JPH07111735B2 - Crack measurement system - Google Patents

Description

The present invention relates to a crack measuring system for analyzing and measuring a crack generated in an object by image processing, and more specifically, for example, a building or a bridge. , Cracks in the range of 0.1 mm to 3.0 mm generated in concrete and metal that compose objects such as automobiles and airplanes as well as structures such as roads, storage dams and waterways are analyzed by image processing However, the present invention relates to a crack measurement system that measures the width, position, shape, distribution state, etc.

(Conventional technology) Buildings, bridges, roads, reservoir dams, waterways and cars,
Cracks occur in reinforced concrete, metal, and the like that make up structures and objects such as airplanes due to aging and various other factors. If you leave such cracks,
Durability and earthquake resistance of structures and objects are reduced.

Therefore, it is necessary to accurately measure the width of cracks of this kind, but generally, in order to perform accurate image processing and measurement of cracks with a width of 0.1 to 3.0 mm, the density of the pixels that make up the image is 0.1. At least 4 pixels per mm (that is, a resolution of 0.025 to 0.05 mm per pixel) is required. When a crack image is constructed with such a pixel density, in addition to the general feature that the crack is "a linear structure" (with a width of 0.2 mm or less), a feature such as "a long and large band-like area" (0.2 mm Width or more).

In addition, image processing of cracks to measure is to process images in which cracks are photographed.
This image shows the shadow created by the cracks. The extraction of cracks in image processing is to extract a portion where the shadow generated in the image is blacker than the other surface of the object, but the surface of the object or the like is caused to have light and darkness other than the crack. Since there are things such as adhesion of dirt, discoloration of the surface, and shadows due to unevenness of the surface, it is not possible to simply binarize and extract the cracks depending on the brightness of the surface of the object.

In order to detect cracks from an image of the surface of an object, it is necessary to examine the entire image of the entire surface of the object.To do this, the entire image is targeted for processing, and an exhaustive search is performed within it. There is a need for an image processing algorithm to detect crack patterns.

(Problems to be Solved by the Invention) As described above, cracks are classified into two parts, a linear structure part and a strip-shaped long and large part. An image processing algorithm for identifying such a crack part is
It is necessary to use two algorithms depending on the crack width, "algorithm to analyze linear structure" or "algorithm to analyze long and large area as crack". There is a problem that it cannot be uniquely determined regardless of the width.

In addition to the cracks, the surface of the object has lightness and darkness due to adhesion of dirt, etc., and there is a problem that the cracks cannot be simply binarized and extracted.

Furthermore, to detect cracks from the image of the surface of the object,
Image processing needs to be performed on the entire surface image,
If you set the image resolution to the pixel density as described above,
For example, when imaging the surface of an object having a size of 50 cm × 50 cm, the image data becomes enormous so that a memory capacity of 400 MB is required, which causes a problem that an image processing load increases and efficiency decreases.

The present invention has been made in view of the above, and an object thereof is, for example, from a very thin crack having a width of about 0.1 mm to a thick crack having a width of about 2 to 3 mm without being affected by surface dirt or the like. The object is to provide a crack measurement system that accurately and efficiently recognizes and measures.

[Structure of Invention]

(Means for Solving the Problems) In order to achieve the above object, the crack measurement system of the present invention is a crack measurement system that performs image processing to measure the state of a crack generated on the surface of an object, and the surface of the object. A hierarchical compression unit that compresses the captured image to a predetermined compression ratio stepwise and performs line-preserving smoothing filtering on the compressed image at each stage, and an image compressed to a predetermined compression ratio by the hierarchical compression unit. A rough crack extracting means for performing crack extraction candidate portion extraction processing using line extraction circular filtering and binarization, skeletonization, vectorization, and noise separation processing to extract a crack portion and calculate a position; A rough crack width calculating means for calculating a crack width by the linear operator method from the crack portion extracted by the rough crack extracting means, and The crack position and width calculated by the rough crack extracting means and the rough crack width calculating means are used as a guide map for calculating the crack position and width in an image having a compression ratio one step smaller, and thereafter compressed by the hierarchical compressing means. Using the crack position and width calculated for a compressed image having a large compression ratio in the reverse order as a guide map, binarization is performed for a crack portion of an image having a compression ratio one step smaller than the compressed image,
It is a gist to have a detailed crack extraction measuring means for performing noise removal, skeletonization, and measurement of crack width in stages until an image of a size before compression by the hierarchical compression means is reached.

