JP6788161B2 - In-object deformation part detection device - Google Patents

Description

In general, the present invention describes spatial characteristics such as the position, shape, and range of deformed parts existing on the surface or inside of an object in the surface temperature distribution of the object or an image drawn based on the same (collectively, the surface). It relates to a device for detecting based on a temperature distribution or an image).

If there is a deformed part of the material properties on the surface or inside of the object that is different from the whole part with relatively uniform material properties, when heated or cooled from the outside or inside, the whole object reflects the deformed part. After undergoing an unsteady temperature distribution, it finally exhibits a stable and steady temperature distribution. The surface temperature distribution of an object during heating or cooling includes a temperature variation portion which is a local temperature distribution reflecting a deformed portion. In such a case, if the surface temperature distribution of the object is measured at a certain point in time, changes existing on the surface or inside of the object in a non-destructive or non-contact manner depending on the presence or absence of a temperature variation portion in the surface temperature distribution. The presence or absence of a shaped portion can be roughly detected.

As an application of the above contents to an actual machine, for example, in the field of civil engineering and construction technology, it exists in deformed parts such as cracks, cavities, floats, and peelings in concrete structures, and at the boundary between steel and concrete in steel-concrete composite structures. A technique for qualitatively detecting cavities, etc., by locally finding locations with large fluctuations in color or shading from colored shading images that image the surface temperature distribution measured with an infrared camera, and qualitatively detecting them as detection target locations. Is commonly used. The method using an infrared camera can be performed non-contact from a place away from the target structure by a handy small device, and also uses a surface temperature distribution that reflects not only the surface of the structure but also the deformed part existing inside. Therefore, it is possible to detect the existence of an internal deformed portion that cannot be seen from the outside. Therefore, compared to the method of using a large and very expensive device that generates electromagnetic waves such as X-rays in contact state, and the method of not contacting by visual inspection or CCD camera but only targeting the phenomenon of the structure surface, the infrared camera Although the method according to is incomplete, it is considered that the method can improve the weaknesses while retaining the advantages of these methods.

On the other hand, the recent development of image treatment technology that handles various images themselves analytically has been remarkable, and by applying these technologies, the spatial characteristics of the deformed part can be further analyzed from the surface temperature distribution of the object. It is considered that the possibility of detection targetly and quantitatively will be improved. For example, a waveform feature extraction technique based on a wavelet transform for detecting a singular point locally contained in a relatively large image as a localized waveform feature peculiar to it, as shown in Non-Patent Document 1. Exists. Considering the structure in the field of civil engineering and construction technology, in the initial stage of the deterioration phenomenon of a general structure, the deformed part begins to appear locally in a part of the structure. Therefore, the surface temperature associated with the initial deformed part appears as a peculiar temperature change localized in a relatively constant surface temperature distribution over a wide range of the entire structure, and thus whether or not the deformed part exists. In the case of analytically detecting the temperature, it is considered that the application of the waveform feature extraction technique corresponding to the deformed portion is effective.

The surface of the structure generally contains various irregularities unrelated to the deformed portion, and when it is in a natural environment, dirt or the like adheres to it, and there are many cases where even minute scratches are present. Such a surface temperature distribution inevitably includes a temperature component that is not caused by the deformed portion as noise. Removing noise from the target surface temperature distribution is desirable in order to improve the image analysis effect for the purpose of detecting deformed parts. As one of such methods, a noise removing method using a wavelet transform by the same inventor as the present application is described in Patent Document 1, and the content thereof is incorporated by reference in the present specification. The present invention also includes the combined use of this noise removing method.

