JP4266864B2 - Corrosion degree evaluation method for hardware - Google Patents

Description

  The present invention relates to a method for evaluating the degree of corrosion of hardware such as a brace that runs over a power transmission line, etc. by recognizing patterns such as rust of the hardware, and in particular, hardware corroded by rust or the like. This is a method for evaluating the degree of corrosion of hardware, which can accurately determine whether or not it can be reused.

Conventionally, hardware such as armatures of steel towers has been used to support power transmission lines and the like. The surface of the hardware is usually plated for rust prevention. However, rust is generated on the hardware due to long-term use of the hardware, and the hardware may be corroded. When the degree of rust is light, the rust can be removed and the hardware can be reused. However, when the degree of rust is significant and corrosion has progressed, the rust is dropped and the hardware is removed. In some cases, it cannot be reused.
Whether or not the hardware can be reused is usually determined by a determination sample based on a photograph of the appearance of the hardware such as rust. Furthermore, it has been proposed to evaluate the degree of deterioration of the steel surface using an image processing method using a computer system (for example, Patent Document 1).
JP 2002-90308 A

However, when the determination is made based on the determination sample, the corrosion status of the hardware is various in color, surface status, and the like, and therefore, a difference in determination of the degree of corrosion by the determiner tends to occur. Furthermore, the determination criterion in this case is determined based on data obtained by examining the corrosion state in detail, but since the number of detailed data is limited, it is difficult to present an optimal determination sample.
Further, in the image processing method using the above-described computer system, only the deterioration degree of the steel material is evaluated, and it is not performed until it is determined whether or not the steel material can be reused.
Therefore, the problem to be solved by the present invention is to acquire the corrosion status as image data without using the above-mentioned specimen for determining the corrosion status of hardware, and to accurately determine whether or not the corroded hardware can be reused by the image data. It is to be able to judge.

In order to solve the above-mentioned problem, the invention according to claim 1 is a method for evaluating the degree of corrosion of hardware in which rust has occurred, wherein the appearance of rust or the like of the hardware is imaged and converted into a signal representing an image. A second step of distinguishing a reusable image of the hardware from the image and a discarded image to be discarded because the hardware cannot be reused, and the reusability of the hardware by a signal representing the image A third step of expressing an image and a discarded image on a feature space; and a fourth step of extracting a reusable image that is difficult to identify and a discarded image that is difficult to identify on the feature space. And a fifth step of generating a discriminant plane between the reusable image that is difficult to identify in the feature space and the discard image that is difficult to identify, and a hardware to be evaluated for corrosion degree Appearance such as rust A sixth step of imaging and converting it into a signal representing the image; and the image obtained in the sixth step is represented on the feature space, and the hardware to be evaluated for corrosion degree is reused based on the discrimination surface And a seventh step of determining whether or not it is possible.
Accordingly, the first step captures the appearance of the hardware such as rust and converts it into a signal representing the image, and the second step reuses the reusable image of the hardware and the hardware among the images. Discard the discarded images that cannot be discarded, and represent the reusable image and the discarded image of the hardware on the feature space by a signal representing the image as the third step, and among the reusable images as the fourth step A difficult-to-identify image and a discarded image that is difficult to identify are extracted, and as a fifth step, a discriminating surface between the reusable image that is difficult to identify and the discarded image that is difficult to identify As a sixth step, an image of the appearance of rust or the like of the hardware to be evaluated for corrosion degree is captured and converted into a signal representing an image, and the image obtained in the sixth step is converted to a signal representing the image as a seventh step. Can be expressed in the feature space, the hardware such corrosion rate evaluated is determined whether reusable, based on the discriminant plane. At this time, since a discrimination plane is generated between the reusable image that is difficult to identify and the discard image that is difficult to discriminate, the reusable image and the discard on both sides of the discrimination plane. Since the distance (margin) between the image and the image can be sufficient, it is possible to accurately determine whether or not the hardware to be evaluated for corrosion degree can be reused based on the determination surface.