(Operation) In the crack measuring system of the present invention, the picked-up image of the object surface is compressed step by step, and the compressed image is subjected to line extraction circular filtering and binarization to extract a crack candidate portion, core conversion, The crack width is roughly calculated by the linear operator method from the crack portion extracted by performing the vectorization and the noise separation processing, and the rough crack position and width are calculated in the image whose compression ratio is one step smaller. As a guide map, and thereafter, the crack position and width calculated for a compressed image having a large compression ratio, which is the reverse of the order of compression in the stepwise compression processing, is used as a guide map, and a cracked portion of an image having a compression ratio one step smaller than this compressed image On the other hand, binarization, noise removal, skeletonization, and measurement of crack width are stepwise converted into an image of the size before the stepwise compression processing. Repeat until it reaches.

(Examples) Examples of the present invention will be described below with reference to the drawings.

FIG. 1 is an overall configuration diagram of a crack measuring system according to an embodiment of the present invention. As shown in the figure, the crack measurement system of the present invention, for example, a crack formed on the surface of a structure made of concrete or the like, for example, a crack having a width of about 0.1 mm to a width of about 3 mm can be accurately processed by image processing. The measurement is performed by reading an image of the photographic film 1 obtained by photographing the surface of the structure with a camera with the photoelectric image reading device 3, and analyzing the read image with the image processing device 5 to display on the surface of the structure. Measuring the width, length, position, etc. of the formed cracks, and finally outputting the characteristic data including these various data, distribution, characteristic diagram and crack diagram from the output device 7 consisting of a personal computer etc. Is. The image reading device 3 has, for example, a CCD camera and a microscope for reading an image on the photographic film 1. In this embodiment, the image reading device 3 reads the image of the photographic film 1 obtained by photographing the surface of the structure with the camera, and the image processing device 5 analyzes the image. Instead of reading the image on the film 1, the surface of the structure may be directly imaged by the image reading device 3 and the imaged image may be directly used.

When reading the photographic film 1 with the image reading device 3, the crack width of 0.1 mm, which is the detection target, is divided into 4 parts to obtain 0.1 mm.
While using a film reading device capable of reading cracks at a reading density of / 4 pixels, shading correction and sensor noise correction are simultaneously performed on the entire frame of the film 1, and the read image information is processed by an image processing device 5 including a computer. Are sequentially sent to, converted into numerical image data, and subjected to image processing.

The image data read from the film 1 by the image reading device 3 and supplied to the image processing device 5 is processed by the image processing device 5.
In the above, the extraction and measurement of cracks are performed by sequentially receiving the rough crack extraction and measurement processing (A) and the detailed crack extraction and measurement processing (B) described below.

(A) Schematic crack extraction and measurement processing (1) Image compression processing and hierarchical structure (2) Crack extraction in 1/32 compressed image (3) Schematic crack position and width measurement (B) Detailed crack extraction and Measurement Processing FIG. 2 is a diagram showing a schematic flow of crack extraction measurement image processing in the crack measurement system shown in FIG. 1 together with an explanatory diagram. Next, description will be made with reference to FIG.

As described above, when the image on the photographic film 1 is read by the image reading device 3 (step 11 in FIG. 2).
0), the read original image data 9 (second
Figure) undergoes compression and hierarchical structuring process (step 12
0), 1/4 → 1/8 → 1/16 → 1/32, and compression is performed stepwise (steps 121, 125, 126, 127). That is, the original image data 9 is first reduced to 1/4 and then sequentially compressed to 1/2 ⁿ to obtain 1/2 ⁿ compressed image data 10 to 1/32 compressed image data 11
Becomes

Note that this compression processing is performed using "line-preserving smoothing filtering" that leaves the line structure and "reduction thinning by the maximum value selection method (because the image data value is larger in the cracked portion than in concrete or other parts)". Be seen. By performing such a gradual compression, the cracks with a wide width, which are considered to be strip-shaped regions, are gradually converted into a line structure with a narrow width as the layers become higher in hierarchy, that is, as they are compressed. In the upper image, the characteristics of the line structure are preserved as they are. By compressing the image in this way, the crack has only linear characteristics regardless of its width, and by applying the line extraction algorithm, it is possible to analyze the macroscopic line structure of the crack. become.

Next, using a line structure extraction algorithm as described later on such a 1/32 compressed image, cracks are extracted by paying attention to the line structure that is the feature, and the extracted cracks are extracted. The image is subjected to processing such as skeletonization, vectorization, and noise separation, and a rough crack pattern is extracted (step 130). In the binarization of cracks performed in the line structure extraction algorithm, the compressed (reduced) image is divided into smaller blocks in order to prevent adverse effects of local stains and stains on the surface of the object as much as possible. It is divided, and a binarization threshold value is calculated for each divided block to perform binarization.

