JPH04303269A - Method and device for extracting contour line - Google Patents

Abstract

PURPOSE:To extract an accurate and clear contour line even when the image of an intrinsic contour line can not be extracted from noise because of a small difference of brightness between the intrinsic contour line part and the noise part or an accurate contour line can not easily be reproduced because a reflectance on the surface of an object to be photographed is high and a belt-like wide countour line is generated with an over-brightness. CONSTITUTION:A belt-like image including a contour line is obtained by picking up (S1) an image having a specific contour line, many segments crossing the belt-like image on respective positions are set up (S2), the brightness levels of the segments on respective positions are sampled (S3) along the segments, the position indicating a maximum brightness level on each segment is regarded as a representative position, and respective representative positions on respective segments are successively connected to extract (S6) the outline.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention is applicable to images where there is a small difference in brightness between the brightness of the original outline and the brightness of the noise, making it difficult to extract the original outline from the noise. Accurate and clear contour lines can be obtained even in cases where the surface of the photographed object has a high reflectance and the contour line is too bright, resulting in a band-like and wide contour line, making it difficult to reproduce an accurate contour line. This invention relates to image processing technology for extracting .

0002

BACKGROUND OF THE INVENTION In conventional image processing methods used in the above-mentioned fields, signals above a certain threshold level are treated as "1" and signals below that level are treated as "0". Background noise has been removed by processing. With just this kind of binarization processing, even if the image should originally appear as a continuous line, the dark parts of the image become discontinuous because there is little difference from the background noise. Therefore, dilation processing is performed to connect the broken lines.

[0003] The expansion process is to expand the binarized contour line data radially around the pixel, and incorporate up to eight adjacent pixels into the same data. That is, it is expanded. This dilation process is repeated until the lines are connected. At this time, if the number of times of dilation processing is reduced, the error will be reduced because unnecessary dilation will not be performed, but if the contour lines are not connected, the number of repetitions will have to be increased, which will increase the processing time. On the other hand, increasing the number of times of expansion processing not only increases processing time but also increases errors due to unnecessary expansion.

[0004] Furthermore, as a result of the above-mentioned dilation process, even though the image should originally appear in the form of clear thin lines, there are many parts where the image becomes wide, so line thinning processing is not recommended. It is being said. The thinning process shrinks the connected outline data from the outside to the inside until the width of the outline of the image becomes one pixel. Conventionally, contour lines have been extracted by such binarization processing, dilation processing, and thinning processing.

[0005]

[Problem to be solved by the invention] As in conventional binarization processing, comparison is made with a constant value threshold level.
There is a problem in that pixels with noise higher than the threshold level, that is, bright noise, cannot be removed. Furthermore, there is a problem in that pixels that are not noise but fall below the threshold level due to darkness are mistaken for noise and are removed. In this way,
There is a problem in that what should originally be a continuous linear image becomes a discontinuous, discontinuous image.

[0006] Although it is true that thin contour lines can be extracted with the dilation processing and thinning processing as described above, there is a problem of error in that the fine curvatures of the original contour lines are smoothed out, and each Since the dilation process and the thinning process must be repeated for each pixel, there is a problem with the processing time. To perform these image processing at high speed,
Another problem is that an expensive image processing device with high processing speed is required. Therefore, in the present invention, the fine curvature of the original contour line is not smoothed out, and the processing time can be shortened.
The aim is to provide highly capable image processing technology at low cost.

[0007]

[Means for Solving the Problems] To this end, in the contour line extraction method of the present invention, by capturing an image having a specific contour line (S1), a band-shaped image including the contour line is obtained; A large number of line segments that cross the image at each position are set (S2), the brightness level at each position on the line segment is sampled along the line segment (S3), and the brightness level is the maximum value on each line segment. The position indicating the line segment is set as a representative position on this line segment (S4, S5), and the representative positions on each line segment are sequentially connected to extract a contour line (S6).

The contour line extracting device of the present invention includes an imaging means 9 for capturing an image having a specific contour line to obtain a band-shaped image including the contour line, and a plurality of image pickup means 9 for capturing an image having a specific contour line to obtain a band-shaped image including the contour line; A line segment setting means 31 that sets a line segment, a sampling means 32 that samples the brightness level of an image along the line segment, and a peak position of the brightness level on each line segment is set as a representative position on this line segment. The apparatus includes representative position determining means 33 and contour line reproducing means 34 for sequentially connecting the respective representative positions on each of the line segments to reproduce the contour line.

[0009]

[Operation] According to the present invention, a band-shaped image including a contour line is traversed by a large number of line segments, and the brightness level at each position on each line segment is sampled along the line segment. Since the position showing the maximum value is taken as the representative position on this line segment, background noise on the line segment is removed and only one point showing the maximum value is determined. Then, since each of the representative positions determined one point on each line segment is sequentially connected to form a contour line, an image of a thinned contour line is obtained.