Japanese Unexamined Patent Publication No. 2006-250892

Signal processing and image processing by wavelet, Hiroki Nakano et al., Kyoritsu Shuppan, 1999

In order to detect spatial characteristics and types such as the position, shape, and range of deformed parts existing on the surface or inside of an object from the surface temperature distribution of the object, two main functions are required. The first function is to detect the temperature variation portion included in the surface temperature distribution and detect the position where the deformed portion exists. The second function is to grasp the spatial characteristics more concretely by associating the characteristics of the temperature variation part with the spatial characteristics such as the shape and range of the deformed part, and based on that, the type of deformation. Is to estimate. For the second function, it is necessary to comprehensively utilize not only the knowledge of image engineering but also the knowledge of heat transfer engineering and structural engineering. That is, the analysis from the heat transfer engineering knowledge about the temperature variation part of the surface temperature distribution near the deformed part, especially the shape of the temperature variation part, and the structural engineering knowledge about the deformed part are combined. For the first time, it seems possible to assume the shape of the deformed part of an object, spatial characteristics such as range, and the type and cause of the object. This is because all the temperature distributions of an object are determined by the thermal and spatial characteristics of the entire object and the deformed parts that are locally present in the object, and the heating conditions. This is because spatial characteristics such as the shape and range of the temperature change part are uniquely reflected in the shape and range of the temperature variation part in the surface temperature distribution near the deformed part according to a certain physical law.

However, for example, in the field of civil engineering and construction technology, a conventional technique for detecting a structural defect portion existing on the surface or inside of a structure by a surface temperature distribution is used for a colored shade image visualized based on the surface temperature distribution. From the change in color or shade that appears, the person finally visually estimates the existence of the deformed part. A method for detecting a deformed portion based only on a change in color or shade appearing in such a colored shade image empirically and sensuously detects the presence or absence of a deformed portion of an object and its approximate position. It can be considered that it is nothing more than.

That is, the conventional method for detecting a deformed portion theoretically and analytically determines the characteristics of the shape of the temperature variation portion in the surface temperature distribution and the spatial characteristics such as the shape and range of the deformed portion of the object. It has not yet included the above-mentioned second function of associating and further identifying the types of deformed parts, such as cracks, cavities, floats, and peelings.

In order to solve the above-mentioned problems of the prior art, the in-object deformed portion detecting device according to the present invention is a device that detects a deformed portion existing on the surface or inside of the object by using the surface temperature distribution of the object. hand,
A detection means for numerically evaluating the shape of the surface temperature distribution and detecting a temperature variation portion due to the deformed portion of the surface temperature distribution based on the evaluation result.
A display means for displaying the detected temperature mutation part on the screen,
It is equipped with.

More specifically, detection for detecting a temperature variation portion caused by the deformed portion of the surface temperature distribution by using a filter for each specific region of image data developed using the surface temperature distribution as two-dimensional data. Means and
A display means for displaying a specific area of the detected two-dimensional data on the screen,
It is equipped with.

<Solution Principle of the Present Invention>
To outline, the present invention utilizes the correlation between the spatial characteristics peculiar to the deformed portion of an object and the shape of the temperature variation portion in the surface temperature distribution of the object when heated. , The position where the deformed part exists is found by theoretically and analytically detecting the part where the shape of the temperature variation part peculiar to the deformed part of the object of interest occurs. It is based on the solution principle such as.

First, the shape of the temperature variation part of the surface temperature distribution, which is peculiar to the deformed part to be detected, is examined.
For example, as shown in the crack model for numerical experiments in FIG. 1, when cracks are present near the surface of the object, the surface temperature distribution near the opening is generally the angle δ (hereinafter, inclination) formed by the crack surface and the surface of the object. The angle δ), the cracked surface changes depending on the depth and range of the inside of the object. For the crack model for numerical experiments as shown in FIG. 1, heat transfer analysis was performed under certain conditions, assuming that the surface on the side where the cracks exist was heated. When the inclination angles δ are 30 °, 45 °, and 60 °, the temperature distributions on the surface where the cracks are present are approximately as shown in FIGS. 2A to 2C, respectively.

Regarding the surface temperature distribution shown in FIGS. 2A to 2C, when the crack opening which is the intersection line between the crack surface and the object surface is considered as one line segment, the surface temperature distributed on the perpendicular bisector of that line segment is calculated. The one shown is shown in FIG. The vertical axis shows the temperature difference from the surface temperature of the part where the influence of cracks is small, and the horizontal axis shows the distance in the x direction (in this case, the number of pixels when the temperature distribution is captured in the image).