Then, for the crack pattern thus extracted, the crack width is roughly calculated by the direct operator method (step 140).

Next, based on the approximate crack pattern and width extracted from the 1/32 compressed image as described above, the cracked portion is searched in the lower layer image step by step in the reverse order of compression and reverse,
The detailed crack extraction and measurement processing is repeatedly performed until the original image at the time of reading the film, and the measurement accuracy of the crack center position and the crack width is improved (step 150).

That is, in this detailed crack extraction and measurement process, the crack data measured in the upper layer image data is used as a guide map to search for cracks in the lower layer image data, and the process is compressed to 1/32 in Fig. 2 as a whole. image
The crack measurement is performed step by step to the lower layer image from 13 to 1/16 → 1/8 → 1/4 → the original image 15 to make the crack measurement finer (steps 151 to 154).

More specifically, since it is not possible to measure the exact position and width of a crack with a macroscopic crack pattern extracted from a 1/32 compressed image, refer to the approximate crack width and By searching for cracks and repeating them until the original image before reading and compressing the film, the accuracy of measuring the crack position and crack width is gradually improved.

The search range is a strip-shaped range around the crack pattern extracted in the upper image. That is, the crack width (calculated at the same time as the crack pattern) on both sides of the crack pattern (line) required in the upper image.
2 to 3 times as many strip-shaped search areas are provided, and the local search range is sequentially binarized to distinguish the cracked portions. After that, the center position of the crack is determined in the distinguished crack portion, and the crack width (thickness of the binarized portion) is also calculated at the same time, and the center position of the crack is advanced as the image processing progresses to the microscopic image. Also, the accuracy of measuring the crack width is improved.

The crack data consisting of the crack position and width retrieved as described above is supplied from the image processing device 5 to the output device 7 and stored in an external storage device such as a magnetic disk thereof, and the crack data is analyzed. A crack diagram and a crack characteristic diagram are created.

FIG. 3 is a flowchart showing a more detailed measurement process of the crack measuring system described above. The detailed crack measurement processing shown in the figure comprises the above-described rough crack extraction and measurement processing and detailed crack extraction and measurement processing following the above-described crack photographic film reading processing, and more detailed processing shown below. In the flow of FIG. 3, steps that are the same as the steps of the flow of FIG. 2 are given the same numbers.

(A) Extraction of rough cracks and measurement processing (1) Image compression processing and hierarchical structure (a) Creation of 1/4 reduced image (b) Line-preserving smoothing filtering (c) 1 / from 1/4 reduced image Stepwise compression to 8,1 / 16,1 / 32 (2) Extraction of cracks in 1/32 reduced image (a) Extraction of line structure Calculation of feature quantity by circular filtering and binarization (c) Core line (d) Vectorization of cracks (e) Noise separation (3) Schematic crack position and width measurement (a) Manual designation of crack measurement area (b) Calculation of conversion coefficient from photo coordinates to actual structure coordinates (c) Schematic crack Confirmation of extraction result (d) Calculation of rough crack width (e) Registration of roughly extracted crack vector and width (B) Detailed crack extraction and measurement processing (1) Selection of subdivided image in which crack exists (2) Crack position Search and binarization (3) Noise Removed by (4) core of (5) Measurement of Crack Width (6) Crack measurement result scaled below with reference to FIG. 4 and later in accordance with the flowchart shown in FIG. 3, it will be sequentially described detailed processing.

First, read the cracked photographic film (step
110) is the entire cracked film (effective screen 50 mm x 5
0 mm) is read by a film reader. For reading, the film is subdivided into small rectangular areas of 1.28 mm x 1.2 mm (corresponding to 512 x 480 pixels for one reading area of the CCD area sensor, which is the sensor part of the film reader), and the film 1 Read all the com. Therefore, the cracked image is composed of about 2000 subdivided images, as shown in FIG. 4, including the effective screen of the photographic film and the reading of the periphery thereof.

Assuming that the image is read at a 1/10 scale, the crack of 0.1mm width is divided into 4 pixels (1 pixel is 2.5μ on the film and 0.0 on the structure).
The reading density is adjusted to 25mm) and the read density is adjusted to 0.1mm / 4 pixels.