Further, the invention according to claim 2 is the fifth aspect of the invention according to claim 1, wherein a discriminant plane is generated between the reusable image that is difficult to identify and the discarded image that is difficult to identify. The step is to use the reusable images that are difficult to identify and the discarded images that are difficult to identify as support vectors for generating the discriminant plane.
Thus, the reusable image that is difficult to identify and the discarded image that is difficult to identify are used as support vectors, and the margin between the reusable image and the discarded image is maximized. The discrimination surface can be formed as follows.

According to the first aspect of the present invention, it is possible to accurately determine whether or not the hardware to be evaluated for corrosion degree can be reused by the computer system.
Further, according to the invention described in claim 2, in addition to the effect of the invention described in claim 1, it is possible to more accurately determine whether or not the hardware to be evaluated for corrosion degree can be reused.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows a system used in the method for evaluating the corrosion degree of hardware according to the present invention, FIG. 2 shows a flowchart at the time of setting in FIG. 1, FIG. 3 shows a discrimination surface in FIG. The flowchart at the time of determination of 1 is shown. 5 shows an example of the mathematical algorithm at the time of setting, FIG. 6 explains the operation of FIG. 5, and FIG. 7 shows an example of the mathematical algorithm at the time of determination. FIG. 8 is an example of a photograph of reusable hardware having a low degree of corrosion, FIG. 9 is an example of a photograph of hardware that is discarded at a high degree of corrosion, and FIG. 10 is related to the present invention. A feature space representing a method for determining the degree of corrosion is shown.

  As shown in FIG. 1, the system used in the method for evaluating the corrosion degree of hardware according to the present invention determines whether or not the system 1 at the time of setting the discrimination surface and the hardware to be evaluated for corrosion degree can be reused. A system 4 for determination is provided. The system 1 at the time of setting includes an input unit 2 (having a CCD camera or the like for imaging) of a set of a hardware image and a determination result of whether or not the image can be reused, and the input unit. 2 includes a determination unit 3 using a pattern recognizer that uses a set of hardware images input from 2 and a determination result obtained by human visual inspection. The determination unit 3 generates a determination surface as shown in FIG. 3 according to the flowchart of FIG. 2 at the time of setting, and the determination unit 3 is configured by a pattern recognizer that recognizes image data on a feature space. Then, the system 4 at the time of determination includes a hardware image input unit 5 (having an imaging CCD camera or the like) for which the corrosion degree is to be determined, the determination unit 3 and the determination result output unit 6. I have. Here, at the time of determination, the determination unit 3 determines whether or not the hardware to be evaluated for corrosion degree can be reused according to the flowchart of FIG. The system 1 at the time of setting and the system 4 at the time of determination are configured by, for example, a computer system. This computer system includes a calculation means such as a CPU, a memory, an input means and a display means. The memory stores a program for carrying out the present invention, such as a program for generating a discriminant plane and a program for the determination method using the discriminant plane.

