Pipe inspection device, pipe inspection method, and program

Abstract

To enable accurate detection of abnormalities within pipes. [Solution] The pipe inspection device of the embodiment includes an abnormality detection unit that, depending on the position or movement of the camera that photographs the inside of the pipe, selects one or more images from the video data of the detection target inside the pipe captured by the camera, determines the presence or absence and level of abnormality of the detection target inside the pipe using the selected images, and generates an abnormality detection result for each detection target based on the determined result.

Description

Technical Field

[0001] Embodiments of the present invention relate to an in-pipe inspection device, an in-pipe inspection method, and a program.

Background Art

[0002] When deposits or rust occur inside pipes such as water pipes due to long-term use, foreign matter deposition and adhesion, and deterioration and peeling of the inner surface of the pipe occur, various abnormalities occur inside the pipe.

[0003] In order to inspect the inside of a pipe such as a water pipe, for example, the inside of the pipe is photographed in advance with a camera, and based on the video of the camera after the photographing, it is diagnosed whether there are any abnormalities in the pipe.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] When visually diagnosing the inside of a pipe such as a water pipe, in addition to the diagnosis taking a very long time, there is a problem that the diagnosis results vary depending on the person making the judgment, and the diagnosis results also vary among individuals.

[0006] On the other hand, there is also a technique for facilitating diagnosis by using the video of the camera as described above. However, due to the flow of water and floating objects in the pipe and the blur caused by the movement of the camera, the detection targets such as deposits and rust may not be correctly reflected in the video of the camera. Therefore, it is difficult to correctly detect abnormalities inside the pipe, and there is a problem that the diagnosis results are likely to vary.

[0007] The problem that this invention aims to solve is to provide a pipe inspection device, a pipe inspection method, and a program that enable accurate detection of abnormalities inside a pipe. [Means for solving the problem]

[0008] The pipe inspection device of the embodiment includes an abnormality detection unit that, depending on the position or movement of the camera that photographs the inside of the pipe, selects one or more images from the video data of the object to be detected inside the pipe that is photographed by the camera, determines the presence or absence and level of abnormality of the object to be detected inside the pipe using the selected images, and generates an abnormality detection result for each object to be detected based on the determined result. [Brief explanation of the drawing]

[0009] [Figure 1] A diagram showing an example of the overall system configuration common to the first to third embodiments. [Figure 2] A diagram showing an example (part 1) of the configuration of the pipe inspection device 1 according to the first embodiment. [Figure 3] A diagram showing an example (part 2) of the configuration of the pipe inspection device 1 according to the first embodiment. [Figure 4] A diagram showing an example (part 3) of the configuration of the pipe inspection device 1 according to the first embodiment. [Figure 5] This diagram shows an example of a camera 4 being inserted into a water pipe 3. [Figure 6] A diagram showing an example of an image included in the video actually captured by camera 4. [Figure 7] This figure shows an example of an image displaying information indicating the cable length (i.e., the location information of camera 4). [Figure 8] A diagram illustrating a method for determining the degree of pipe blockage due to rust from images. [Figure 9] A diagram showing an example of a table used to determine abnormal levels of rust. [Figure 10] A diagram showing an example of a table used to determine abnormal levels in sediments. [Figure 11] A diagram illustrating a method for determining the size of abnormal areas containing foreign objects or degradation from an image. [Figure 12] A graph for explaining a method of determining, for each section composed of a plurality of frames, the one with the highest abnormality level as the abnormality level of that section. [Figure 13] A graph for explaining a method of performing level determination only for frames determined that camera 4 is not moving. [Figure 14] A graph for explaining a method of performing level determination for the frame with the least movement amount of camera 4 for each section. [Figure 15] A table showing a setting example defining the movement and moving direction of camera 4 that can be applied for each detection target. [Figure 16] A table showing a setting example defining a determination unit that can be applied for each detection target. [Figure 17] A diagram showing an example of a screen for displaying the detected and determined results on a display. [Figure 18] A flowchart showing an example of the operation of the in-pipe inspection device 1 according to the first embodiment. [Figure 19] A diagram showing an example of the configuration of the in-pipe inspection device 1 according to the second embodiment. [Figure 20] A flowchart showing an example of the operation of the in-pipe inspection device 1 according to the second embodiment. [Figure 21] A diagram showing an example of the configuration of the in-pipe inspection device 1 according to the third embodiment. [Figure 22] A flowchart showing an example of the operation related to the setting unit 12 of the in-pipe inspection device 1 according to the third embodiment.

