 (a), one point of interest _xi is set in the target image data, and the point of interest is placed on the same straight line in a direction perpendicular to the direction of the above-described straight line detection matrix. test point x _i-1 either side of the two at the position of specific distance 1 of the point x _i, take x _{i + 1.}

As shown in FIG. 5 (a), if the above target point x _i are present on the parallel line segments to each other (hatched portion), because the spacing between the extracted parallel lines is a particular interval 1 Suffices that at least _one of the left and right inspection points x _i−1 and x _{i + 1} exists on another parallel line segment. Therefore, if the pixel value of point x is P (x), and P (x) is binary (0,1),
If present on one line of the parallel lines of target point x _i is a particular interval _{1, P (x i) AND} (P (x i-1) OR P (x i + 1)) = 1 ... ( Equation 3) When the pixel value P (x) is multi-valued (for example, 0 to 255), this (Equation 3) sets MIN {P (x _i ), MAX (P (x _i−1 ) as the threshold value p. , P (x _{i + 1} ))} ≧ p (Expression 4)

From the pixel values of these three points x _i , x _i-1 , x _{i + 1} , the above (Equation 3)
Alternatively, if the value of (Equation 4) is calculated and set as the output pixel value, the points of interest _xi are sequentially scanned in the image data, so that the line segments forming the parallel lines at the interval 1 are sequentially extracted. Is done.

As described above, when the logical operation is performed according to (Equation 3) or (Equation 4), when a plurality of mutually parallel line segments are arranged for each specific interval 1, as shown in FIG. All are extracted.

On the other hand, when it is desired to extract only two parallel line segments having an interval of 1 as parallel lines, the left and right inspection points x
_i-1 and x _{i + 1} should not exist simultaneously on parallel line segments. Therefore, if the pixel value P of the point x (x) is the binary, as shown in FIG. 6 _{(a), P (x i} ) AND (P (x i-1) XOR P (x i + 1 )) = 1 ... (if equation 5) and, P (x) is _{multi-valued, MIN [P (x i)} , {MAX (P (x i-1), P (x i + 1)) −MIN (P (x _i−1 ), P (x _{i + 1} ))}] ≧ p (Expression 6)

However, in a graphic like FIG. 6 (b) (the same graphic as shown in FIG. 5 (b)), even if the logical operation is performed by the above (Equation 5) or (Equation 6), the line at both ends is Remains (FIG. 6 (c)). In this case, the above processing is performed twice (after the scan of the point of interest x _i is completed once,
By scanning again from the beginning and performing the same processing), the lines at both ends can be erased.

As described above, a parallel line figure at a specific interval in one direction can be extracted using one load matrix. To extract a parallel line figure in any direction, the above processing must be performed in multiple directions. (Using a load matrix in a plurality of directions), and it is necessary to take the total or logical sum of the results (see FIG. 1).

In the embodiment shown in FIG. 1, a matrix for eight directions is prepared as a load matrix for detecting a straight line at an angle of every π / 8, the above processing is performed for eight directions, and the sum of these resultant images is calculated. However, the present invention is not limited to eight directions.

〔The invention's effect〕

As described above in detail, according to the present invention, when extracting parallel lines from a figure on an image such as a map or a mechanical drawing,
Regarding figures of any shape, product sum operation, threshold processing,
Since processing can be performed by combining basic image processing techniques such as logical operation, high-speed processing can be performed by dedicated hardware, parallel processing, or the like.

[Brief description of the drawings]

FIG. 1 is a system diagram illustrating an embodiment of the present invention, FIG. 2 is a diagram of an image showing a result of performing parallel line extraction processing at a specific interval on an image in which parallel line figures at different intervals are drawn,
FIG. 3 is an explanatory diagram of a parallel line extracting process, FIG. 4 is a diagram showing an example of a load matrix for detecting a straight line, FIG. 5 is an explanatory diagram of an example of a logical operation, and FIG. It is explanatory drawing of the example of. It should be noted that, in the reference numerals used in the drawings, 1... Convolution integral 2... Threshold processing 3... Logical operation...

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-61-118883 (JP, A) JP-A-2-16686 (JP, A) JP-A-63-193282 (JP, A) (58) Investigation Field (Int.Cl. ⁷ , DB name) G06T 7/00

Claims (3)

(57) [Claims] A first step of preparing a load matrix having a positive value at a central portion and a negative value at a peripheral portion in a straight line, and performing convolution integration by a product-sum operation of the load matrix and image data; A second step of extracting a line figure in a specific direction from the image data by performing a threshold process on the result obtained in the first step so that a pixel value less than a predetermined threshold is set to 0; In the image data obtained in the second step, three inspection points are set at positions equidistant from each other on the same line in an azimuth perpendicular to the azimuth of the load matrix, and a logical sum of pixel values at the inspection points at both ends is taken. And a third step of extracting only a line figure at a specific interval by performing an AND operation of the result of the OR operation and the pixel value at the central inspection point. Row line shape extraction method. 2. A first step of preparing a load matrix composed of a linearly linear center portion having a positive value and a peripheral portion having a negative value, and performing convolution integration by sum-of-products operation of the load matrix and image data; A second step of extracting a line figure in a specific direction from the image data by performing a threshold process on the result obtained in the first step so that a pixel value less than a predetermined threshold is set to 0; In the image data obtained in the second step, three inspection points are set at positions equidistant from each other on the same line in an azimuth perpendicular to the azimuth of the load matrix. A third step of performing only an OR operation and performing an AND operation on the result of the exclusive OR operation and the pixel value at the central inspection point to extract only a line figure at a specific interval. Parallel line shapes extraction method characterized. 3. A plurality of load matrices are prepared, and convolution integration is performed on the image data with the respective load matrices and threshold processing is performed, thereby extracting line figures in a plurality of directions from the image data. 3. The parallel line figure extracting method according to claim 1, wherein a logical sum operation or an arithmetic sum operation is performed between pixels at the same position for each image data from which a line figure in a specific direction is extracted.