The structural feature of the temperature variation part in this surface temperature distribution is that there are a maximum temperature point and a minimum temperature point with the crack opening as a boundary, and the temperature is high on the side where the crack surface spreads across the opening. It is lower on the other side. Further, the smaller the inclination angle δ, the larger the temperature difference between high and low, and the direction of the gradient of the surface temperature sandwiching the line segment of the crack opening is substantially perpendicular to this line segment. Therefore, it is considered that the crack occurrence location can be found by analytically detecting the characteristic temperature variation portion in the vicinity of such a crack opening.

Next, the shape characteristics of the temperature variation portion in the surface temperature distribution when a rectangular parallelepiped-shaped cavity exists inside the object as shown by the quarter model of FIG. 4 will be considered. When one of the surfaces of the entire object in which four quarter models are combined is heated, the surface temperature distribution on the heated surface is approximately as shown in FIG. The temperature variation portion of this surface temperature distribution is characterized in that the highest temperature point exists directly above the central part of the cavity and gradually decreases toward directly above the vertical side surface of the cavity. The smaller the distance between the cavity and the heating surface of the object, the higher the surface temperature directly above the cavity. On the other hand, regarding the thickness of the cavity measured in the vertical direction from the surface of the object, when the thermal conductivity between the cavity and the substance constituting the object is significantly different, for example, a structure made of concrete is filled with air. When the cavity is present, even if the thickness of the cavity is small, it acts as a heat insulating layer sufficiently, so that the difference in the thickness of the cavity has a small effect on the surface temperature distribution. Therefore, even deformations such as floating and peeling, which correspond to the case where the thickness of the cavity is small, can be represented by the cavity model. Similarly, the crack width (or thickness), which is the distance between the cracked surfaces in the above-mentioned cracks, has a small effect on the temperature variation portion of the surface temperature distribution.

When an object contains a deformed portion, for example, a crack or a cavity as described above, in order to detect a temperature variation portion included in the surface temperature distribution of the object after heating, for example, in the temperature variation portion. A filter consisting of spatially localized functions having a similar shape is scanned against a specific region of the surface temperature distribution, and the correlation with the function in each region of the surface temperature distribution is obtained, and the correlation is high. It is conceivable to judge that there is a high possibility that such a deformed part exists in the place.

As a specific function used for this scanning, for example, a wavelet of a two-dimensional wavelet transform can be considered. The basic equation in that case is as in Equation 1 when the coordinate axes are x in the horizontal direction and y in the vertical direction.

ここに、ｔ（ｘ，ｙ）は座標（ｘ，ｙ）における表面温度分布に関連する関数であって、表面温度分布そのものだけでなく、表面温度分布に基づく明度画像又はカラー画像を構成する数値などでも代用できる。ψ（ｘ，ｙ）は２次元マザーウェーブレットと呼ばれる関数であり、空間的に局在化した関数である。（ｘ_０，ｙ_０）は原点位置のパラメータ、ａはψ（ｘ，ｙ）を移動及び縮小・拡大させるための変数である。

Here, t (x, y) is a function related to the surface temperature distribution at the coordinates (x, y), and is a numerical value constituting not only the surface temperature distribution itself but also a brightness image or a color image based on the surface temperature distribution. Can be used as a substitute. ψ (x, y) is a function called a two-dimensional mother wavelet, which is a spatially localized function. (X ₀ , y ₀ ) is a parameter of the origin position, and a is a variable for moving, reducing, and expanding ψ (x, y).

In order to scan the surface temperature distribution of an object using the two-dimensional wavelet transform and detect a specific temperature variation part, the two-dimensional mother wavelet ψ (x, y) appropriately characteristics the characteristics of the temperature variation part. It is necessary to set a function that can be transformed. Among such functions, for example, a case where a two-dimensional Gabor wavelet using a Gabor function effective for detecting the local periodicity and directionality of a signal is applied will be described here.