Mode for Carrying Out the Invention

[0010] Hereinafter, embodiments will be described with reference to the drawings.

[0011] <The First Embodiment> First, the first embodiment will be described.

[0012] (System Configuration) Figure 1 shows an example of the overall system configuration including the pipe inspection device according to the first embodiment.

[0013] In this embodiment, as an example, we will describe the inspection of the inside of a water pipe (water supply pipe), particularly an inspection conducted while the pipe is full of water as a non-stop water supply inspection. However, this embodiment is not limited to this example and can also be applied to the inspection of the inside of various pipes, such as sewer pipes and other pipes.

[0014] The pipe inspection device 1 according to this embodiment performs an inspection of the inside of a water pipe 3 based on images from a camera 4 that photographs the inside of the pipe, as shown in Figure 1, at a site 2. The images captured by the camera 4 are recorded on a recording medium 5 as video data (file) consisting of multiple consecutive images (frames). Attribute information indicating the attributes associated with the camera is also recorded on this recording medium 5. The pipe inspection device 1 takes in the video data and attribute information recorded on the recording medium 5 and uses them to perform an inspection of the inside of the pipe.

[0015] Note that the system configuration shown in Figure 1 is just one example and is not limited to this example. For example, the pipe inspection device 1 may include a camera 4 for taking images, or it may include a recording medium 5 that stores video data and attribute information.

[0016] (Configuration of pipe inspection device 1) Figures 2 to 4 show examples of the configuration of the pipe inspection device 1 according to the first embodiment. Figure 2 shows one example of the configuration of the pipe inspection device 1 (part 1). Figure 3 shows another example of the configuration of the pipe inspection device 1 (part 2). Figure 4 shows yet another example of the configuration of the pipe inspection device 1 (part 3). Some or all of the functions of the individual elements shown in each of Figures 2 to 4 can be realized as a program to be executed by a computer.

[0017] The pipe inspection device 1 shown in Figure 2 includes a video input unit 10, an attribute information input unit 11, a camera position / motion detection unit 20, an anomaly detection unit 30, an integrated determination unit 40, and an output processing unit 50.

[0018] The video input unit 10 receives video data from the camera 4 that films the inside of the water pipe 3.

[0019] The attribute information input unit 11 inputs attribute information of the camera 4. This attribute information is supplied to the camera position / motion detection unit 20, the anomaly detection unit 30, and the integrated determination unit 40, and is used as needed in the processing of each unit.

[0020] The camera position / motion detection unit 20 uses video data input by the video input unit 10 and attribute information input by the attribute information input unit 11 to detect the position or movement of the camera 4 inside the pipe.

[0021] The anomaly detection unit 30 selects one or more images from the video data of the detection target captured by the camera 4 according to the position or movement of the camera 4 detected by the camera position / motion detection unit 20, uses the selected images to determine whether there is an anomaly in each detection target inside the pipe and the level of the anomaly, and generates an anomaly detection result for each detection target based on the determined result. The objects to be detected here refer to any of the following: (1) rust, (2) sediment, (3) partial adhesion of foreign matter or deterioration of corrosion protection on the pipe surface, (4) film-like dirt adhering to the entire inside of the pipe, and (5) suspended matter.

[0022] The integrated determination unit 40 integrates the abnormality detection results generated for each detection target by the abnormality detection unit 30 to generate an integrated determination result indicating the level of abnormality within the pipe.

[0023] The output processing unit 50 outputs information including the anomaly detection result generated for each detection target by the anomaly detection unit 30 and the integrated judgment result generated by the integrated judgment unit 40 to a display device or storage device.