Generally, when an image is formed through a lens, the image plane illuminance differs between the center of the lens and its periphery, and the image around the lens becomes darker than the center (this phenomenon is called dimming around the lens). In addition, the light receiving element 1 that constitutes the CCD area sensor
Each one (512 x 480 elements in total) has slightly different sensitivity. For this reason, the read image data does not have a uniform numerical value even if a film having a uniform density is read because of the peripheral dimming of the lens and the non-uniformity peculiar to the CCD area sensor element.

In film reading in cracked image processing, such image non-uniformity is corrected (shading correction and sensor noise correction) every time a subdivided image is read (step 111 and step 112). As shown in FIG. 5, the correction method is to take in data (think of it as the brightest uniform image) from the film reader without setting the film in advance, divide the cracked subdivided image sequentially read by this data, The method of normalizing the read image data is adopted.

That is, the shading and sensor noise correction shown in FIG. 5 is performed by setting the subdivided image data O (i, j) read by setting the film 1 as shown in FIG.
As shown in FIG. 3B, the image data S (i, j) read without setting the film 1 is divided, and this quotient is multiplied by a constant to perform shading correction and sensor noise correction image data P (i, j). Is getting Note that i = 1 to 512 and j = 1 to 480.

After reading the image data as described above, in the rough crack extraction and measurement processing, the rough crack pattern extraction and measurement processing is performed from the image read as the subdivided image of the entire surface of the cracked photography film. Compression and hierarchical structure (Step 12
In (0), a 1/4 reduced image is created (step 121).

In the creation of this 1/4 reduced image, the minimum detected crack width of 0.1 mm becomes the size of one pixel of the image data so that the line-preserving smoothing filtering (step 123) is performed efficiently. Perform reduction.

Specifically, since the original image has 0.025 mm / 1 pixel, it is reduced by 1/4. In this reduction method, an average value is calculated from 16 pixels of 4 pixels × 4 pixels, and this value is set as a new pixel.

Then, as shown in FIG. 6 (a), the 1/4 reduced image is divided into the 1st to 8th points for each point (i, j) of the reduced image.
Calculate the average value AVen (n = 1 to 8) in the direction and use this average value
The maximum value is retrieved from AVEn and line-preserving smoothing filtering is performed using the value of the point (i, j) (step 123). The cracked part is because the image data value is higher than the surrounding area. This is done so that isolated shadows other than cracks do not mix as noise in the subsequent compression process.The values of linear shapes such as cracks do not change, but the isolated points are smooth. It is adapted to reduce the value. The length of the filter was set to a range of about 1 mm on the structure considering the linearity of cracks.

Next, the 1/4 reduced image obtained as described above is compressed stepwise so as to be a power of 2 as 1/8, 1/16, 1/32, as shown in FIG. 6 (c) ( Steps 125-127). Image compression is performed by thinning out by the maximum value selection method that selects the maximum value from the 2 × 2 4 pixels as a representative value, as shown in FIG. The thick cracks are gradually converted into a thin line structure.

For this reason, for example, it is conceivable to adjust the minimum film reading density to 0.025 mm / 1 pixel (corresponding scale on the surface of the measured object such as concrete) for the minimum detected crack width of 0.1. Therefore, when performing line-preserving smoothing filtering, the image is reduced to 0.1 mm / 1 pixel (by simple calculation of average value of 4 pixels x 4 pixels and thinning).
However, processing is performed on this image. as a result,
A 1/32 compressed image with one pixel of 0.8 mm is generated from an original image with one pixel size of 0.025 mm (on the measured object) (see FIGS. 6 (a) and 6 (b)).

A crack pattern is then extracted from the highest layer image compressed to 1/32 as described above (step 130). In this process, as shown in FIG. 7, the line structure extraction circular filtering is used to calculate the characteristic amount S of the line structure (crack) in each pixel of the image, and the binarization is performed based on the average feature amount of the image. It is determined (binarization) whether the pixel is a crack candidate portion according to the threshold value (step 132).

When the line structure extraction circular filter is used for the image, a large feature amount S is calculated where the pixel is on the line structure. As the parameters of the circular filter necessary for this calculation, the values (target window size WS = 1, processing radius R = 5) obtained as a result of the original research are used.

In calculating the binarization threshold, the entire image is shown in FIG.
As shown in Fig. 3, it is divided into 36 small blocks, and a binarization threshold value for determining whether each of these blocks has a line structure (a value obtained by averaging the feature amount S calculated for each pixel in the target block is a binary value). The threshold value is set as the threshold value, and the line element candidate points above the average value and the background below the average value) are individually calculated.

