JP2008276805A - Method of creating image for electric wire inspection - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To rapidly and efficiently perform processing for extracting an electric wire part region from an original image. <P>SOLUTION: A direction extending along a direction of an electric wire 4 in the original image 1 out of a horizontal direction and a vertical direction of the original image 1 of the photographed electric wire 4 is set as an electric wire longitudinal direction. A search region 3 in which a horizontal direction range and a vertical direction range are predetermined is distributed in three or more parts in the electric wire longitudinal direction of the original image 1, and an image of the electric wire 4 which is a search object is given beforehand as a template 2. A region most similar to the template 2 is searched within each search region 3, and whether a search result on each search region 3 satisfies predetermined regularity is determined. An image including at least the search result which satisfies the regularity is extracted from the original image 1 and set as an image for electric wire inspection. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a method for creating an image for wire inspection. More specifically, the present invention relates to a technique incorporated and used in a method for detecting an abnormality in the shape or color of an electric wire by image processing.

  In order to maintain and manage the transmission line, a patrol inspection of the transmission line by a helicopter is conventionally performed. One of the purposes of this inspection is the early detection of abnormal parts of electric wires damaged by lightning strikes. Early detection of abnormalities in wires is essential for maintaining the reliability of power supply. The inspection is performed by photographing the electric wire with a video camera mounted on the helicopter, playing the video image, and visually observing the video image by the operator. The operator confirms whether or not there is a broken wire that has a part of the electric wire cut or an arc mark that may cause a broken wire in the future. In order to reduce oversight of the abnormal part of the electric wire, a check by slow reproduction of the video image or a double check by two workers may be performed.

On the other hand, an electric wire abnormality detection method and apparatus in which an image processing technique is applied to a power transmission line inspection work have also been proposed (see Patent Document 1 and Non-Patent Document 1). For example, in the technique of Patent Document 1, as shown in FIG. 29 (A), edges A ₁ and A ₂ that are contour information are detected from the top to the bottom of the left and right ends of the screen, and then the wires are known in advance. B ₁ and B ₂ points that are separated by the diameter of the wire region, and the region surrounded by A ₁ , A ₂ , B ₁ , and B ₂ is defined as the wire region, and based on the concentration information in the wire region, Abnormalities such as wire breakage and arc marks are detected.

JP-A-4-156212 Tonooka et al., "Application of Image Processing Technology to Transmission Line Inspection Work-Automatic Wire Fault Detection Device-" Mitsubishi Electric Technical Report, pp.52-56 Vol.66 No.11 1991

  However, a method for finding an abnormality in the electric wire by visual observation of the video image by the operator requires a great deal of labor and time. When humans continue to work monotonously, their attention is not sustained, and oversight can occur due to fatigue and other effects. Also, it is not light as labor to carry out the work of squinting for a long time.

On the other hand, in the technique of Patent Document 1, as shown in FIG. 29 (B), the strand break 100 that has bounced outward from the back side of the region surrounded by A ₁ , A ₂ , B ₁ , B _2. Cannot be detected. In addition, as shown in FIG. 29C, when _one of the edges A ₁ and A ₂ hits a portion of a strand that has bounced outward, A ₁ , A ₂ , B ₁ , B ₂ The enclosed area 101 is greatly deviated from the true electric wire area, and accurate electric wire abnormality detection cannot be performed. As a result, the operator needs to be reviewed. Further, in order to avoid the above situation, it is conceivable to detect all edges, but if such a detailed edge detection process is performed on the entire image, a great amount of processing time is required.

  Needless to say, the shorter the time from the shooting by the helicopter to the completion of the wire inspection, the better. This is because if it is short, a serious problem that needs to be confirmed locally can be dealt with early and the risk can be reduced.

  Therefore, an object of the present invention is to provide a method for creating an image for electric wire inspection that can perform the process of cutting out the region of the electric wire portion from the original image at high speed and efficiently.

  In order to achieve such an object, the method for creating an image for electric wire inspection according to claim 1 is characterized in that the length along the direction of the electric wire in the original image of the horizontal direction and the vertical direction of the original image in which the electric wire is photographed is The search area in which the horizontal range and the vertical range are determined in advance is distributed in three or more locations in the longitudinal direction of the electric wire of the original image, and the image of the electric wire to be searched is given in advance as a template, and is most similar to the template The search area is searched in each search area, it is determined whether the search result for each search area satisfies a predetermined regularity, and an image including at least the search result satisfying the regularity is cut out from the original image An image for wire inspection is used.

  Since the electric wire is a curved shape with a very small curvature, it can be regarded as almost a straight line, and the electric wire position in the screen has a certain regularity. By performing template matching within the search area and determining whether the search results at three or more locations satisfy a certain regularity, a high-speed and efficient electric wire search is performed.

  According to a second aspect of the present invention, in the method for creating an electric wire inspection image according to the first aspect, when the search result does not satisfy the predetermined regularity, or the similarity of the search result does not satisfy the predetermined criterion. If so, try to add a new template. Therefore, a template can be added as needed, and it is not necessary to prepare a plurality of templates assuming all situations from the beginning.

  According to a third aspect of the present invention, in the method for creating an electric wire inspection image according to the first or second aspect, a template having the highest similarity of search results is selected from a plurality of templates. Therefore, the accuracy of template matching can be improved.

  According to the method for creating an image for electric wire inspection according to claim 1, since an electric wire can be searched at high speed and efficiently, an image for electric wire inspection which is a process of cutting out the region of the electric wire portion from the original image is created. It becomes possible to carry out at high speed and efficiently.

  Furthermore, according to the method for creating the electric wire inspection image according to the second aspect, a template can be added as necessary, and the burden of preparing a plurality of templates is reduced by assuming all situations from the beginning.

  Furthermore, according to the method for creating the electric wire inspection image according to the third aspect, the template with the highest similarity of the search results is selected from the plurality of templates, so that the accuracy of template matching can be improved.

  Hereinafter, the configuration of the present invention will be described in detail based on embodiments shown in the drawings.

  In the present embodiment, an embodiment in which the method for creating an image for wire inspection according to the present invention is used by being incorporated in a wire abnormality detection method and apparatus and program by image processing shown in FIGS. 1 to 28 will be described. For example, as shown in FIG. 11, the wire abnormality detection method includes pre-processing (S1) for obtaining an original image in which a wire is photographed, processing (S2) for cutting out a region of a wire portion from the original image, and a cut-out target image. In contrast, the processing includes detecting a wire abnormality (S3), and outputting the image in which the target image cannot be cut out successfully or the wire abnormality is detected (S4). In addition, the process (S2) which cuts out the area | region of an electric wire part from an original image corresponds to the preparation method of the image for electric wire inspection of this invention.

  In the electric wire abnormality detection process (S3), for example, as shown in FIG. 13, the electric wire is sound using the process (S306) for detecting the edge pixel constituting the actual outline of the electric wire and the detected edge pixel. A process for obtaining an ideal contour line (S307), a process for detecting an abnormality in the shape of the electric wire based on information on the edge pixel and the ideal contour line (S308), and a color information value of an area surrounded by the ideal contour line And the process of detecting an abnormality in the color of the electric wire (S309).