[0024] On the other hand, the pipe inspection device 1 shown in Figures 3 and 4 also includes the video input unit 10, attribute information input unit 11, camera position / motion detection unit 20, anomaly detection unit 30, integrated determination unit 40, and output processing unit 50 mentioned above, but the arrangement of some of these elements differs from that of the pipe inspection device 1 shown in Figure 1.

[0025] In this embodiment, as shown in Figure 2, the video input unit 10 does not include a camera 4 that actually films the inside of the pipe, but is basically configured to input video data from the outside. In this case, the video data of the image captured by the camera 4 is input to the pipe inspection device 1 via the recording medium 5, and the pipe inspection device 1 uses this data to detect abnormalities.

[0026] Alternatively, the camera 4 may be included within the pipe inspection device 1. In that case, as shown in Figure 3, the camera 4 itself will be included in the video input unit 10. Furthermore, sensors such as a gyro sensor or an accelerometer may be used in the video input unit 10 to acquire position information or motion information of the camera 4. In that case, as shown in Figure 4, the gyro sensor or accelerometer may be mounted in the video input unit 10 as a camera position / motion detection unit 20, for example, on the camera 4 or an accessory to the camera 4.

[0027] (Details of individual components) The individual components that make up the pipe inspection device 1 are described in detail below.

[0028] • Video input section 10 The video input unit 10 can accommodate both cases: with or without hardware such as a camera 4 for capturing video. In the example configuration shown in Figure 2, there is a separate device for recording video data from camera 4 that captures images inside the pipe, so all that is needed is a means to read the video data recorded there via the recording medium 5. For example, the video data file can be saved to a recording medium 5 such as an SD card, and the video data can be input by reading that file with a reader that reads the recording medium 5.

[0029] On the other hand, in the configuration examples shown in Figures 3 and 4, the pipe inspection device 1 includes a camera 4. This camera 4 may be implemented as a camera head equipped with, for example, an ITV camera. The ITV camera can digitize the optical information obtained through the camera lens using an A / D converter and output it as video data.

[0030] Figure 5 shows an example of a camera 4 inserted into a water pipe 3. Here, as shown in Figure 5, the camera 4 moves inside the water pipe 3 at site 2, so it may be waterproofed as a whole. The camera 4 may also be equipped with a function to illuminate the inside of the pipe and a rotation mechanism to keep the camera 4 always up and down and horizontal. A long cable 6 is connected to the camera 4, and the camera 4 is inserted into the water pipe 3 by pushing the cable 6 from a fire hydrant 7 or the like, and the inside of the water pipe 3 is photographed. For photography, the camera 4 is inserted on both the upstream and downstream sides. To achieve this, a jig placed from the fire hydrant 7 to the water pipe 3 allows control of the direction in which the camera 4 is inserted.

[0031] Furthermore, camera 4 may be equipped with sensors such as the aforementioned gyro sensor and accelerometer to obtain information about the orientation and movement of camera 4, and these sensors may be used to detect the orientation and movement of camera 4. In this case, the type of sensor does not matter as long as it can measure the movement and position of camera 4. The information about the orientation and movement of camera 4 can be used by the camera position / motion detection unit 20 when detecting the movement and position of camera 4.

[0032] The video from camera 4 is recorded on recording medium 5 via cable 6 by a recording device (not shown) located outside the fire hydrant 7. Any video recording method is acceptable; there are no restrictions on the method. For example, it could be analog NTSC, digital uncompressed, or digitally encoded.

[0033] Figure 6 shows an example of an image included in the video actually captured by camera 4. Normally, there is almost no foreign matter inside the cylindrical pipe, but if there is an abnormality inside the pipe, the video may show conditions such as foreign matter accumulating and adhering inside the pipe as shown in Figure 6(a), deterioration and peeling of the inner surface of the pipe, or foreign matter floating in the water and flowing inside the pipe as suspended material as shown in Figure 6(b).

[0034] Attribute information input section 11 Next, the attribute information input unit 11 will be described. In the example configuration shown in Figure 2, similar to the video input unit 10, there is a separate device for recording video data from camera 4 that photographs the inside of the pipe. Therefore, it is sufficient to have a means to read the attribute information related to the video from camera 4 along with the video data recorded there via the recording medium 5. The attribute information may be input, for example, from camera 4 or from a keyboard or mouse input device. Alternatively, information obtained from sensors inside camera 4 may be used as attribute information. The attribute information generated in any of these forms can be saved as a data file on a recording medium such as an SD card, and the attribute information can be input by reading that file with a reader that reads the medium.