When ψ (x, y) is redefined as a two-dimensional Gabor wavelet (hereinafter abbreviated as Gabor wavelet), one of the general ones is as shown in Equation 2. This is generated by applying a two-dimensional Gaussian window to the complex vibration e ^{i (2πfy + φ)} that vibrates in the y direction. By applying a two-dimensional Gaussian window, the complex vibration e ^{i (2πfy + φ)} is localized around the origin of the Gaussian window.

ここに、ｆは中心周波数、σ_ｘ、σ_ｙは窓幅、及びφは位相をそれぞれ示す。
このようなガボールウェーブレットを用いることによって、温度分布ｔ（ｘ，ｙ）のガボールウェーブレット変換Ψ（ｘ，ｙ）は式３で与えられる。

Here, f is the center frequency, σ _x , σ _y is the window width, and φ is the phase.
By using such a Gabor wavelet, the Gabor wavelet transform Ψ (x, y) of the temperature distribution t (x, y) is given by Equation 3.

また、式２のフーリエ変換Ｆ（ｕ，ｖ）は式４のように求められる。

Further, the Fourier transform F (u, v) of Equation 2 is obtained as in Equation 4.

ここに、Ａ＝２πσ^２、σ_ｕ＝１／（２πσ_ｘ）、σ_ｖ＝１／（２πσ_ｙ）である。このように、ガボールウェーブレット変換は周波数領域で中心周波数ｆを中心とするバンドパスフィルタになっており、帯域幅はσ_ｕ及びσ_ｖによって決定される。

Here, A = 2πσ ² , σ _u = 1 / (2πσ _x ), σ _v = 1 / (2πσ _y ). As described above, the Gabor wavelet transform is a bandpass filter centered on the center frequency f in the frequency domain, and the bandwidth is determined by σ _u and σ _v .

However, since the wavelet of Equation 2 is a function that oscillates only in the y direction, it is necessary to rotate it around the origin in order to respond to the detection of a change traveling in an arbitrary direction. Therefore, ψ _r (x, y, θ _r ) obtained by rotating ψ (x, y) counterclockwise at the origin by an angle θ _r is defined by equations 5, 6 and 7.

以上のようにして得られたガボールウェーブレットψ（ｘ，ｙ）を原点の反時計回りに適当な間隔で設けた角度θ_ｒｉだけ回転させたψ_ｒ（ｘ，ｙ，θ_ｒ）によって温度分布ｔ（ｘ，ｙ）のガボールウェーブレット展開係数Ψ_ｒｉを求める。例えば、その展開係数Ψ_ｒｉの絶対値の二乗和が最大となる角度θ_ｒｉの値θ_ｒｏを求める。

The temperature distribution t by ψ _r (x, y, θ _r ) obtained by rotating the Gabor wavelet ψ (x, y) obtained as described above by an angle θ _ri provided at appropriate intervals counterclockwise at the origin. Find the Gabor wavelet expansion coefficient Ψ _ri of (x, y). For example, the value θ _ro of the angle θ _ri that maximizes the sum of squares of the absolute values of the expansion coefficient Ψ _ri is obtained.

Arranging the absolute values of the Gabor wavelet expansion series corresponding to θ _ro in descending order, for example, if only the expansion coefficient of the upper p% (expansion coefficient residual ratio) is left and the others are set to 0, then the inverse transform is performed and returned. , A surface temperature distribution is obtained in which the temperature variation portion peculiar to the deformed portion similar to the complex vibration e ^{i (2πfy + φ)} is emphasized. The method of selecting the expansion coefficient is not limited to this.