  For example, in the present embodiment, an aerial image of a wire taken by a video camera mounted on a helicopter is used as an image of the wire that is the inspection target. For example, the video camera generates an image of 30 frames per second. Each frame image obtained from the video camera is converted into an RGB color model that can be processed by a computer in the pre-processing (S1), and further an 8-bit grayscale image for simplifying and speeding up the processing. Is converted to Each frame image converted as described above is used as an original image in this embodiment. Each pixel constituting the original image has a brightness value (luminance value) of 256 gradations from 0 (black) to 255 (white) as a color information value. The resolution of the original image is, for example, 640 pixels in the horizontal direction and 480 pixels in the vertical direction. However, the original image is not limited to a gray scale image and may be a color image, and the resolution is not limited to the above example.

  For example, in the original image of this embodiment, it is assumed that the electric wire is photographed over the left and right direction of the screen. For this reason, in this embodiment, the screen horizontal direction is the electric wire longitudinal direction, and the screen vertical direction is the electric wire transverse direction. When the electric wire is photographed in the vertical direction of the screen, the vertical direction of the screen may be the longitudinal direction of the electric wire, and the horizontal direction of the screen may be the horizontal direction of the electric wire.

  Since the portion where the elongated electric wire is photographed is a very small portion of the original image, even if the edge detection process (S306) is performed on the entire original image, it does not contribute much to the edge detection accuracy, and a great amount of processing time is required. It takes. As in the present embodiment, a process (S2) for cutting out a portion where an electric wire has been photographed from the original image is performed, the extracted region is set as a target image, and an edge detection process (S306) is performed on the target image. As a result, the processing time can be greatly reduced. For example, in the present embodiment, the number of pixels in the horizontal direction, which is the wire longitudinal direction of the target image, is the same as the number of horizontal pixels in the original image, and the number of vertical pixels in the wire crossing direction of the target image is the number of vertical pixels in the original image. I try to make it smaller. Specifically, the resolution of the target image is set to 640 × 80 pixels, for example. In this case, the area to be subjected to the edge detection process is narrowed to 1/6 of the original image, and the processing time of the edge detection process can be reduced to about 1/6 of the case where the entire original image is processed.

  In the target image cut-out processing of the present embodiment, for example, as shown in FIG. 12, three or more search areas in which the horizontal range and the vertical range are predetermined are distributed in the horizontal direction, which is the wire longitudinal direction of the original image. Along with (S205), an image of the electric wire to be searched is given in advance as a template (S202), an area most similar to the template is searched in each search area (S206), and a search result for each search area is predetermined. It is determined whether or not regularity is satisfied, and an image including at least a search result satisfying the regularity is cut out from the original image to be a target image (S210).

  The search area in the present embodiment is, for example, vertically long as shown in FIG. 23, the number X of pixels in the horizontal range is the same as the number of horizontal pixels in the template, and the number Y of pixels in the vertical range is the vertical pixels in the template. The number of pixels is larger than the number of pixels and equal to or less than the number of pixels in the vertical direction of the original image. Here, reference numeral 1 in FIGS. 23 and 24 indicates an original image, reference numeral 2 indicates a template, reference numeral 3 indicates a search area, and reference numeral 4 indicates an image of an electric wire. However, the size of the search area is not limited to the above example. The larger the search area, the more advantageous it is that the possibility of template matching failure can be reduced. Conversely, the smaller the search area, the more advantageous it is that the processing speed can be increased. For example, if the position of the electric wire greatly changes beyond the vertical range Y of the search area between successive frames, the electric wire is not included in the search area, and template matching in the search area fails. The possibility of the failure can be reduced by increasing the vertical range Y, and the failure can be eliminated if it is equal to the number of pixels in the vertical direction of the original image. On the other hand, by making the horizontal range X larger than the number of horizontal pixels of the template, it is possible to search in the horizontal direction in each search region. For example, when an electric wire equipped with a snow ring is photographed, there is a possibility that a region similar to the template does not appear in the vertical direction. By doing so, the possibility of template matching failure can be reduced.

  Further, for example, as shown in FIG. 24, the search areas in the present embodiment are distributed at three places with a predetermined distance in the horizontal direction which is the longitudinal direction of the electric wire. However, the number of search areas may be three or more. Here, in the present embodiment, each search area is not moved in the horizontal direction, but is moved only in the vertical direction. For example, each search area is moved in the vertical direction so that the search result in the previous frame is located at the center. This is because the position and angle of the electric wire in the screen do not vary greatly between successive frame images, and the electric wire can be tracked well in a range Y smaller than the number of pixels in the vertical direction of the original image. It is also a device.

  For example, the degree of difference due to a square error is used to calculate the similarity between an area having the same size as the template in the search area (hereinafter referred to as a candidate area) and the template. In this case, if the brightness value of the template coordinates (x, y) is v (x, y) and the brightness value of the candidate area coordinates (x, y) is v ′ (x, y), the dissimilarity err is It is calculated by the following formula.

  A candidate area having the smallest difference err is set as an area most similar to the template in the search area, that is, a search result. However, the method of calculating the similarity between the candidate area and the template is not limited to the above example. For example, by using a method such as using cross-correlation for similarity calculation or determining similarity in the frequency domain, it is possible to perform more stable template matching with reduced influence of illuminance change.

  In the present embodiment, as “regularity of search results” that is a criterion for determining success or failure of search, it is determined whether or not the phases of displacement of search results are aligned. The phase here refers to that used in “topology”. For example, if two graphics expressed by line segments have the same shape, it is determined that the two graphics have the same phase, that is, the phases are aligned. In the case of this embodiment, if the three search results are, for example, the first, second, and third search results in order from the left, the slope of the line segment connecting the first search result and the second search result, If the slopes of the line segments connecting the second search result and the third search result can be regarded as the same, it is determined that the phases are aligned. The reason why the success or failure of the search can be determined depending on whether or not the phases of the displacements of the search results are aligned will be described below. That is, the electric wires are not in contact with other objects except for the steel tower portion, and are in a state of hanging according to the laws of physics. Therefore, the electric wire has a curved shape with a very small curvature, and can be regarded as a straight line within a small range within the screen without any problem. Therefore, the displacement of the wire position in the screen is relative to the movement of the helicopter or camera. The helicopter keeps flying at a constant speed along the electric wire and does not change its advancing direction suddenly. In addition, the camera does not rotate and only pans in the vertical and horizontal directions. As described above, the electric wire can move in the vertical direction in the three search areas with time change, but the movement is usually aligned in the three search areas. Therefore, if the phase of the displacement of each search result is not aligned, it is considered that the search has failed.

An example of numerically expressing whether or not the phase of displacement of each search result is aligned is shown below. _Let y _1i , y _2i , and y _3i be the predetermined coordinate positions of each search result, for example, the vertical coordinate positions of the center point in each search result. Subscripts i = 0,..., P. Here, y ₁₀ , y ₂₀ , y ₃₀ are the coordinate positions in the current original image, y ₁₁ , y ₂₁ , y ₃₁ are the coordinate positions in the original image one frame before, and y _1p , y _2p , y _3p are The coordinate position in the original image before p frames is used. Then, based on the following formula, an average dy12rate of the difference between the distance between y _1i and y _2i and the distance between y _2i and y _3i in a certain period is calculated, and if the average dy12rate is within a predetermined constant value It is determined that the phase of the displacement of each search result is aligned, and when it exceeds the certain value, it is determined that the phase of the displacement of each search result is not aligned.