First, at the time of setting, as shown in the flowchart of FIG. 2, the input unit 2 captures a hardware image corroded by rust as sample data, and converts the characteristics of the captured image into a digital signal.
This digital signal is, for example, data obtained by dividing one hardware image (collection of a large number of pixel data) into several regions and using the average value of the color (RGB color signal) of the pixel data in each region as a parameter. (Step S1). Note that the parameters are not limited to RGB color signals, but may be luminance, the distribution degree of pixels of a specific color, or the like, or these parameters may be used in combination. Next, with respect to the captured image, the degree of corrosion is relatively small among the images, and the degree of corrosion is relatively advanced among the reusable image of the hardware and the images. It is determined that the image is discarded because the hardware cannot be reused (step S2). Based on this determination, for example, reusable images are classified into the +1 class, and discarded images are classified into the −1 class (see FIG. 6). The hardware shown in FIG. 8 is an example that can be reused because the degree of rust 21 is light and has a low degree of corrosion. The hardware shown in FIG. 9 can be reused because the degree of rust 22 is extremely high in degree of corrosion. This is an example of what should be discarded.
Next, the input unit 2 inputs the set of the hardware image and the result of the human visual determination to the determination unit 3, whereby the captured reusable image and the discarded image are input to the feature space of the determination unit 3. 10 (see FIG. 3 and FIG. 6) (step S3). Here, the feature space is a space represented by vectorizing (coordinated) image data with parameters constituting the data, or a non-linear mapping of data vectorized with parameters to a higher dimensional space. Refers to the space. An example represented two-dimensionally for simplicity is shown in FIG. Note that Z ₁ , Z _2, etc. on the feature space 10 in FIG. 6 correspond to the hardware images X ₁ , X _2, etc. Next, extracting the difficult reuse image identification (e.g. Z1, Z ₃₎ the identification of difficult disposal image (e.g. Z2, Z ₄₎ in the feature space 10 (step S4).
Next, a discriminant surface 13 (for example, a hyperplane) is placed in the middle between a reusable image (for example, Z ₁ , Z ₃ ) that is difficult to identify on the feature space 10 and a discarded image (for example, Z ₂ , Z ₄ ) that is difficult to identify. Form (step S5) and finish. An example of the discrimination surface 13 is shown in FIG.

  Next, the determination is made according to the flowchart shown in FIG. First, the appearance of the hardware to be determined by the input unit 5 is imaged as described above and converted into image data (step S11). Next, this image data is expressed on the feature space 10 (step S12), and whether or not the image can be reused is determined based on the discrimination surface 13 (step S13). At that time, it is determined that the image located on the right side of the determination surface 13 in FIG. 3 is reusable, and the image located on the left side of the determination surface 13 is determined as a discarded image that cannot be reused. Finally, the output unit 6 displays the determination result (step S14) and ends.

An example of a detailed mathematical algorithm at the time of setting shown in FIG. 5 is as follows when a pattern recognition method called a support vector machine is used.
First, it is assumed that there are, for example, m pieces of image data X _i as sample data. That is, i = 1,..., M (step S1).
Next, an image is visually determined by human eyes. As described above, the reusable image is classified into the +1 class, the discarded image is classified into the −1 class, and y = + 1 or y = −. 1 (step S2).
Next, the image data X _i is expressed as Z _i on a higher-dimensional feature space by nonlinear transformation (step S3).
Φ: X _i → Z _i
Here, Φ indicates a nonlinear mapping.
Next, a reuse image and a discard image that are difficult to identify are extracted (step S4). Specifically, it is as follows.
Expression of the discrimination surface 13: f (Z) = W ^T Z _i + b
Objective function: (1/2) ‖W‖ ² → Minimize Restriction at this time: y _i (W ^T Z _i + b) ≧ 1 (i = 1,..., M)
Here, the margin (distance) 14 between the reusable image pattern 11 and the discarded image pattern 12 shown in FIG. 10 corresponds to 2 / ‖W‖, and minimizing the above objective function is margin 14. Is equivalent to maximizing. B is a constant.
The problem of minimizing the above objective function can be solved by using the Lagrange multiplier α _i to the following dual problem.
Objective function:
_mm
Σαi− (1/2) Σα _i α _j y _i y _j K (X _i , X _j ) → maximize
^{i = 1 i, j = 1}
Here, the kernel function K is K (X _i , X _j ) = Φ (X _i ) ^T Φ (X _j )
As a constraint condition at this time, αi ≧ 0 (i = 1,..., M)
_m
Σα _i y _i = 0
^{i = 1}
There is. At this time, by obtaining image data X _i whose Lagrange multiplier α _i is not 0, for example, Z ₁ , Z ₂ , Z ₃ and Z _{4 in} FIG. 6 become support vectors.
Next, the discriminant plane 13 is generated in the middle of the reusable images (for example, Z ₁ and Z ₃ ) that are difficult to identify on the feature space 10 and the discarded images (for example, Z ₂ and Z ₄ ) that are difficult to identify (step S5). ). Specifically, the discriminant function f (Φ (X)) representing the discriminant plane 13 is as follows.
_m
f (Φ (X)) = Σα _i y _i k (X, X _i ) + b = 0
^{i = 1}