In the binarization, a value having a feature amount larger than the binarization criterion in each block is set as “1” as a line element candidate point, and the other portion is set as “0” as a background, whereby the crack candidate portion is determined. Extract and do. By calculating the binarization standard by dividing it into small blocks, it is possible to suppress erroneous determination caused by a change in the binarization standard due to local stains on the image.

Next, the crack candidate portion obtained by the line structure extraction circular filtering is subjected to the core line by the Hilditch thinning method, that is, the processing for obtaining the center line of the crack candidate portion is performed (step 134).

After the skeletonization, the position coordinates on the image of the point sequence forming the center line are arranged in the order of connection, and the position coordinate value sequence of the crack portion, that is, the rough crack vector (see FIG. 18) is obtained (step 136). ).

In order to remove portions other than cracks from the crack candidate point sequence thus obtained, isolated points, isolated areas are removed (small area portions are removed), continuity check (short crack candidate point sequences are removed), and whisker-like noise. Removal (removal of noise generated at the time of skeletonization (see FIG. 19)) and loop area filling processing are performed (step 138).

Note that the values shown in FIG. 20 are set to the parameters used in the program such as noise removal processing.

Next, for the rough crack pattern extracted in the 1/32 compressed image as described above, the rough crack width is further calculated by the straight line operator method (step 140). The operator designates (arbitrary polygonal area) (step 141).

Then, the conversion coefficient from the photographic coordinates to the coordinates of the actual structure, that is, the two-dimensional projective conversion coefficient (at the step of detailed crack extraction and measurement in the next step 150, the crack measurement data is converted from the coordinate system on the image into the coordinates on the actual structure. An eight-element coefficient for scale conversion to the system is calculated (step 143). Note that at this stage, the coefficient to be used in the next step 150 is calculated as follows: "If the conversion coefficient is also determined when the entire screen for checking the extraction cracks is displayed, it is necessary to display it again in the subsequent processing. It disappears and the subsequent processing can be automatically advanced without the intervention of the operator. ”
The projective transformation coefficient is calculated by solving the simultaneous equations by associating the coordinates of the four reference points (points whose coordinate values are known on the object surface) projected on the cracked photograph with the coordinates on the image data. It is possible (see FIG. 8).

Next, the outline crack extraction result is confirmed, and the additional correction of the extraction leakage crack and the deletion of the misidentified crack are manually performed (step 145). Then, the crack width is calculated by the linear operator method (see FIG. 9) for each constituent point of the rough crack vector finally checked and confirmed by the operator in this way (step 146), and the roughly extracted crack position and its position are calculated. Register the width (step 148). The approximate crack width and the crack position are referred to as initial values in the detailed crack extraction and measurement in the next step.

Next, detailed crack extraction and measurement processing (step 150) are performed. This is because the rough crack vector (crack center line position and width) obtained from the image at the top of the layer compressed to 1/32 in the rough crack extraction and rough crack width measurement processing described above is used as a guide map for the one step lower layer. The crack in the compressed image, that is, the 1/16 compressed image is binarized to determine the center position and measure the width (see Fig. 10). The crack search range is a band-shaped range around the crack pattern extracted in the upper image. That is, a strip-shaped search area having a width two to three times as large as the crack width (calculated at the same time as the crack pattern) is provided on both sides of the crack pattern (line) obtained in the upper image, and the local search range is sequentially divided into two. Thresholding (local iterative discrimination binarization method that is an improvement of Otsu's binarization method) is performed to distinguish cracked portions. Then, the center line position of the crack is determined in the distinguished crack portion, and the crack width (thickness of the binarized portion) is calculated.

Then, this process is performed from 1/32 compressed image to 1/16, 1/8, 1/4 compressed image and further to the original image. Detailed crack extraction and measurement are performed by repeating the procedure. In this case, the crack position and width measured at each stage are used as a guide map in the image planning of the next layer, which allows the crack position and width to be accurately measured.

Next, the smaller the size of the image to be processed, the shorter the image processing will be completed. Therefore, in order to perform detailed crack measurement, it is important to reduce the processing unit as much as possible. Therefore, the image of each layer is divided into several small blocks and managed as shown in FIG.
When processing an image one step below based on the rough crack vector obtained from the upper layer image, as shown in FIG. 11, a small block to be processed, that is, a subdivided image in which a crack exists is selected. (Step 161). As a result, image processing of unnecessary portions is omitted, and the entire processing can be shortened.