However, the determination of whether or not the phase of the displacement of each search result is aligned is not necessarily limited to that based on the above calculation. For example, the following more general calculation method may be used. For example, y1 = (y ₁₀ ,..., Y _1p ), y2 = (y ₂₀ ,..., Y _2p ), y3 = (y ₃₀ , y _3i using the above-described vertical coordinate positions y _1i , y _2i , y _3i . .., Y _3p ), and the cross-correlation of y1, y2, and y3 is calculated. Here, cor12 is a cross-correlation vector between y1 and y2, cor23 is a cross-correlation vector between y2 and y3, and cor31 is a cross-correlation vector between y3 and y1. In each cross correlation vector cor12, cor23, cor31, it is assumed that the maximum value is the j1, j2, j3th element, respectively. When these j1, j2, and j3 are both equal to or less than a certain value, it is determined that the phase of the displacement of each search result is aligned, and when the value exceeds the certain value, the phase of the displacement of each search result is not aligned. Judge. Note that j1, j2, and j3 are 0 when the phases are completely aligned. In the case of this calculation method, it is possible to detect whether or not the phase of displacement of each search result is aligned with higher accuracy. However, the calculation time is longer than the method based on Equation 2.

  Here, even if the wires are of the same system, the actually captured images of the wires are not always constant. The causes include, for example, a difference in brightness due to the illuminance setting of the camera, a difference in brightness due to a mountain shadow, etc., and a difference in the surface state of the wire such as dirt or paint attached to the wire. It is difficult to express a plurality of variations of these electric wire images with one template, and it is desirable to use a plurality of templates properly. However, it is not always easy to prepare a template of sufficient and appropriate quantity and quality from the beginning.

  Therefore, in the present embodiment, in order to facilitate the work of preparing a plurality of templates and improve the accuracy of template matching, for example, as shown in FIG. 12, the search result does not satisfy a predetermined regularity (S208; No) ) Or when the similarity of the search results does not satisfy a predetermined criterion (S209; No), a new template can be added (S213).

  For example, when it is determined that the phases of the displacements of the search results are not aligned, that is, when the search fails, the user interface screen is displayed so that the operator can add a new template. Even if it is determined that the phase of displacement of each search result is aligned, if the similarity between the search result and the template is smaller than a certain value, in other words, the difference exceeds a certain value. If it is larger, there is a possibility that something other than the electric wire is returned as a search result, and a new template is considered necessary. Therefore, for example, the user interface screen is used to notify the situation where the similarity is smaller than a certain value, and the worker can add a new template. For example, the similarity is displayed as a graph, and the portion where the similarity is smaller than a certain value is displayed in different colors to indicate to the operator that a new template is necessary. The above-described user interface screen includes an original image in which the phase of the displacement of the search result is not aligned or an original image in which the similarity between the search result and the template is less than a certain value, and the operator uses a pointing device such as a mouse. Can be used to designate the electric wire portion in the original image, and the designated electric wire portion can be registered as a new template. The operator only needs to determine whether or not to use the wire portion in the original image displayed on the user interface screen as a new template, and can prepare a sufficient and appropriate amount and quality template with less effort. .

  Here, if the search fails or the similarity of the search results does not satisfy a predetermined criterion, a new template may be added immediately from the above user interface screen, but added after a series of processing. You may make it do. For example, an original image in which the search has failed or an original image in which the similarity between the search result and the template is less than a certain value is recorded in the log without cutting out the target image (S214). Then, after finishing the target image clipping process for the subsequent original image, the operator adds a new template and performs the target image clipping process again for the original image recorded in the log. Also good.

  Next, a method for performing a stable electric wire search using a plurality of templates will be described. For example, in the present embodiment, a template having the highest similarity of search results is selected from a plurality of templates. For example, calculate the similarity between the current search area and all prepared templates, select the template with the highest similarity, and use the selected template for processing the original image from the current and next time. Try to use it. However, the search area of the original image at the current time is searched using all templates, the search result having the highest similarity is adopted, and the template from which the search result is obtained is used as the original image from the next time. You may make it use for the process with respect to. Alternatively, a search is performed using a plurality of templates in a predetermined order with respect to the search area of the original image at the current time, and the first template that satisfies the predetermined criteria for the similarity of the search results is determined from the current time and the next time The original image may be used for processing. In this case, for example, the template may be used in order from the newest template, or may be used in order from the most recently used template, or may be used in order from the template with the highest use frequency. Also good.

  Next, the edge pixel which comprises the actual outline of an electric wire is detected with respect to the object image cut out from the original image. Among the target images obtained by cutting out the electric wire portion from the original image, only the electric wire has the edge, that is, the contour information clearly appears, and the electric wire region can be accurately extracted by detecting the edge pixel.

  For example, with respect to an image as shown in FIG. 7, as indicated by an arrow in the figure, the luminance value between two adjacent pixels in one direction in the vertical direction that is the electric wire crossing direction, for example, in the upward direction in the figure. FIG. 8 shows the relationship between the difference value obtained and the vertical coordinate position. Among the points protruding from the difference value 0 appearing in FIG. 8, in other words, the difference value is larger or smaller than both of the difference values on both sides (hereinafter also referred to as a peak), the absolute value is large. The thing has shown the electric wire part. Therefore, if both ends of the peak indicating the electric wire portion can be detected, the electric wire region can be accurately extracted.

  Here, there is a method of searching for a peak where the absolute value of the difference value exceeds the threshold and determining the pixel constituting the peak as an edge pixel. However, in this case, an appropriate threshold fluctuates depending on the sun direction, weather, and the like Therefore, accurate edge detection cannot be performed under a certain threshold.

  Therefore, in the edge detection processing of the present embodiment, for example, as shown in FIG. 14, for each pixel row in the wire crossing direction in the target image, the difference in color information value between two adjacent pixels facing in one direction in the wire crossing direction. The value is obtained (S306-5). Then, as shown in FIG. 15, a point where the absolute value of the difference value is maximum is obtained (S306-7-1), and is within a predetermined range from the maximum point in both directions in the electric wire crossing direction and A point that is farthest from the maximum point and protrudes from the difference value 0 is obtained, and the pixels constituting each protruding point are determined as edge pixels (S306-7-6, S306-7-11).

  The above principle will be described with reference to FIG. 9 which is an enlarged view of FIG. A point P1 where the absolute value of the difference value is maximum is searched for in FIG. Further, a peak P2 that is within a predetermined range Δj and is farthest from the maximum point P1 around the maximum point P1 is searched. Here, the range Δj for searching for the peak P2 is determined based on the thickness of the electric wire in the target image, for example. For example, it is assumed that the thickness of the electric wire is about 40 pixels, and the outermost peak P2 is searched for in the range Δj of ± 20 pixels around the maximum point P1. And let the pixel which comprises the peak P2 be an edge pixel. Here, each of the peaks P2 is composed of two pixels, and the one closer to the maximum point P1 is an edge pixel. As a result, two edge pixels are obtained. One of the two edge pixels is called a first edge pixel, and the other is called a second edge pixel. In the present embodiment, the upper edge pixel in the target image is the first edge pixel, and the lower edge pixel is the second edge pixel. The above processing is performed for all the vertical pixel rows of the target image.