[0035] On the other hand, in the configuration examples shown in Figures 3 and 4, even if the attribute information input unit 11 is not present, it is still possible to perform inspections inside the pipe if the video input unit 10 is present. However, in this embodiment, an attribute information input unit 11 is provided to input attribute information related to the camera in order to enable more accurate diagnosis of the inside of the pipe.

[0036] The attribute information entered in the attribute information input unit 11 includes, for example, information indicating the length of the cable 6 when the camera 4 is inserted into the water pipe 3, which represents the position information of the camera 4 inside the pipe. Figure 7 shows an example of an image displaying information indicating the cable length (i.e., the position information of the camera 4).

[0037] Information indicating the cable length can be recorded by superimposing it onto a predetermined location in the video data, making it usable in the camera position / motion detection unit 20 described later. If the video is recorded in digital format, the information indicating the cable length may be recorded in synchronization with the video data rather than superimposing it onto the video data.

[0038] The attribute information input unit 11 also allows for the input of other attribute information. For example, attribute information such as the location of the water pipe (place name, latitude and longitude, address, etc.), the type of water pipe, the diameter of the water pipe, the year the water pipe was laid, the date and time of the last inspection and the diagnostic results at that time, as well as the camera model and lens parameters, can be recorded in association with the recording medium used to record the video, making it possible to input this information into the attribute information input unit 11.

[0039] The attribute information input unit 11, which inputs this attribute information, may be installed in the same camera unit as the camera 4 in the video input unit 10 shown in Figures 3 and 4, or it may be installed in a separate unit different from the said camera unit, or it may be installed in a separate unit connected to the end of the cable 6 to which the camera 4 is connected.

[0040] Furthermore, if the camera unit is equipped with sensors such as a gyro sensor or an accelerometer, information indicating the position and movement of the camera 4 may be recorded as attribute information based on these sensors, and this attribute information may be used by the camera position / motion detection unit 20 described later.

[0041] • Camera position and motion detection unit 20 The camera position and motion detection unit 20 detects the current position of the camera 4 inside the pipe, as well as whether or not the camera 4 is moving. In the configuration examples shown in Figures 2 and 3, it is assumed that video data and attribute information are input to the camera position and motion detection unit 20 via the video input unit 10 and the attribute information input unit 11, and the camera position and motion detection unit 20 uses this video data and attribute information to detect the position and motion of the camera 4. In the configuration example shown in Figure 4, for example, the camera position and motion detection unit 20 is mounted on the camera 4 hardware itself (or on an accessory part of the camera 4). In this case, for example, sensors such as a gyro sensor or an accelerometer may be mounted in the camera 4 hardware, and the movement of the camera 4 may be detected using the sensor information, or the same detection may be performed by software processing by the CPU (Central Processing Unit).

[0042] The following describes a method for detecting the position and movement of camera 4 using video data.

[0043] (1) Detection of camera position First, let's explain how to detect the position of camera 4. The simplest method for detecting the position of camera 4 is to read the cable length information (i.e., the position information of camera 4) from the attribute information input by the attribute information input unit 11. If the video data is in digital format, this can be achieved simply by embedding the cable length information corresponding to each frame in the video data and reading that value.

[0044] If information indicating cable length is superimposed and recorded within the video data, this information can be read using optical character recognition (OCR). In this case, recognition is possible by finding areas where characters are arranged horizontally within each image, without specifying a character display area within camera 4. For example, 0-9 and the decimal point "·" can be registered as templates, and the image can be searched using template matching. By utilizing the fact that there are areas where characters that match any of these characters are arranged horizontally, character areas can be found, and at the same time, the numerical value can be read from the template information of the most matched number.