Further, among the selected small blocks, image processing is performed only on a local area where cracks exist, thereby achieving more efficient processing.

Next, the crack position is searched and binarized by the local and sequential iterative binarization method (step 163). This is a small block selected based on the rough crack vector measured in the upper image, and only the small local area near the crack is selected and binarized to obtain a crack candidate area (see FIG. 11). . The small area near the crack is actually
As shown in Fig. 12, it is an elliptical small area along the crack.

When binarizing cracks, one frame of image data (50
cm × 50 cm to 100 cm × 100 cm) In addition to the problem that one binarization threshold cannot be determined for the whole, there is a unique problem of Otsu's discriminant binarization method, that is, the binarization target area (cracking). ) And the area ratio of the background (surface such as concrete) are extremely biased, it has been pointed out that an appropriate binarization threshold cannot be calculated. Therefore, in this crack measurement system, each of the rough crack vectors is The local binarization small area is selected by narrowing down the binarization area to a local elliptical small area centered on the constituent point, and selecting the area ratio of the binarization portion and the background. The size of the small area is determined with reference to the crack width calculated from the image with the compression ratio one step higher so that the deviation does not occur and is appropriate (see FIG. 10).

In addition, in order to reduce the influence of noise when binarizing each small area, especially for binarizing narrow cracks, "binarization → expansion → noise removal → next binarization area determination The above cycle is repeated to adopt a repetitive binarization process that determines cracks while removing noise (No. 13).
See figure). It should be noted that the method of determining the end of iteration is to compare the previous binarized area and the binarized area of that time by 90%
Suppose that the above changes are no longer observed.

Of the cycles described above, the number of expansion treatments also differs depending on the crack width. That is, it is necessary to increase the number of times of expansion for a thick crack but increase the number of times of expansion, and to reduce the number of times for a thin crack in order to prevent uneven distribution of the area ratio between the cracked portion and the background (see FIG. 13). It was also confirmed that if the width becomes thicker than a certain amount, a good binary image can be obtained without executing the above-described repeated binarization processing.

As described above, by limiting the number of repetitions and the number of expansions using the crack width measured in the upper image,
2 cracks accurately with a measuring method that matches the width of the cracks
It can be quantified. Fig. 21 shows the correspondence between each parameter used for binarization and the crack width measured in the upper image.

Next, with respect to the binarized crack candidate region, removal of isolated noise (step 165) and skeletonization (step 16).
Do 7).

First, among the perforated portions formed inside the crack candidate region, which have a small area, the holes are filled as shown in FIG. This is to prevent the crack width from being measured small when measuring the crack width.

Then, by moving the approximate crack vector to the center of the crack candidate region, skeletonization processing is performed to gradually bring the center line position of the crack closer to the true center position as shown in FIG. At the time of skeletonization, an isolated area (noise) that remains even after binarization is removed at the same time. This is because when there are multiple candidates for the destination of a rough crack vector due to isolated area noise, the connectivity of the cracks is focused on, the destination of the vector is determined based on the principle of majority, and the crack moves to the isolated area. This is a method not to do so (see Fig. 16).

This skeletonization and isolated noise removal will be described in more detail. First, for a crack candidate region binarized based on the rough crack vector extracted in the image at the higher stage,
Number each area. Next, at each constituent point of the rough crack vector, a point where a straight line perpendicular to the crack direction and the boundary portion of the crack candidate region intersect is obtained. The midpoint of this intersection is set as the new rough crack vector (
(See Figure 15). As a result, the centerline position of the crack is obtained in the image one step lower, and a rough crack vector with improved one step accuracy can be obtained.

When there are multiple candidates for the destination of the new rough crack vector due to isolated noise, etc., pay attention to the connectivity of the cracks, aggregate the destination numbers of the points that make up the series of new rough crack vectors, and use the majority vote to determine the most frequent The region having the higher destination number of is determined as the destination of the new rough crack vector (see FIG. 16). Then, for the constituent points of the new rough crack vector that has not moved to the area determined by the majority vote, the area having the number is searched again and the center is moved so that the center line does not move to the isolated area ( (See Figure 16).

Next, the crack width is measured at the center position of the crack (step 169).

As shown in FIG. 17, the width of the crack is measured by providing a circle inscribed in the crack centering on the center position of the crack made into a core and calculating the diameter of the inscribed circle. At this time, as shown in Fig. 14, the center position of the inscribed circle is swung in a direction perpendicular to the crack direction so as not to underestimate the crack width depending on the shape of the crack candidate portion, and the circle is gradually expanded to have the maximum diameter. Search for an inscribed circle such that and make the diameter of the circle the crack width.