  Next, based on the edge pixel detected by the above-mentioned process, the process which calculates | requires the ideal outline in case an electric wire is healthy is performed. For example, as shown in FIG. 16, low-reliability pixels that do not satisfy a predetermined criterion are excluded from the first edge pixel group and the second edge pixel group, and the remaining first edge pixel group and second edge pixel group are respectively straight lines. To obtain two ideal contour lines (S307-1 to S307-3).

  For example, in the present embodiment, the following processing is performed on the first edge pixel group and the second edge pixel group, respectively, so that low-reliability pixels that greatly impair the linearity are removed. That is, as shown in FIG. 17, each edge pixel is a pixel of interest, and a vertical range that is a predetermined range in the horizontal direction that is the longitudinal direction of the wire around the pixel of interest, for example, ± 15 pixels is the wire transverse direction of the edge pixel. The average of the coordinate positions in the direction is obtained (S307-1-3, S307-1-5). Then, an edge pixel as a pixel of interest whose vertical coordinate position is significantly different from the average, that is, an edge pixel as a pixel of interest whose absolute value of the difference between the vertical coordinate position and the average is equal to or greater than a predetermined threshold value. The pixels are excluded as low-reliability pixels (S307-1-4, S307-1-6).

  In addition, an edge pixel in which the distance between the paired first edge pixel and the second edge pixel is extremely narrow or wide, in other words, the vertical direction between the first edge pixel and the second edge pixel having the same horizontal coordinate position. Edge pixels whose direction coordinate position difference is not less than a predetermined upper limit value or less than a lower limit value are excluded as low reliability pixels (S307-1-7, S307-1-8).

  As a method of applying a straight line to the edge pixel group excluding the low reliability pixels, for example, a least square method is used in the present embodiment. However, the method of linear approximation is not limited to this example. The straight lines obtained based on the first edge pixel group and the second edge pixel group excluding the low reliability pixels are referred to as a first ideal contour line and a second ideal contour line, respectively.

  In the shape abnormality detection process using the ideal outline (S308 in FIG. 13), the edge pixels that are separated from the ideal outline by a predetermined distance in the vertical direction that is the electric wire crossing direction are horizontal in the electric wire longitudinal direction. When a predetermined number of lines continue in the direction, it is determined that there is a possibility that an abnormality has occurred in the shape of the electric wire due to a broken wire or the like. For example, when an edge pixel that is 10 pixels or more away from the ideal contour line appears over a length of 20 pixels, it is determined that there is a possibility of a wire abnormality such as a broken wire. This utilizes the fact that when a part of the electric wire is cut off and jumps to the outside, the edge pixels come together to some extent at some distance from the ideal contour line. And according to the above-mentioned process, the shape abnormality of the electric wire resulting from a wire break etc. can be detected from the relationship between an ideal outline and an edge pixel, and an operator's labor can be reduced significantly.

  On the other hand, in the color abnormality detection process (S309 in FIG. 13), a representative value in the wire transverse direction is obtained from the color information value of the pixel row in each wire transverse direction sandwiched between the first and second ideal contour lines, and the color information When representative values that deviate from the threshold value continue for a predetermined number in the longitudinal direction of the electric wire, it is determined that there is a possibility that the color of the electric wire may be abnormal due to wire breakage, arc marks, or scratches. .

  For example, in this embodiment, the average value of the luminance values of the vertical pixel columns sandwiched between the first and second ideal contour lines is set as the representative value in the vertical direction. However, the present invention is not limited to this example, and the mode value of the luminance value of each vertical pixel column sandwiched between the first and second ideal contour lines may be used as the representative value in the vertical direction.

  In the present embodiment, the threshold value of the color information to be compared with the representative value is calculated for each target image. For example, as shown in Formula 3, a value obtained by multiplying the standard deviation of the representative value obtained for a certain target image by α is added to the average value of the representative values obtained for the target image, and the value is added to the upper limit of the target image. And a value obtained by subtracting a value obtained by multiplying the standard deviation of the representative value by α from the average value of the representative values is set as the lower limit threshold value. Here, α is an arbitrary coefficient. When α was changed from 0.5 to 1.5, it was possible to accurately detect an abnormal electric wire particularly when α = 0.9. Therefore, in this embodiment, α = 0.9 is set.

<Equation 3>
Threshold = average value of representative values ± α × standard deviation of representative values

  When a representative value that is equal to or higher than the upper limit threshold is continuous for 20 pixels or more in the horizontal direction, or when a representative value that is equal to or lower than the lower limit threshold is continuous for 20 pixels or more in the horizontal direction, It is determined that there is a possibility of wire abnormality. This utilizes the fact that when an electric wire has an arc mark or a flaw, the abnormal portion becomes darker or whitish than a normal portion of the electric wire. And according to the above-mentioned processing, it is possible to automatically and accurately detect an abnormality in the color of the electric wire due to arc marks or scratches from the tendency of the color information value of the region surrounded by the ideal outline, Can greatly reduce labor. In addition, by using the color information value of the region surrounded by the ideal outline, it is possible to prevent color information values such as a background that is not an electric wire from being included, and to accurately detect a color abnormality.

  Here, in the above-described color abnormality detection process, the average value of the luminance values of the pixel rows in the electric wire transverse direction sandwiched between the first edge pixel and the second edge pixel may be used as the representative value in the electric wire transverse direction. It is done. However, in this case, if there is a break in the wire part of the wire is jumping outside, the luminance value of the non-wire part, for example, the luminance value of the background is also included in the average value calculation, and an accurate color abnormality It cannot be detected. As in this embodiment, by calculating the average of the luminance values in the region sandwiched between the first and second ideal contour lines and using it as the representative value, the luminance value of the portion that is not the electric wire is included in the calculation of the average value. This provides an advantage of preventing and performing accurate color abnormality detection.

  Here, the target image of the present embodiment is obtained from an aerial image, and an abnormal portion of the electric wire appears over several frames. Therefore, in the present embodiment, it is determined that an abnormality has occurred in the electric wire when the target image that is determined to have the possibility that the electric wire has an abnormality continues for a predetermined number, for example, two frames or more. I have to. Thereby, a more reliable electric wire abnormality detection can be performed. However, if the flight speed of the helicopter is slow, etc., abnormal parts appear in more continuous frames. Therefore, if more than a number of frames continue, it may be determined that an abnormality has occurred in the electric wire. good.

  The wire abnormality detection method described above can be implemented as a wire abnormality detection device using, for example, a known computer. For example, as shown in FIG. 10, the electric wire abnormality detection device 10 includes an input interface 11 to which video images of electric wires are input, and a hard disk as an external storage device 12 in which data such as original images and target images are recorded. A RAM as a main storage device 13 in which temporary work data and the like are recorded, an output device 14 in which an image in which an abnormality is detected is output, a central processing unit (CPU) 15 and the like. . The hardware resources are electrically connected through the bus 16, for example. The input interface 11 converts, for example, a signal input from a video camera or read from a medium such as a DVD on which video is recorded into data that can be processed by a computer, and each frame constituting the video is an image. It has a function of recording in the external storage device 12 as data. As such an input interface 11, for example, an existing NTSC-RGB converter, a frame grabber, an image capturing board for a personal computer, or the like may be used. The output device 14 is a display, for example, and displays an image in which an abnormality is detected, a user interface screen, and the like. The external storage device 12 stores an electric wire abnormality detection program in which the method for creating an image for electric wire inspection according to the present invention is incorporated, and the computer is read and executed by the CPU 15 so that the computer can detect the electric wire abnormality. It functions as the detection device 10.