[0045] In the example image in Figure 7, the templates "1", "6", "·", and "4" have high matching scores in the upper right region of the screen, and since their positions along the Y-axis are almost the same and they are aligned left and right, it is possible to read the value as 16.4. For character recognition in this case, techniques such as those described in "O. Hori, "A video text extraction method for character recognition", Proceedings of the Fifth International Conference on Document Analysis and Recognition. ICDAR '99 (Cat. No.PR00318)" (hereinafter, "Reference 1") may be used.

[0046] Alternatively, information specifying the model of camera 4 may be included in the attribute information in advance, so that an area is set up to display information indicating the cable length according to the model.

[0047] In the case of a cable-pushing type camera, the cable 6 bends when pushed in and stretches when pulled back, so the length of the cable 6 tends to be displayed differently even at the same location. Therefore, it is possible to improve the accuracy of the position by comparing the images taken during the push-in and pull-back phases and correcting the cable length value during the pull-back phase to the official value for similar images. In this case, a method of comparing similar images may be used, or a more accurate position may be obtained by using the 3D information reconstruction technique using a monocular camera as shown in "Seki, "3D Shape Reconstruction and Self-Localization Estimation (SLAM) and High-Density 3D Shape Reconstruction from Moving Camera Images", IPSJ Research Report Computer Vision and Image Media (CVIM), 2014-CVIM-190, 2014" (hereinafter, "Reference 2").

[0048] Furthermore, if the type of water pipe 3 is known, the location of each connection point between the pipes can be determined. Therefore, if the location of each connection point appears in the video as concentric circles with a certain level of brightness change (edges), the position of camera 4 can be determined by counting the number of such concentric circles with edges detected.

[0049] (2) Detection of camera movement Next, we will explain how to detect the movement of camera 4. A simple method is to determine the optical flow across the entire image, and if, when you find the vanishing point from the direction of the flow, you can see that many flow vectors originate or end at the vanishing point, then you can tell that camera 4 is moving.

[0050] However, when photographing inside the water pipe 3, if there are floating objects inside the water pipe 3, these objects will flow towards the back of the pipe or from the back to the front, which can cause a vanishing point vector to be generated even if the camera 4 is stationary. Therefore, it may be possible to determine whether or not the camera 4 is moving by detecting the blurring (motion blur) that occurs when the image changes within the exposure time when the camera 4 moves, causing the entire image to become blurred. For example, the edge intensity in the image can be calculated and these values ​​accumulated. When the accumulated edge intensity exceeds a certain threshold, it is determined that the camera is stationary, and when it falls below a certain threshold, it is blurred and therefore moving.

[0051] Furthermore, to reduce the influence of floating objects, the area to be accumulated may be limited, such as limiting the pixels to the peripheral part of the tube (excluding the central part). Also, edge strength may be determined not by the edges with adjacent pixels, but by applying an FFT to the entire image and accumulating the power of high-frequency components. For example, by pre-determining the frequencies generated by floating objects and excluding those frequencies, the presence or absence of motion can be detected with high accuracy.

[0052] In addition, although this method is limited to video stored in the NTSC (National Television Standards Committee) analog signal format, in this case, images are recorded alternately in the vertical direction, one row at a time, using an interlaced method. Therefore, if there is movement in camera 4, a striped noise called combing noise occurs. This is because the shooting time differs between odd-numbered and even-numbered rows, causing a change in brightness wherever a change occurs, resulting in a striped change in brightness. Therefore, for this striped area, when Y(x,y) is the brightness value Y at coordinates x,y, a predetermined coefficient α may be used to determine if there are a predetermined number of pixels that satisfy the following equation, and thus determine that camera 4 is moving.

[0053] |Y(x,y)-Y(x,y-1)| > |(x,y)-y(x,y-2)|+α According to the above formula, under normal circumstances, brightness changes smoothly in space, so the brightness difference between two lines above is usually greater than the brightness difference between one line above. However, when combing noise occurs, the brightness difference between one line above and two lines above becomes greater than the brightness difference between two lines above. This method utilizes this phenomenon.

[0054] In the configuration examples shown in Figures 2 and 3, if information from sensors such as the gyro sensor and accelerometer mounted on the camera 4 is input from the attribute information input unit 11 to the camera position / motion detection unit 20, then the movement of the camera 4 can be easily detected by reading that information. In this case, it becomes possible to accurately detect the movement of the camera 4 regardless of the underwater conditions.