Next, it is examined what kind of complex vibration ^{ei (2πfy + φ)} suitable for detecting the temperature variation portion peculiar to the crack. One of the features of the temperature variation portion shown in FIG. 3 is that, as described above, when the origin is placed at the crack opening, the positive and negative of the temperature are reversed with the origin as the boundary. This is similar to the characteristics of trigonometric functions in which the positive and negative values are reversed with the origin as the boundary. Further, as shown in FIG. 3, the overall temperature distribution of the temperature variation portion is unidirectional in a substantially uniform shape along the crack opening while this sinusoidal-like pattern slightly changes the temperature at the peak. It is configured to be laminated. Therefore, as the complex vibration e ^{i (2πfy + φ)} suitable for detecting such a temperature variation portion, for example, one having a sinus function of sin (2πfy) and a function similar thereto seems to be suitable only in the real part. .. The value of the Gabor wavelet transform Ψ based on the Gabor wavelet ψ _r (x, y, θ _r ) having the above characteristics is the angle with respect to the peripheral region of the magnitudes σ _x and σ _{y in} the temperature distribution t (x, y). The value correlates with the trigonometric function of the frequency f traveling in the _θr direction and a function similar thereto.

Furthermore, a complex vibration ^{ei (2πfy + φ)} suitable for detecting a characteristic temperature variation portion included in the surface temperature distribution of the portion directly above the cavity will be examined. As described above, the temperature variation portion included in the surface temperature distribution of the portion where the rectangular parallelepiped-shaped cavity is contained as shown in FIG. 4 has the highest temperature point directly above the central portion of the cavity, and is located on the vertical side surface of the cavity. It is characterized by a gradual decrease toward the top. This temperature variation portion is similar to the characteristic of the cosine function, in which the value has an apex at the origin when the origin is placed in the center of the cavity, and the value gradually decreases as the origin moves away from the origin. Therefore, as the complex vibration e ^{i (2πfy + φ)} suitable for detecting such a temperature variation portion, for example, one having a cosine function of cos (2πfy) and a function similar thereto only in the real part seems to be suitable. ..

Further, as a method of improving the detection accuracy of the presence of the temperature variation portion and specifying the spatial characteristics of the deformed portion in more detail, for example, a so-called pattern recognition or machine learning method can be adopted. That is, in advance, for an object model having various deformed parts, a numerical experiment or a test piece experiment considering the thermal characteristic conditions, ambient environmental conditions, boundary conditions, heating conditions, etc. of the object is performed, and the cause is caused by the deformed parts. The characteristics of the temperature variation part in the surface temperature distribution of the object generated by the above are obtained as training data. Next, the measurement data of the surface temperature distribution of the object having the deformed portion is collated with the teacher data, and the portion where the temperature variation portion having a high correlation with the teacher data exists is detected. Finally, it is determined that there is a high possibility that there is a deformed part that has a high correlation with the deformed part of the object model that generated the teacher data.

The spatial characteristics of the deformed part include the opening length, width, inclination angle, depth of the tip of the crack from the object surface, etc. for the crack, and the plane seen from the object surface side for the cavity. Shape and area, depth from the surface of the object on the upper and lower surfaces, and the like. However, it is not limited to these.

By the way, as described above, since noise exists in the surface temperature distribution of an actual object, the surface temperature distribution is disturbed by the influence of the noise. As a method generally used in engineering, the standard deviation of the surface temperature distribution of an object having a deformed portion is σ, and the standard deviation σ _w is 2σ in the surface temperature distribution of the crack model for numerical experiments shown in FIG. 2B. The surface temperature distribution with white noise added is shown in FIG. Since noise also disturbs the shape of the temperature variation portion characteristic of the deformed portion, it becomes a major factor that makes it difficult to detect the temperature variation portion from the surface temperature distribution.

In the above method, the above-mentioned deformed portion detection method is applied to the surface temperature distribution in which the noise is reduced by applying the noise removal method described in JP-A-2006-250892 to the surface temperature distribution. The detection accuracy is improved.

In the present invention, the case where the shape of the characteristic temperature variation portion due to the existence of the deformed portion is expressed by using the trigonometric function or the cosine function and a function similar thereto has been described, but the characteristic temperature. By evaluating the shape of the deformed part numerically, for example, the existence of the deformed part is detected from the correlation with a function similar to the three-dimensional curvature, and the details of the spatial characteristics of the deformed part are estimated. Furthermore, it is conceivable to classify the types of deformation.