  An example of processing executed by the wire abnormality detection device 10 by the wire abnormality detection program of the present embodiment will be described along the flowcharts of FIGS. First, pre-processing is performed (S1 in FIG. 11). FIG. 22 shows a flowchart detailing this preprocessing. Image data is read into the external storage device 12 from the video camera or the medium on which the video is recorded via the input interface 11 (S101). Then, the image data recorded in the external storage device 12 is converted into, for example, an 8-bit grayscale image (S102). A Gaussian filter is applied to the original image thus obtained to remove noise components from the original image (S103). In order to avoid such a situation, sudden noise such as impulse noise appears in the image and a large peak is generated in the difference calculation in the edge detection process, causing erroneous detection of the edge. It is. This completes the preprocessing.

  After the preprocessing is completed, the target image is cut out (S2 in FIG. 11). FIG. 12 shows a detailed flowchart of the target image cut-out process. In this process, first, 1 is set to the frame counter n (S201). Then, a template registration process is performed (S202). For example, the screen shown in FIG. 25 is displayed. In this screen, when the operator clicks the “shoot” button 21 with a pointing device such as a mouse, the original image of the nth frame, that is, the first frame is displayed in the window 22. Reference numeral 4 denotes an image of the electric wire. Further, when the operator clicks a point on the electric wire image 4 with a pointing device, an area 23 having a predetermined size centered on the clicked point is displayed. Then, when the operator clicks the “confirm tracking point” button 24 with a pointing device, the region 23 is registered as a template. Then, an area similar to the template is automatically searched in the vertical direction at a position away from the center of the area 23 in the horizontal direction by a predetermined range, for example, 30 pixels left and right, and the search result 28 is shown in FIG. As shown on the screen. The three areas 28, 23, 28 are the first search results, and the respective vertical coordinate positions are recorded. When the operator clicks the “tracking start” button 25 with the pointing device, the following wire position tracking process is performed. If a template has already been registered, the above template registration process (S202) may be skipped.

  When the n-th frame original image exists (S203; Yes), the original image is read (S204). Here, since the wire search is performed for the original image of the first frame when the above template is registered, it is possible to add 1 to the frame counter n in the template registration process (S202). good. Three search areas are arranged for the read original image (S205). Each search area is arranged such that the vertical coordinate position of the previous search result at the corresponding horizontal position is located at the center. Then, a template matching process for searching for a region most similar to the template in each search region is performed (S206). FIG. 27 shows an example of a screen in which electric wire tracking is performed. Reference numeral 3 indicates a search area, and reference numeral 29 indicates a search result. Here, in the window 27, for example, a graph 31 showing a change with time of the result of calculating the difference between the central search result and the template in each frame is displayed. A straight line 30 in the window 27 is a reference line representing the minimum value of the dissimilarity, and indicates that the dissimilarity is higher, that is, the similarity is lower as the graph 31 is further away from the reference line 30.

  The vertical coordinate position of each search result is obtained (S207), and using the recorded vertical coordinate position of the past search result, it is determined whether the phase of the displacement of each search result is aligned based on, for example, Equation 2. (S208). When the phase of the displacement of each search result is aligned (S208; Yes) and the similarity of each search result satisfies the defined criteria (S209; Yes), an image including at least each search result is cut out from the original image. The nth target image is recorded in the external storage device 12 (S210). Then, the vertical coordinate position of each search result is recorded in order to arrange the next search area and to determine whether the phase of the displacement of each search result is aligned (S211). Then, 1 is added to the frame counter n (S212), and the processing from S203 onward is repeated for the original image of the nth frame which is the next frame.

  On the other hand, if the phase of the displacement of each search result is not aligned (S208; No), or if the similarity of each search result does not satisfy the defined criteria (S209; No), the operator can add a template (S213). ). For example, the screen shown in FIG. 28 shows an example in which the appearance of the electric wire changes because the background of the electric wire changes and the illumination condition changes, and the similarity of the search results does not satisfy the defined criteria. The graph 31a is far away from the reference line 30 and shows that the degree of difference is high. For example, the screen is displayed for a certain period of time to indicate to the worker that a new template is necessary. The operator can add a new template by clicking the “add reference data” button 26 displayed on the screen with a pointing device and clicking a point on the wire image 4 in the window 22. When a template is added (S213; Yes), a template matching process using the added template is performed (S206). A graph 31b in FIG. 28 shows a state in which the degree of difference is reduced by adding a template.

  When the template is not added by the operator (S213; No), the original image is recorded in the log (S214). The log is a recording area provided in the external storage device 12, for example. Then, 1 is added to the frame counter n (S212), and the processing from S203 onward is repeated for the original image of the nth frame which is the next frame.

  When the above processing is performed for all the original images (S203; No), the target image cutout processing is terminated, and the wire abnormality detection processing is performed (S3 in FIG. 11). FIG. 13 shows a detailed flowchart of the electric wire abnormality detection process. In this process, 1 is set in the image counter k (S301), 0 is set in the shape abnormality counter z (S302), and 0 is set in the color abnormality counter q (S303). If the k-th target image exists (S304; Yes), the target image is read (S305), the edge detection process (S306), the ideal contour estimation process (S307), the shape are read for the read target image. An abnormality detection process (S308) and a color abnormality detection process (S309) are performed. After completion of these processes, 1 is added to the image counter k, and the processes from S304 onward are repeated for the kth target image corresponding to the next frame. If the above process is performed about all the target images (S304; No), an electric wire abnormality detection process will be complete | finished.

  FIG. 14 shows a detailed flowchart of the edge detection process. In this process, 1 is set to the pixel number i in the horizontal direction, which is the longitudinal direction of the electric wire (S306-1). Then, it is determined whether the value of the horizontal pixel number i exceeds the horizontal pixel number i_max of the target image (S306-2). If the horizontal pixel number i is equal to or less than i_max (S306-2; Yes), 1 is set to the pixel number j in the vertical direction that is the electric wire crossing direction (S306-3). Then, it is determined whether the value of the vertical pixel number j exceeds the value obtained by subtracting 1 from the number of vertical pixels j_max of the target image (S306-4). If the vertical pixel number j is equal to or smaller than j_max−1 (S306-4; Yes), the difference between the luminance value of the pixel at the coordinate (i, j) and the luminance value of the pixel at the coordinate (i, j + 1) is obtained. The difference value (j) is recorded (S306-5). Then, 1 is added to the vertical pixel number j (S306-6), and the processes after S306-4 are repeated. When the value of the vertical pixel number j exceeds j_max-1 (S306-4; No), all the luminance value difference values between two adjacent pixels in the vertical pixel row at the horizontal pixel number i are obtained. Next, edge pixel search processing is performed (S306-7).