The above steps will be explained with reference to the flowchart shown in FIG. S1 is an imaging processing step to obtain a band-shaped image D including a contour line, S2 is a step of setting a line segment Fn-Gn that crosses the band-shaped image D, and S3 is a step of sampling along the line segment Fn-Gn. S4 is a maximum value determination step of selecting the coordinate Jmax of the point with the highest brightness, and S5 indicates the maximum value in the n-th line segment Fn-Gn. S6 is a representative position determination step in which the coordinates Jn of the representative position of the coordinate Jmax of the point are determined, and S6 is a connection processing step in which the contour line K is reproduced by connecting all the coordinates Jn of the representative position representing each line segment Fn-Gn. be.

[0011]The operation of the contour line extracting device of the present invention will be explained based on FIG. First, by the imaging means 9,
An image having a specific outline is captured to obtain a band-shaped image D including the outline, and a line segment setting means 31 sets a large number of line segments Fn-Gn that cross the band-shaped image D at each position,
The sampling means 32 samples the brightness level En of the image along this line segment Fn-Gn. Then, the representative position determination means 33 sets the peak position of the brightness level En on each line segment as the representative position Jn on this line segment, so that the parts other than the peak position on this line segment are slightly bright. However, all are excluded. Furthermore, if there is a part that is even slightly brighter than other parts on this line segment, that part is taken as the representative position, so that each representative position Jn can be obtained for each line segment. Therefore, by sequentially connecting each representative position Jn on each line segment by the contour line reproducing means 34, a thinned contour line K is reproduced. Note that as the imaging means, a still video camera (SV camera), which is a camera suitable for recording images in a data state that is easy to process, can be used.

0012

DESCRIPTION OF THE PREFERRED EMBODIMENTS Below, a contour extraction device used in the contour extraction method of the present invention will be explained in detail based on a tunnel internal cross-section measuring device as an embodiment thereof. FIG. 3 is a configuration diagram of the tunnel internal cross-section measurement device, FIG. 4 is an explanatory diagram showing the imaging state inside the tunnel, FIG. 5 is an image of the tunnel internal cross-section obtained by the SV camera, and FIG. 6 is the luminance distribution in the radial direction. The curve shown in FIG. 7 is a thinned correction contour line. In each figure,

001
3] 1 is a magnetic disk on which image data captured by the SV camera 9 is recorded;
, a background image A, a calibration image B, and a plurality of measurement images C1, C2, . . . , Cn, . . . are recorded. 2 is an SV regenerator for reproducing the image data of the magnetic disk 1; 3 is equipped with software and hardware for performing image processing on the basis of the image data output from the SV regenerator 2 to obtain a contour line. 4 is a personal computer that controls each peripheral device and has a built-in data processing program; 5 is a display device for displaying captured images and processed images; 5' is character data and graphic data; a CRT display device that displays
6 is a printer, 7 is a plotter, and 8 is a floppy disk device for storing programs and data.

The SV camera 9, as shown in FIG.
It is installed with its optical axis aligned in the direction of the central axis of the tunnel, and is controlled by a programmable controller 11 together with a laser beam irradiation device 10.

First, an image of the sky inside the tunnel in a state where no laser beam is irradiated is taken as a background image A. Next, an image containing arbitrary two points at a predetermined distance on a plane whose cross section is to be measured is captured as a calibration image B in the tunnel interior. Then, a plurality of measurement images C1, C are obtained by dividing the laser beam into the tunnel's interior at predetermined angles and rotating it, and photographing the tunnel's interior each time with an SV camera.
2,..., Cn,... are recorded on the magnetic disk 1. This magnetic disk 1 is set in the SV player 2 and image data is read out. Read measurement images C1, C2,...
, Cn, . . . are each calibrated in size based on the component image B, and are also subjected to subtraction processing from the background image A to obtain an image in which the influence of the background has been eliminated. Next, all of the subtracted images are added to form one continuous band image D as shown in FIG. 5. This band-shaped image D has a band-like shape, with bright areas being thick and dark areas being thin, depending on differences in reflectance, inclination, etc. of the surface of the tunnel inner wall.

Next, a process of obtaining a thinned contour line from band-shaped image data will be explained. In FIG. 5, the band-shaped image D includes the contour line of the cross section of the tunnel interior, but due to the condition of the tunnel surface, etc., there are bright and thick parts D1, slightly bright and thin parts D2, and other dim parts D3. There is. In the thick portion D1, the exact shape of the contour line is unclear, and in the slightly bright portion D2, the difference in brightness from the dim portion D3 due to the influence of background noise, etc. is small, and the contour line is not clear. These image data are stored at predetermined addresses in the image memory of the image processing board 3.

Next, a radial brightness distribution curve En as shown in FIG. 6 is obtained from the image in FIG. 5 by the following calculation process. In FIG. 5, P is the center point set at a position corresponding to approximately the center of the tunnel, and the line segment Fn - Gn is
In the circle centered on the center point P, the reference radial direction R
It is a radial line segment with an angle from 0 to θn. The outline D is cut across the line segment Fn - Gn to obtain a brightness distribution curve En showing the change in brightness at each position on the line segment Fn - Gn, as shown in FIG. That is, the brightness distribution curve En is obtained by sequentially reading out the brightness data of the addresses corresponding to the line segment Fn - Gn from the image memory. Next, in this brightness distribution curve En, the position of the portion Jmax of the brightness peak is set as cross-sectional position data Jn in the direction of the angle θn. That is, the line segment Fn read in the direction of the angle θn
- Select the maximum value among the brightness data of Gn, and position data J corresponding to the address where that brightness data is stored.
It outputs n.