According to the present invention, structural defects existing on the surface or inside of a structure can be automatically detected without relying on human experience and feeling. Moreover, since it detects defective parts inside the structure from the surface temperature distribution of the object, it is extremely useful as a non-destructive inspection device for bridge piers, roads, tunnels, and the like.

This is a crack model for numerical experiments used for finite element analysis of concrete slabs with cracks near the center. The surface temperature distribution just above the crack is shown in the result of heat transfer analysis under certain conditions when the inclination angle δ of the crack in the crack model of FIG. 1 is 30 °. For the case where the inclination angle δ of the crack of the crack model shown in FIG. 1 is 45 °, the surface temperature distribution just above the crack in the heat transfer analysis result under the same conditions as in the case of FIG. 2A is shown. For the case where the crack inclination angle δ of the crack model of FIG. 1 is 60 °, the surface temperature distribution immediately above the crack in the heat transfer analysis result under the same conditions as in the case of FIG. 2A is shown. Regarding the surface temperature distribution shown in FIGS. 2A to 2C, when the crack opening which is the intersection line between the crack surface and the object surface is considered as one line segment, the surface temperature distributed on the perpendicular bisector of that line segment. Is shown. A quarter of the numerical experimental cavity model used for finite element analysis of a concrete slab with a rectangular parallelepiped cavity near the inside of the center is shown. The surface temperature distribution just above the cavity is shown as a result of heat transfer analysis of the hollow model of FIG. 4 under certain conditions. The surface temperature distribution of the crack model shown in FIG. 2B with Gaussian white noise added to the surface temperature distribution is shown. An isotherm image of the cracked portion when the inclination angle δ of the crack of the crack model shown in FIG. 1 is 45 °, and the image when noise is not included is shown. An image when noise is added to the image of FIG. 7A is shown. The result of performing the detection processing by the function based on the 2D Gabor wavelet including the sine function on the noise-free surface temperature distribution shown in FIG. 7A is shown. FIG. 7B shows the result of performing detection processing by a function based on a two-dimensional Gabor wavelet including a sine function on the surface temperature distribution including noise shown in FIG. 7B. There are cracks of δ = 30 ° near the center of the lower right range, δ = 45 ° in the upper left range, and δ = 60 ° in the lower left range of the image range divided into four, and in the upper right range. An image containing noise in the absence of cracks is shown. It is the result of performing the detection process by the function based on the 2D Gabor wavelet including the sine function, and shows the case where the expansion coefficient residual ratio p is 3%. It is the result of performing the detection process by the function based on the 2D Gabor wavelet including the sine function, and shows the case where the expansion coefficient residual ratio p is 1%. The result of performing the detection process by the function based on the two-dimensional Gabor wavelet including the sine function, and the case where the expansion coefficient residual ratio p is 0.5% is shown. A crack model for a test piece experiment is shown. An image showing the surface temperature distribution after heating of the crack model shown in FIG. 11 is shown. The image shown in FIG. 12 is the result of performing detection processing by a function based on a two-dimensional Gabor wavelet including a sine function, and shows a case where the expansion coefficient residual ratio p is 100%. The result is shown in the same detection process as in FIG. 13A when the expansion coefficient residual ratio p is 10%. Similar to FIG. 13B, this is the result of the detection process in which the expansion coefficient residual ratio p is 10%, and shows the case where noise is removed in advance. An image is shown when the surface temperature distribution in the cavity model having the rectangular parallelepiped-shaped cavity shown in FIG. 4 includes noise. The image shown in FIG. 15 is the result of performing detection processing by a function based on a two-dimensional Gabor wavelet including a cosine function, and shows a case where the expansion series ratio p is 5%. The block diagram of the apparatus is shown.

Hereinafter, an embodiment of the present invention will be described with reference to the block diagram of FIG.
In FIG. 17, the in-object deformation portion detection device 1 executes an arithmetic process described later based on the infrared camera 2 that images the outer surface of the structure and outputs the surface temperature as image data, and the input image data. The computer 3 is provided with a display means 4 for displaying the processing result of the computer 3. Further, the storage means 5 connected to the computer 3 stores an area for storing the input image data as two-dimensional image data, and the above-mentioned filter and teacher data. From the input means 6, the designation of the filter to be used and various other data are input.