  FIG. 15 shows a detailed flowchart of the edge pixel search process. In this process, the difference value (j ′) having the maximum absolute value is obtained from the recorded difference values, and the vertical pixel number j ′ corresponding to the difference value (j ′) is obtained (S306-7-1). . A value obtained by adding, for example, 20 to Δ ′ as Δj is set as J (S306-7-2). Then, it is determined whether or not the difference value (J) corresponding to the vertical pixel number J constitutes a peak. Specifically, whether difference value (J)> 0, difference value (J)> difference value (J-1), and difference value (J)> difference value (J + 1) (S306-7-) 3) or whether difference value (J) <0 and difference value (J) <difference value (J-1) and difference value (J) <difference value (J + 1) (S306-7-) 4). When the difference value (J) constitutes a peak (S306-7-3; Yes or S306-7-4; Yes), the pixel at the coordinates (i, J) is recorded as the first edge pixel (S306). -7-6). When the difference value (J) does not constitute a peak (S306-7-3; No and S306-7-4; No), 1 is subtracted from the vertical pixel number J (S306-7-5), and J is the center. One pixel is brought closer to the vertical pixel number j ′, and it is examined again whether or not the difference value (J) corresponding to the vertical pixel number J forms a peak. Similarly, a value obtained by subtracting 20 as Δj from j ′ is set as J ′ (S306-7-7), and it is determined whether or not the difference value (J ′) corresponding to the vertical pixel number J ′ constitutes a peak ( S306-7-8, S306-7-9), when forming a peak, the pixel at coordinates (i, J ′) is recorded as the second edge pixel (S306-7-11), and no peak is formed 1 is added to the vertical pixel number J ′ (S306-7-10), and J ′ is brought close to the vertical pixel number j ′ centered by one pixel, and whether or not the difference value (J ′) forms a peak. Judge again. As described above, a pixel that constitutes a peak that is in the range ± Δj determined from the point j ′ where the difference value is maximum and is farthest from the maximum point j ′, that is, an edge pixel is obtained.

  When the first edge pixel and the second edge pixel are obtained for the horizontal pixel number i, the process returns to the process of FIG. 14, and 1 is added to the horizontal pixel number i (S306-8), and S306-2 for the horizontal pixel number i. The following process is repeated. As described above, the first edge pixel and the second edge pixel are obtained for each vertical pixel column constituting the target image, in other words, for each horizontal pixel number. When the value of the horizontal pixel number i exceeds the horizontal pixel number i_max of the target image (S306-2; No), the edge detection process in the target image is terminated.

  When the edge detection process is completed, an ideal contour estimation process (S307) is performed. FIG. 16 shows a detailed flowchart of the ideal contour estimation process. In this process, first, a removal candidate pixel search process is performed (S307-1). Further, in this process, as shown in detail in FIG. 17, 1 is set to the pixel number i in the horizontal direction which is the longitudinal direction of the electric wire (S307-1-1). Then, it is determined whether the value of the horizontal pixel number i exceeds the horizontal pixel number i_max of the target image (S307-1-2). If the horizontal pixel number i is less than or equal to i_max (S307-1-2; Yes), the average value of the coordinate positions in the vertical direction that is the wire crossing direction for the first edge pixels existing in the horizontal pixel numbers i−Δi to i + Δi Is determined as Ave1 (S307-1-3). Δi is set to 15, for example. Then, an edge pixel in which the absolute value of the difference between the vertical coordinate position of the first edge pixel of the horizontal pixel number i and the average value Ave1 is equal to or greater than a predetermined threshold is recorded as a removal candidate pixel (S307-1). -4). Similarly, for the second edge pixels existing in the horizontal pixel numbers i−Δi to i + Δi, the average value of the coordinate positions in the vertical direction is obtained and set as Ave2 (S307-1-5), and the second edge pixel of the horizontal pixel number i Edge pixels whose absolute value of the difference between the vertical coordinate position and the average value Ave2 is equal to or greater than a predetermined threshold value are recorded as removal candidate pixels (S307-1-6). Further, the difference between the positions of the first edge pixel and the second edge pixel in the horizontal pixel number i in the vertical direction is obtained and set as ΔD (S307-1-7). When ΔD is greater than or equal to a predetermined upper limit value or less than a lower limit value, the first edge pixel and the second edge pixel at the horizontal pixel number i are recorded as removal candidate pixels (S307-1-8). Then, 1 is added to the horizontal pixel number i (S307-1-9), and the processing from S307-1-2 onward is repeated for the horizontal pixel number i. As described above, an edge pixel greatly deviating from the average of surrounding edge pixels whose vertical coordinate position is within a range of ± 15 pixels, or an edge where the distance between the paired first edge pixel and the second edge pixel is extremely narrow or wide Pixels are recorded as removal candidate pixels. When the value of the horizontal pixel number i exceeds the horizontal pixel number i_max of the target image (S307-1-2; No), the removal candidate pixel search process in the target image ends.

  When the removal candidate pixel is obtained, the process returns to the process of FIG. 16, and the remaining pixels excluding the removal candidate pixel, that is, the low-reliability pixel that impairs the linearity of the edge, from the first edge pixel group are used. A first contour line is created (S307-2). Similarly, a second contour line is created by the least square method using the remaining pixels obtained by removing the removal candidate pixels from the second edge pixel group (S307-3). Thereby, the ideal outline estimation process in the target image is finished, and the shape abnormality detection process (S308) and the color abnormality detection process (S309) are performed on the target image.

  18 and 19 show flowcharts detailing the shape abnormality detection processing. In this processing, 0 is set in the counters w and w '(S308-1 and S308-2), and 1 is set in the horizontal pixel number i which is the longitudinal direction of the electric wire (S308-3). Then, it is determined whether the value of the horizontal pixel number i exceeds the horizontal pixel number i_max of the target image (S308-4). If the horizontal pixel number i is equal to or less than i_max (S308-4; Yes), the vertical coordinate that is the electric wire crossing direction between the first edge pixel at the horizontal pixel number i and the point on the first contour line at the horizontal pixel number i. A difference in position is obtained and set as ΔH1 (S308-5). Similarly, the difference in the vertical coordinate position between the second edge pixel at the horizontal pixel number i and the point on the second contour line at the horizontal pixel number i is obtained, and is set as ΔH2 (S308-6). Then, it is determined whether the absolute value of ΔH1 and the absolute value of ΔH2 are equal to or greater than a predetermined value H_max (S308-7, S308-9). H_max is set to 10, for example. When ΔH1 is greater than or equal to H_max, 1 is added to the counter w (S308-8), and when ΔH2 is greater than or equal to H_max, 1 is added to the counter w ′ (S308-10). Then, it is determined whether the value of the counter w or the counter w ′ is equal to or greater than a predetermined value w_max (S308-11). For example, w_max is 20. If both of the counters w and w ′ do not satisfy w_max (S308-11; No), 1 is added to the horizontal pixel number i (S308-13), and the processing from S308-4 onward is repeated for the horizontal pixel number i. . On the other hand, if both ΔH1 and ΔH2 are less than H_max, 0 is set to the counters w and w ′ (S308-12), 1 is added to the horizontal pixel number i (S308-13), and the horizontal pixel number The process from S308-4 onward is repeated for i.