Next, an example of the mathematical algorithm at the time of determination shown in FIG. 7 is as follows when a pattern recognition method called a support vector machine is used as in the case of the setting.
First, the appearance of the hardware to be judged is imaged, and the hardware image data is represented as hardware image = X.
(Step S11). Then, this image is expressed on the feature space 10 (see FIG. 3) (step S12). Here, Z is a map of X.
Φ: X → Z
Next, new input data X that is to be determined as to whether or not it can be reused is input to an identification function f (Φ (X)) that represents the discriminant plane below, thereby identifying which side of the discriminant plane the input data X is on. Whether or not reuse is possible is determined (step S13).
_m
f (Φ (X)) = Σα _i y _i k (X, X _i ) + b
^{i = 1}
Next, the output unit 6 outputs a determination result: y = sign (f) = + 1 or −1. Here, sign (f) is a sign function and outputs 1 when f> 0 and 1 when f ≦ 0.

From the above, it is possible to determine whether or not the hardware subject to corrosion degree evaluation can be reused based on the determination surface 13. At that time, as shown in FIG. 10, since the margin 14 between the reusable image pattern 11 and the discarded image pattern 12 is maximized on the feature space 10 by the discrimination surface 13, Thus, it is possible to reliably determine whether or not the new data of the hardware to be evaluated for corrosion degree can be reused. For example, it is determined that the new discarded image data 15 is correctly determined and included in the discarded image pattern 12.
Therefore, it is possible to automatically and accurately determine whether or not the hardware subject to corrosion degree evaluation can be reused by the system shown in FIG.

It is a block diagram which shows the system used for the corrosion degree evaluation method of the hardwares which concern on this invention. It is a flowchart at the time of the setting of FIG. It is explanatory drawing which shows the discrimination surface in FIG. It is a flowchart at the time of determination of FIG. It is explanatory drawing which shows the example of the mathematical algorithm at the time of the said setting. It is explanatory drawing of an effect | action of FIG. It is explanatory drawing which shows the example of the mathematical algorithm at the time of the said determination. FIG. 8 is an example of a photograph of reusable hardware having a low degree of corrosion. It is an example of the photograph of the hardwares which the degree of corrosion advances and is discarded. It is explanatory drawing which shows the feature space showing the corrosion degree determination method which concerns on this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 System 2 at the time of setting 2 Input part of pair of hardware image and determination result 3 Determination part 4 by pattern recognition 4 System at the time of determination 5 Input part of hardware image 6 Output part of determination result 10 Feature space 11 Reusable image pattern 12 Discarded image pattern 13 Discrimination surface 21, 22 Rust

Claims (2)

A method for evaluating the degree of corrosion of hardware in which rust has occurred,
A first step of imaging the appearance of the hardware such as rust and converting it to a signal representing an image;
A second step of distinguishing a reusable image of the hardware from the reusable image of the hardware that cannot be reused and discarded.
A third step of expressing a reusable image and a waste image of the hardware on a feature space by a signal representing the image;
A fourth step of extracting the reusable images that are difficult to identify and the discarded images that are difficult to identify on the feature space;
A fifth step of generating a discriminant plane between the reusable image that is difficult to identify and the discarded image that is difficult to identify on the feature space;
A sixth step of imaging the appearance of rust or the like of the hardware to be evaluated for corrosion degree and converting it into a signal representing the image;
A seventh step of expressing the image obtained in the sixth step on the feature space, and determining whether or not the corrosion degree evaluation target hardware is reusable based on the determination surface;
A method for evaluating the degree of corrosion of hardware, comprising: A method for evaluating the degree of corrosion of hardware according to claim 1,
The fifth step of generating a discriminant plane between the reusable images that are difficult to identify and the discarded images that are difficult to identify is that the reusable images that are difficult to identify and the discarded images A method for evaluating the degree of corrosion of hardware, characterized in that a method for determining a support surface for generating a discriminant plane is used which is difficult to identify.