The computer 3 stores a program that executes the above-mentioned solution principle, and by executing the program, the shape of the surface temperature distribution is numerically evaluated from the input image data, and the surface is based on the evaluation result. The temperature variation part caused by the deformed part of the temperature distribution is detected. More specifically, it realizes a function as a detection means 31 for detecting a temperature variation portion caused by a deformed portion of the surface temperature distribution by using a filter for each specific region of two-dimensionally developed image data. .. The filter used here is for inputting the image data of a specific area into the above-mentioned formulas 1 to 7 and performing arithmetic processing. Further, the evaluation means 32 for evaluating the correlation between the surface temperature distribution of the teacher data regarding the surface temperature distribution of the object model having the deformed portion on the surface or the inside and the surface temperature distribution of the object are compared with each other. It also realizes a function as an estimation means 33 for estimating a deformed portion from the evaluated teacher data having a large correlation.

Using this computer 3, it was verified whether or not to detect a characteristic temperature variation portion due to the presence of a deformed portion contained in the surface temperature distribution.
First, the verification results for the surface temperature distribution when the crack inclination angle δ of the crack model for numerical experiments shown in FIG. 2B is 45 ° are shown. Regarding the surface temperature distribution when the noise shown in FIG. 2B is not included and the surface temperature distribution when the noise shown in FIG. 6 is included, the respective surface temperature distributions are shown by isotherms in FIGS. 7A and 7B. When the noise shown in FIG. 7A is not included, the isotherms are regularly drawn in a substantially arc shape on both sides of the position of the straight crack opening in the center. On the other hand, when the noise shown in FIG. 7B is included, the regular isotherms are disturbed and buried in the noise, making it difficult to clearly see the presence of cracks. In fact, the infrared image drawn from the surface temperature distribution of the cracked object is close to that of FIG. 7B, and it is not easy to visually recognize the existence of the crack.

8A and 8B show the results of performing detection processing on the surface temperature distributions shown in FIGS. 7A and 7B by a function based on a two-dimensional Gabor wavelet including a sine function in order to detect a temperature variation peculiar to a crack. Is shown. The location of the characteristic temperature variation near the cracked opening, with or without noise, is clearly indicated by the elongated rectangle.

Next, the verification results for the case where the surface temperature distribution when cracks having inclination angles δ of 30 °, 45 °, and 60 ° as shown in FIGS. 2A to 2C are mixed on one surface are shown. FIG. 9 shows the case where the noise shown in FIG. 7B is included, and δ = 30 ° near the center of the lower right range, δ = 45 ° in the upper left range, and lower left in the range of the entire image divided into four equal parts. An image is shown in the case where there is a crack of δ = 60 ° in the range of and no crack is present in the upper right range. 10A, B, and C are the results of detection processing by a function based on a two-dimensional Gabor wavelet including a sine function, and the above-mentioned expansion coefficient residual ratios p are 3%, 1%, and 0, respectively. The case where it was changed to 5% is shown. When p = 3%, the location of the temperature variation characteristic of cracks is clearly shown regardless of the size of the tilt angle. Further, it can be seen that as the expansion coefficient residual ratio p is reduced, only cracks having a small inclination angle can be detected.

Next, the verification results of the surface temperature distribution measured by the crack test piece experiment are shown. The test piece is made of concrete having a size of 450 × 600 × 200 mm as shown in FIG. 11, and has a thickness of 1 mm (width 250 mm, length) so that the angles formed with the surface are 30 ° and 60 °. An acrylic plate that simulates cracks (200 mm and 150 mm) is inserted. FIG. 12 is an image of the surface temperature distribution on the surface after heating when the surface provided with the simulated cracks is directly above and heated by sunlight during the daytime on a clear day.