  When at least one of the counters w and w ′ is equal to or greater than w_max (S308-11; Yes), edge pixels that are separated by 10 pixels or more from the ideal contour line appear continuously, so that the target image Is recorded as a shape abnormality candidate image (S308-14). Then, 1 is added to the shape abnormality counter z (S308-15), and it is determined whether the counter z is equal to or greater than a predetermined value z_max (S308-16). z_max is set to 2, for example. If the shape abnormality counter z does not satisfy z_max (S308-16; No), the shape abnormality detection process for the target image is terminated. On the other hand, when the shape abnormality counter z is equal to or greater than z_max (S308-16; Yes), when z_max is set to 2, the shape abnormality candidate images are continuous for two or more frames. Record in the log (S308-17). Then, the shape abnormality counter z is reset to 0 (S308-18), and the shape abnormality detection process for the target image ends. On the other hand, when the processing is performed until the value of the horizontal pixel number i exceeds the horizontal pixel number i_max of the target image (S308-4; No), 20 edge pixels that are separated by 10 pixels or more from the ideal contour line are consecutive. In this case, the shape abnormality counter z is reset to 0 (S308-18), and the shape abnormality detection process for the target image ends.

  FIG. 4 shows an image obtained by photographing an electric wire that is causing a wire break. FIG. 2 shows the result of performing the above edge detection processing on the image of FIG. Further, the result of obtaining the ideal contour based on the remaining edge pixels excluding the low reliability pixel is identified by the above-described ideal contour estimation process from the edge pixels shown in FIG. As shown in FIG. FIG. 1 shows a result of displaying the edge pixel of FIG. 2 and the ideal outline of FIG. 3 in an overlapping manner. In the figure, reference numeral 40 indicates a first edge pixel, reference numeral 41 indicates a second edge pixel, reference numeral 42 indicates a first ideal contour line, and reference numeral 43 indicates a second ideal contour line. FIG. 1 shows a state in which 20 pixels or more of edge pixels that are 10 pixels or more away from the ideal contour line appear continuously, and the above-described shape abnormality detection processing accurately detects the wire that has been broken. I can confirm that I can do it.

  Next, FIG. 20 and FIG. 21 show flowcharts detailing the color abnormality detection processing. In this processing, 0 is set to the counters u and u '(S309-1, S309-2), and 1 is set to the pixel number i in the horizontal direction which is the longitudinal direction of the wire (S309-3). Then, it is determined whether the value of the horizontal pixel number i exceeds the horizontal pixel number i_max of the target image (S309-4). If the horizontal pixel number i is equal to or less than i_max (S309-4; Yes), the average value of the luminance values of the pixels located between the first contour line and the second contour line at the horizontal pixel number i is obtained, and the average The value is set as a representative value Val (i) (S309-5). Then, 1 is added to the horizontal pixel number i (S309-6), and the processing from S309-4 onward is repeated for the horizontal pixel number i. Thereby, the representative value Val (i) is obtained for each horizontal pixel number i. When the horizontal pixel number i exceeds i_max (S309-4; No), the upper limit threshold value Val ′ and the lower limit threshold value Val ″ are obtained by Expression 3 using the obtained representative values Val (1) to Val (i_max) ( S309-7).

  Then, the horizontal pixel number i is set to 1 again (S309-8), and it is determined whether the horizontal pixel number i is equal to or less than i_max (S309-9). If the horizontal pixel number i is not more than i_max, it is determined whether the representative value Val (i) corresponding to the pixel number i is not less than the upper limit value Val ′ or not more than the lower limit value Val ′ (S309-10, S309). -12). When the representative value Val (i) is greater than or equal to the upper limit value Val ′, 1 is added to the counter u (S309-11), and when the representative value Val (i) is less than or equal to the lower limit value Val ″, 1 is added to the counter u ′. (S309-13) Then, it is determined whether the value of the counter u or the counter u ′ is equal to or larger than a predetermined value u_max (S309-14), where u_max is set to 20, for example. If both do not satisfy u_max (S309-14; No), 1 is added to the horizontal pixel number i (S309-16), and the processing from S309-9 onward is repeated for the horizontal pixel number i. If (i) is not greater than or equal to the upper limit value Val ′ and not less than the lower limit value Val ″, 0 is set in the counters u and u ′ (S309-15), and 1 is added to the horizontal pixel number i (S309-16). ), S3 for the horizontal pixel number i 9-9 repeats the following processing.

  When at least one of the counters u and u ′ is greater than or equal to u_max (S309-14; Yes), a representative value Val (i) that is greater than or equal to the upper limit value Val ′ or less than or equal to the lower limit value Val ″ appears continuously for 20 pixels. Therefore, the target image is recorded as a color abnormality candidate image (S309-17), and 1 is added to the color abnormality counter q (S309-18), and the counter q is equal to or greater than a predetermined value q_max. In step S309-19, q_max is set to 2. If the color abnormality counter q is less than q_max (S309-19; No), the color abnormality detection process for the target image is terminated. When the color abnormality counter q is equal to or greater than q_max (S309-19; Yes), when q_max is set to 2, the color abnormality candidate images are continuous for two frames or more. (S309-20) Then, the color abnormality counter q is reset to 0 (S309-21), and the color abnormality detection process for the target image is terminated, while the value of the horizontal pixel number i is the target image. When the processing is performed until the number of horizontal pixels i_max exceeds (S309-9; No), the representative value Val (i) that is equal to or higher than the upper limit value Val ′ or lower than the lower limit value Val ″ may appear continuously for 20 pixels. In this case, the color abnormality counter q is reset to 0 (S309-21), and the color abnormality detection process for the target image ends.

  FIG. 6 shows an image in which an electric wire with an arc mark is photographed. FIG. 5 shows the result of the above color abnormality detection process performed on the image of FIG. In the figure, reference numeral 44 indicates a representative value in each horizontal pixel number, and reference numeral 55 indicates a lower limit threshold. FIG. 5 shows a state in which a representative value equal to or lower than the lower limit threshold appears continuously for 20 pixels or more, and it can be confirmed that an electric wire having an arc mark can be accurately detected by the color abnormality detection process.

  When the electric wire abnormality detection process is completed, an abnormal image recorded in the log is output (S4 in FIG. 11). In this process, the original image in which the target image could not be cut out and the target image determined to have an abnormality in the shape and color of the electric wire are displayed on the output device 14. The operator can visually determine whether there is a problem by looking at the image displayed on the output device 14. For example, in the color abnormality detection process of the present embodiment, there is a possibility of determining that there is an abnormality even when there is a stain or the like on the electric wire. Is trying to do. Note that still images may be arranged and displayed in an album form so that the operator can easily see them. Further, the actual video image may be displayed on the output device 14 based on the frame number of the image recorded in the log.

  According to the method for creating an image for electric wire inspection of the present invention, template matching is performed within a limited search area, and it is determined whether three or more search results satisfy a certain regularity. Therefore, it is possible to quickly and efficiently create a wire inspection image, which is a process of cutting out the region of the wire portion from the original image. And, according to the wire abnormality detection method and apparatus and program by image processing incorporating the method for creating the wire inspection image of the present invention, the operator does not need to visually check all the aerial images of the wire from the beginning, What is necessary is just to confirm a part of image and image | video which the electric wire abnormality detection apparatus 10 presents, and an operator's labor is reduced significantly. Furthermore, by connecting the video camera to the electric wire abnormality detection device 10 and processing the image obtained from the video camera in real time, it becomes possible to detect the abnormal portion of the electric wire at the site where the helicopter patrols, It becomes possible to reduce human resources and equipment resources related to work and to greatly reduce the overall cost.