Next, the crack measurement data is scale-converted (step 171).

When the measurement of the original image at the bottom of the hierarchy is completed, the crack measurement values (position and width are expressed in the coordinate system on the image up to this stage) are used to determine the coordinate system on the surface of the object such as concrete ( Scale conversion is performed in order to express in actual size. Scale conversion is performed by the two-dimensional projective conversion shown in FIG.
The projective transformation coefficient obtained at the final stage of extraction of rough cracks is used here to transform into actual dimensions.

As described above, when the detailed crack extraction and the measurement process are completed, the measured crack data is stored in the external storage device (magnetic disk) of the output device 7 from the image processing device 5 and the crack data is analyzed. A crack diagram and a crack characteristic diagram are created. Figure 22 shows the output format of crack data. In this table, the crack position data is to be represented by three-dimensional coordinate values.

〔The invention's effect〕

As described above, according to the present invention, a picked-up image of an object surface is compressed stepwise, and a crack candidate portion is extracted by using line extraction circular filtering and binarization for this compressed image, and the core line is formed. , The crack position and width are roughly calculated by the linear operator method from the crack portion extracted by performing the vectorization and noise separation processing, and the rough crack position and width are calculated in the image with the compression ratio one step smaller. Is used as a guide map, and thereafter, the crack position and width calculated for a compressed image with a large compression ratio, which is the reverse of the order of compression by the stepwise compression, is used as a guide map, and the compression ratio is one step higher than the compressed image used as this guide map. Binary compression, noise removal, skeletonization, and crack width measurement for cracks in small images Since it is repeated until the image of the size is reached, the crack has only the linear characteristic regardless of the width by the image compression, and the line extraction image processing algorithm can be applied. It is possible to reduce the risk of erroneously recognizing typical linear noise as a crack, reduce the effect of local noise such as dirt and stains on the object surface, and extract the extracted rough crack pattern stepwise. The center line position of the crack while expanding on the image below the hierarchy,
Since the width is measured, the positional accuracy and the crack width measurement accuracy can be improved while maintaining the connectivity of the cracks, and the crack pattern extracted in the higher hierarchy is used as a guide map to analyze the images in the lower hierarchy sequentially. A micro image without cracks can be removed from the object to be processed, the efficiency of the processing can be improved, and the measurement values can be gradually refined by stepwise measurement, and accurate measurement can be performed.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a crack measuring system according to an embodiment of the present invention, FIG. 2 is a diagram showing a schematic flow of crack extraction measurement image processing in the crack measuring system shown in FIG. FIG. 3 is a flow chart showing a more detailed measurement process of the crack measuring system of FIG. 1, FIG. 4 is an explanatory view showing reading of a crack photograph by an image reading device, and FIG. 5 is a shading correction in reading a crack photographic film. Explanatory drawing showing the method of sensor noise correction, FIG. 6 (a), (b),
(C) is an explanatory diagram of line-preserving smoothing filtering, an explanatory diagram of thinning by the maximum value selection method, and an explanatory diagram of image layering and compression method. FIG. 7 shows binarization reference calculation in line extraction circular filtering. Illustration of block division,
FIG. 8 is an explanatory diagram of coordinate conversion from photographic image coordinates to coordinates on an actual structure, FIG. 9 is an explanatory diagram of the linear operator method, and FIG.
Figure is an explanatory diagram of detailed crack extraction, Fig. 11 is an explanatory diagram showing the relationship between the crack data extracted at the upper part of the hierarchy and the lower layer image data selected based on this data, and Fig. 12 is the local using general crack vector constituent points. Repeated sequential discrimination 2
Fig. 13 is an explanatory diagram of the quantification method, Fig. 13 is an explanatory diagram of repetitive binarization and expansion of the cracked portion, and Fig. 14 is an explanation showing movement of the measurement position, expansion of the inscribed circle, and removal of the influence of noise by filling holes FIG. 15, FIG. 15 is an explanatory diagram of detailed crack core line formation by moving the outline crack vector constituent points, and FIG. 16 is an explanatory diagram of crack core line formation that also uses isolated point ignoring based on a majority decision,
FIG. 17 is an explanatory diagram showing the definition of crack width, FIG. 18 is an explanatory diagram of thinning and vectorization, FIG. 19 is an explanatory diagram showing occurrence of whisker-like noise due to thinning, and FIG. 20 is a crack image. Figure showing image processing method and main parameters adopted for processing,
Fig. 12 shows the binarization parameters in the local and sequential iterative binarization method, Fig. 22 shows the format of the crack data, and Fig. 23 shows the unit processed image at each stage of the stepwise crack measurement method. It is a figure which shows a size. 1 ... Film, 3 ... Image reading device, 5 ... Image processing device, 7 ... Output device.