The image shown in FIG. 12 was subjected to detection processing by a function based on a two-dimensional Gabor wavelet including a sine function. FIG. 13A shows the result of the detection process with the expansion coefficient residual ratio p as 100%. In this case, since the influence of minute stain-like temperature changes remains, the temperature changes in the vicinity of the cracked openings are hidden behind them and do not appear clearly. Therefore, when the expansion coefficient residual ratio p is set to 10%, the result shown in FIG. 13B can be obtained. The influence of minute temperature changes is almost eliminated, and the presence of crack openings with an inclination angle of δ = 30 ° can be clearly detected.

To clarify the existence of a crack opening with an inclination angle of δ = 60 °, either remove the temperature variation part of the crack opening with an inclination angle of δ = 30 °, or scan by changing the frequency f of the Gabor wavelet. It is effective to do.
Here, the effect of using the above-mentioned noise removal method using the wavelet transform as a pretreatment for the surface temperature distribution will be verified. This noise removal method is based on wavelet degeneracy using multiple resolution analysis using orthogonal wavelets. FIG. 14 shows the result of performing the detection process with the expansion coefficient residual ratio p set to 10% after performing this noise removal. The influence of the minute noise remaining in FIG. 13B has been completely removed, and only the isotherms related to cracks are drawn. In this case as well, if the temperature variation portion of the crack opening with an inclination angle δ = 30 ° is removed or the frequency f of the Gabor wavelet is changed and scanned, the crack opening with an inclination angle δ = 60 ° can be further increased. Can be emphasized.

Next, the verification results for the surface temperature distribution of the cavity model for numerical experiments shown in FIG. 4 are shown. It was investigated that the characteristic temperature variation due to the existence of the cavity can be detected by a function based on the two-dimensional Gabor wavelet including the cosine function. FIG. 15 shows an image when the surface temperature distribution of the cavity model includes noise, and the cavity exists at a position directly below the central portion of the image. In this case, the presence of cavities is not clearly visible in the image due to noise. FIG. 16 shows the detection results in the case where the residual development series ratio p = 5%. In FIG. 16, the existence position of the temperature variation portion characteristic of the cavity is detected and displayed as a region surrounded by a substantially elliptical line.

1 In-object deformation part detection device 2 Infrared camera 3 Computer 4 Display means 5 Storage means 31 Detection means 32 Evaluation means 33 Estimating means 51 Crack opening 52 Crack surface 53 Concrete surface 54 Crack surface 55 Crack opening 56 Acrylic plate

Claims (4)

It is an in-object deformation part detecting device that detects a deformed part existing on the surface or inside of the object by using the surface temperature distribution of the object.
A means for expanding the surface temperature distribution as two-dimensional data in a computer memory,
It is a detection means for detecting a temperature variation portion caused by the deformed portion by acting a filter on the surface temperature distribution for each specific region of the developed two-dimensional data, and the action of the filter is a two-dimensional Gabor. Using a wavelet, the surface temperature distribution is wavelet-transformed to obtain an expansion series, only a part of the expansion series is left and the rest is removed, and the expansion series remaining after removal is wavelet-reverse-transformed to specify the above. A detection means that reconstructs the surface temperature distribution for each region,
A display means for displaying the reconstructed surface temperature distribution for each specific region on the screen, and
With
If the deformed portion is cracked, the two-dimensional Gabor wavelet contains a sine function, and if the deformed portion is hollow, floating or peeled, the two-dimensional Gabor wavelet contains a cosine function. Deformation part detection device. Claim 1 further includes estimating the inclination angle of the cracked surface with the surface of the object by changing the expansion coefficient residual ratio of the two-dimensional Gabor wavelet when the deformed portion is a crack. The device for detecting a deformed portion in an object according to the above. The object according to claim 1 or 2, further comprising preliminarily performing noise removal by wavelet degeneracy using multiple resolution analysis of orthogonal wavelets on the surface temperature distribution for each specific region of the developed two-dimensional data. Degenerate part detection device. The device for detecting a deformed portion in an object according to any one of claims 1 to 3, further comprising an infrared camera for measuring the surface temperature distribution of the object.

Images