  The above-described embodiment is an example of a preferred embodiment of the present invention, but is not limited thereto, and various modifications can be made without departing from the gist of the present invention. For example, in the above-described embodiment, each process is performed using a grayscale image as an original image, but each process may be performed using a color image as an original image. For example, when the original image is a grayscale image, the target image may be judged as a search result in which the ridge lines of the mountains are aligned in the phase of displacement in the clipping process of the target image. Such a failure can be avoided by performing template matching using the hue value in addition to the value.

  Further, the edge detection process is not necessarily limited to the example of the above-described embodiment, and a conventional edge detection method may be adopted. In order to further increase the processing speed, a part or all of the processing shown in FIGS. 11 to 22 may be implemented as hardware. In particular, since the edge detection processing part is a general-purpose method, it has good compatibility with a commercially available image processing board and can be easily implemented as hardware. By using hardware, processing at the video rate can also be expected.

  Each parameter such as a threshold, for example, a range Δj for searching for a peak in the edge detection process, a reference H_max of a distance between an edge pixel for determining a shape abnormality and an ideal outline, a threshold coefficient α for determining a color abnormality, a continuous abnormality The criteria w_max and u_max for candidate pixels and the criteria z and q for consecutive abnormal candidate images may be adjusted as appropriate according to the shooting conditions.

  Also, in the target image cut-out process, if there is no object similar to an electric wire in the original image, the search result sticks to the top or bottom of the screen, and it is determined that the phase of the search result displacement is aligned. There is a case. For this reason, when the search result sticks to the upper end or the lower end of the screen, a conditional branch that determines that the search has failed may be inserted. In this case, for example, a standard deviation in a certain period of displacement of each search result may be calculated, and it may be determined that sticking has occurred when the standard deviation is less than a certain value.

  In the above-described embodiment, the template can be added when the wire search fails in the target image cut-out process. However, when the wire search fails, the search area is enlarged and the search is performed again. You may do it. In the above embodiment, the number of search areas in the target image cut-out process is three, but the number of search areas may be further increased when there is a margin in the performance and processing time of the computer. . For example, the electric wire search may be performed at five or more locations, and the search results at three or more locations that satisfy the linearity may be selected from these search results.

  In the above-described embodiment, the wire abnormality detection process is performed on the target image cut out from the original image. However, in some cases, the wire abnormality detection process may be performed using the original image as it is as the target image. Moreover, you may use the target image obtained by the above-mentioned cutting-out process for the image processing for electric wire inspections other than a shape abnormality detection process and a color abnormality detection process.

It is a graph explaining the principle of the electric wire abnormality detection method in which the creation method of the image for electric wire inspection of the present invention is incorporated, and is displayed by superposing the edge pixel of the electric wire and the ideal outline. The vertical axis of the graph indicates the vertical pixel number, and the horizontal axis indicates the horizontal pixel number. It is the graph which displayed the edge pixel of the electric wire, a vertical axis shows a vertical pixel number, and a horizontal axis shows a horizontal pixel number. It is the graph which displayed the ideal outline of the electric wire, a vertical axis shows a vertical pixel number, and a horizontal axis shows a horizontal pixel number. The image which becomes the origin of Drawing 1-Drawing 3 is shown, and the image of the electric wire which has raise | disconnected the strand is shown. It is a graph explaining the principle of the electric wire abnormality detection method incorporating the preparation method of the image for electric wire inspection of this invention, and displays the representative value and threshold value of the color information value of an electric wire. The vertical axis of the graph indicates the luminance value, and the horizontal axis indicates the horizontal pixel number. FIG. 5 shows the original image of FIG. 5 and shows an image of an electric wire with arc marks. The image of the electric wire for demonstrating the principle of edge detection is shown. It is a graph which shows the result of having calculated | required the difference value of the luminance value between two adjacent pixels above the perpendicular direction shown by the arrow of FIG. The vertical axis of the graph indicates the vertical pixel number of the lower pixel for which the difference value is obtained, and the horizontal axis indicates the magnitude of the difference value. FIG. 9 is an enlarged view of a part of FIG. It is a schematic block diagram which shows one Embodiment of the electric wire abnormality detection apparatus in which the production method of the image for electric wire inspection of this invention is integrated. It is a flowchart which shows an example of the process performed by execution of the electric wire abnormality detection method and apparatus and program which incorporate the production method of the image for electric wire inspection of this invention. 12 is a flowchart showing an example of a method for creating an image for wire inspection according to the present invention, which is a detailed process of the target image cutout process in FIG. 11. It is a flowchart which shows an example of the process which detailed the electric wire abnormality detection process of FIG. It is a flowchart which shows an example of the process which detailed the edge detection process of FIG. It is a flowchart which shows an example of the process which detailed the search process of the edge pixel of FIG. It is a flowchart which shows an example of the process which detailed the ideal outline estimation process of FIG. FIG. 17 is a flowchart illustrating an example of a process in which the removal candidate pixel search process of FIG. It is a flowchart which shows an example of the process which detailed the shape abnormality detection process of FIG. 13, and shows the first half part of a process. It is a flowchart which shows the second half part of FIG. It is a flowchart which shows an example of the process which detailed the color abnormality detection process of FIG. 13, and shows the first half part of a process. It is a flowchart which shows the second half part of FIG. It is a flowchart which shows an example of the process which detailed the pre-process of FIG. It is a conceptual diagram explaining the principle which searches an electric wire in the preparation method of the image for electric wire inspection of this invention. It is a conceptual diagram explaining the principle which searches an electric wire in the preparation method of the image for electric wire inspection of this invention. It is a figure which shows the example of the screen which the electric wire abnormality detection apparatus in which the preparation method of the image for an electric wire inspection of this invention is integrated displays. It is a figure which shows the other example of the screen which the electric wire abnormality detection apparatus in which the preparation method of the image for an electric wire inspection of this invention is integrated is displayed. It is a figure which shows the other example of the screen which the electric wire abnormality detection apparatus in which the preparation method of the image for an electric wire inspection of this invention is integrated is displayed. It is a figure which shows the other example of the screen which the electric wire abnormality detection apparatus in which the preparation method of the image for an electric wire inspection of this invention is integrated is displayed. (A) is a conceptual diagram explaining the principle which extracts an electric wire area | region by a prior art, (B) and (C) are conceptual diagrams which show the problem at the time of using the same method.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Original image 2 Template 3 Search area 4 Electric wire 10 Electric wire abnormality detection apparatus 29 Search result

Claims (3)

  One of the horizontal direction and the vertical direction of the original image in which the electric wire is photographed is defined as a direction along the direction of the electric wire in the original image as the electric wire longitudinal direction, and the search area in which the horizontal range and the vertical range are predetermined Distributing three or more locations in the longitudinal direction of the electric wire of the original image, giving an image of the electric wire to be searched in advance as a template, searching for a region most similar to the template in each search region, and for each search region It is determined whether or not the search result satisfies a predetermined regularity, and an image including at least the search result satisfying the regularity is cut out from the original image and used as an electric wire inspection image How to create an image.   The new template is added when the search result does not satisfy the predetermined regularity or when the similarity of the search result does not satisfy a predetermined criterion. How to create an image for wire inspection.   The method for creating an image for electric wire inspection according to claim 1 or 2, wherein the template having the highest similarity of the search results is selected from the plurality of templates.