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical indication location G01N 21/84 Z (72) Inventor ▲ Ken Ihara 13-16 Tamura-cho, Atsugi-shi, Kanagawa No. Asia Consulting Co., Ltd. (72) Inventor Kazuo Oda 13-16 Tamura-cho, Atsugi-shi, Kanagawa Within Asia Consulting Co., Ltd. (56) References JSCE 45th Annual Meeting 45 , P. 374-375

Claims (5)

[Claims] 1. A crack measuring system for image-processing and measuring a state of a crack generated on the surface of an object, wherein the surface of the object is compressed stepwise to a predetermined compression ratio and compressed. Hierarchical compression means for performing line-preserving smoothing filtering on the image at the stage, and for cracked candidate parts using line extraction circular filtering and binarization for the image compressed to a predetermined compression ratio by the hierarchical compression means. Extraction processing, skeletonization, vectorization, and noise separation processing are performed to extract a crack portion and calculate the position, and a crack width is calculated from the crack portion extracted by the rough crack extraction means by a linear operator method. Means for calculating the approximate crack width, and cracks calculated by the approximate crack extracting means and the approximate crack width calculating means The position and the width are used as a guide map for calculating the crack position and the width in the image having the compression ratio smaller by one step, and thereafter, the crack calculated in the compressed image having the large compression ratio in the reverse order of the compression by the hierarchical compression means. Using the position and width as a guide map, binarization, noise removal, skeletonization, and crack width measurement are performed stepwise for the crack portion of the image whose compression ratio is one step smaller than that of the compressed image before compression by the hierarchical compression means. A crack measurement system, which has a detailed crack extraction and measurement means for performing the process up to the size image. 2. The hierarchical compressing means compresses an image of the surface of an object in stages up to a predetermined compression ratio, and points constituting the image at each stage compressed by the compressing means. Line preserving smoothing filtering to convert thick cracks into thin line structures, thin cracks are preserved as thin line structures, and line preserving smoothing filtering means to reduce noise other than cracks is featured. The crack measuring system according to claim 1. 3. The detailed crack extracting and measuring means 2
The binarization is based on each component point of a new rough crack vector obtained by noise removal and skeletonization based on the rough crack vector obtained by the rough crack extraction unit vectorizing in an image having a compression ratio one step higher. Local binarization small region selection means for narrowing the binarized region into a local elliptical small region to be binarized to obtain a crack candidate region, and an uneven area ratio between the binarized part and the background The small area determining means determines the small area based on the size of the crack width calculated by the approximate crack width calculating means, and the width is narrow so as to reduce the influence of noise when binarizing the small area. Repetitive binarizing means for repeating the process of determining a small area for binarizing, expanding, removing noise and next binarizing for cracks, and repeating times in the repetitive binarizing means. And cracking measuring system according to claim (1) wherein the and a number limiting means for limiting the inflation times. 4. The noise removal and skeletonization in the detailed crack extraction and measurement means are performed on a crack candidate region obtained by binarization in an image having a compression ratio one step higher,
Crack candidate area numbering means for numbering each area, and a straight line and a crack candidate perpendicular to the crack direction at each point of the rough crack vector obtained by noise removal and skeletonization for an image with a compression ratio one step higher. A center position searching means for obtaining the midpoint of the point where the boundary parts of the regions intersect as a new rough crack vector, a majority decision judging means for judging a crack candidate area to which this new rough crack vector belongs most, and a majority judging means The crack measuring system according to claim 1, further comprising: a re-moving unit that moves a new rough crack vector that does not belong to the determined crack candidate region to the center position of the determined crack candidate region. 5. The crack width measured by the detailed crack extraction and measurement means is inscribed circle setting means for setting a circle inscribed in a crack candidate area around a new approximate crack vector obtained by skeletonization, and cracks. A maximum inscribed circle searching means for searching for an inscribed circle having a maximum diameter by gradually expanding the circle while swinging the center position of the inscribed circle in a direction perpendicular to the crack direction so as to eliminate the influence of the shape of the candidate region. The crack measuring system according to claim 1, further comprising: a filling means that fills an island-like void portion left inside the crack before the inscribed circle is obtained.