[0018] The above processing is performed from an angle of 0 degrees to an angle of 20 degrees.
This is performed for each line segment Fn - Gn set in steps of 1 degree up to 0 degree, and cross-sectional position data J0 to J200 corresponding to each angle θ0 to angle θ200 is obtained. When the cross-sectional position data J0 to J200 obtained in this way is displayed on the CRT display device 5', a contour line K reproduced in a thinned state as shown in FIG. 7 is obtained. In the above, an example was shown in which the range from angle 0 degrees to angle 200 degrees was set in steps of 1 degree, but it is of course possible to set the range and steps in various ways.

When the step angle is increased, the number of cross-sectional position data Jn decreases, and the contour line K
becomes intermittent. In that case, it is preferable to interpolate the adjacent cross-sectional position data Jn, Jn+1 using a straight line or a curved line. Alternatively, a continuous contour line can be obtained by connecting the broken portions by conventional dilation processing/thinning processing. In this way, when it is desired to roughly measure the cross section of the tunnel interior, the step angle is increased to reduce the amount of data, thereby reducing the processing time. Moreover, when it is desired to improve measurement accuracy, it is preferable to reduce the step angle.

Next, the obtained contour line K and design cross-section line L are displayed in a superimposed manner on the CRT display device 5', and the numerical value of the difference between the contour line K and the design cross-section line L is displayed. It is possible to inspect the accuracy of construction work. These image data can be printed and drawn by the printer 6 and the plotter 7, and recorded and stored in the floppy disk device 8.

In this way, according to the above-mentioned tunnel interior cross-section measuring device, even if the image taken by the SV camera 9 has an unclear band-like outline, the luminance in the radial direction of that part can be determined. Since the peak portion is extracted, a clear, thinned contour line can be reproduced. Also, even if the outline is not bright enough to be blended into background noise, if that part is even slightly brighter than other parts in the radial direction, the brightest part will be played as the position of the outline. be able to.

[0022] In addition, in the conventional expansion processing/thinning process, finely curved parts are smoothed and become inaccurate, but according to the above-mentioned tunnel hollow cross section measurement device, the expansion processing/thinning process is performed. Since it is possible to reproduce a thinned contour line without any distortion, it is possible to obtain the effect of improving measurement accuracy. Furthermore, since the above image processing can be performed in a shorter time than conventional image processing using dilation processing and line thinning processing, it is possible to achieve the effect of shortening processing time and improving work efficiency.

[0023]

As described above, according to the contour line extraction method of the present invention, it is possible to reproduce an accurate contour line without smoothing out the small bends of the original contour line. This has the effect that lines can be extracted and reproduced in a short time. According to the contour extraction device of the present invention, contours can be reproduced in a short time even with an image processing device whose processing speed is not so fast, so a high-performance contour extraction device can be provided at low cost. There is also an economic effect.

[Brief explanation of drawings]

FIG. 1 is a schematic flowchart of a contour line extraction method according to the present invention.

FIG. 2 is a schematic configuration diagram of a contour extraction device according to the present invention.

FIG. 3 is a configuration diagram of a tunnel internal cross-section measuring device according to an embodiment of the present invention.

FIG. 4 is an explanatory configuration diagram showing the imaging state of the embodiment of the present invention.

FIG. 5 is an image of a cross-section of the tunnel interior taken by an SV camera.

FIG. 6 is a brightness distribution curve of the radial method of the image of FIG. 5;

FIG. 7 is a corrected contour line obtained by thinning the image in FIG. 5;

[Explanation of symbols]

S1 Imaging processing S2 Set line segment S3 Sampling processing S4 Maximum value determination S5 Representative position determination S6　　　Consolidation processing of representative positions D Strip image Fn-Gn line segment En brightness distribution curve Jn Cross-sectional position data K Regenerated contour line 9 Imaging means, SV camera 31 Line segment setting means 32 Sampling means 33 Representative position determining means 34 Contour line reproduction means

Claims (2)

[Claims] Claim 1: Obtain a band-shaped image including the contour by capturing an image with a specific contour, set a number of line segments that cross this band-shaped image at each position, and Sample the brightness level of the image, and on each line segment,
A contour line extraction method characterized in that a position where the brightness level shows a peak is taken as a representative position on this line segment, and the respective representative positions on each of the line segments are successively connected to reproduce the contour line. 2. Imaging means for capturing an image with a specific contour line to obtain a strip image including the contour line; and line segment setting means for setting a large number of line segments that cross the strip image at each position. sampling means for sampling the brightness level of the image along the line segment; representative position determining means for determining the peak position of the brightness level on each line segment as a representative position on the line segment; 1. A contour line extracting device comprising: contour line reproducing means for sequentially connecting representative positions to reproduce a contour line.