[0055] In the example configuration shown in Figure 4, the hardware itself that detects motion using sensors such as a gyroscope and an accelerometer is included within the camera position / motion detection unit 20. In this case as well, motion detection is possible by utilizing the sensor information.

[0056] As a result, it becomes possible to detect the current position of camera 4 and whether or not camera 4 is moving.

[0057] • Anomaly detection unit 30 The anomaly detection unit 30 detects the presence or absence of an anomaly in the water pipe 3 for each detection target from the video data and determines the level of the anomaly. In this embodiment, the detection targets are assumed to be (1) rust, (2) sediment, (3) partial adhesion of foreign matter and deterioration of corrosion protection on the pipe surface, (4) film-like dirt adhering to the entire inside of the pipe, and (5) floating matter. The methods for detecting each detection target will be described below.

[0058] (1) Chorus Rust occurs when the inner surface of a pipe corrodes and rusts, often appearing as a reddish-brown color. Since the inner surface of a pipe does not normally turn reddish-brown, rust can be detected by finding the reddish-brown areas through color detection. Furthermore, in recent years, it has become possible to detect objects learned at the pixel level using deep learning. The rust area can also be determined at the pixel level using the techniques described in "A New Image Region Segmentation Method Using Deep Learning," Toshiba Review Vol. 74, No. 5, September 2019 (hereinafter, "Reference 3") or "Fast-SCNN: Fast Semantic Segmentation Network," https: / / arxiv.org / pdf / 1902.04502.pdf (hereinafter, "Reference 4").

[0059] The level of rust abnormality may be determined from the video data based on the shape and blockage rate of the pipe. The shape of the pipe may be determined by creating multiple circles using isodensity lines to obtain the brightness gradient of the pipe, or by applying a circular extraction filter. For example, as shown in the example image in Figure 8, the brightness of the inner surface of the cylindrical pipe decreases (becomes darker) the further it is from the camera 4, so this can be used to create multiple circles using isodensity lines based on multiple brightness criteria. If rust is present at the connection points of the water pipe 3, the rust adheres along that circle, and the rust adheres along one of the concentric circles. By creating a circle along the inside of the rust (shown by the dashed line in Figure 8) relative to the outermost circle where the rust adheres, the blockage rate can be determined from the ratio of this circle to the circumscribing circle.

[0060] In determining the abnormal level of rust, for each abnormal level, a predetermined threshold may be set for, for example, the occlusion rate, and the higher the calculated occlusion rate, the higher the abnormal level may be judged. Figure 9 shows an example of a table used to determine the abnormal level of rust. As shown in the table in Figure 9, if the relationship between abnormal levels S, A, B, C, and D and the occlusion rate (a specific range based on a threshold) is set in advance, the abnormal level can be determined from the calculated occlusion rate value.

[0061] In addition to using the occlusion rate as an indicator, other values ​​may also be used as indicators, such as the total area (number of pixels) of rust on the screen, the size of rust clumps determined by labeling, or the area of ​​rust within a predetermined region to be judged.

[0062] (2) Sediment In sediment detection, foreign matter accumulated at the bottom of the pipe is detected as sediment. In this case, detection may be performed using the techniques described in references 3 and 4 above. On the other hand, when camera 4 retracts, the movement of accumulated foreign matter as it is stirred up by camera 4 is more likely to come into the camera 4's field of view. For this stirred foreign matter, the amount of motion vectors within the screen may be determined using optical flow, or the amount of edges may be detected, and if these values ​​exceed a predetermined standard, it may be determined that sediment is present.

[0063] For sediment, the abnormality level can be determined according to predetermined criteria by calculating the area and volume. The area and volume can be calculated from the amount of pixel values ​​in the image. However, since the calculated area and volume values ​​need to be corrected according to the distance from camera 4, it is desirable to determine the abnormality level by correcting them by multiplying each pixel coordinate by a coefficient. Alternatively, the abnormality level can be determined in proportion to the volume by accumulating the calculated values ​​along the time axis. Furthermore, since the sediment is basically located in the lower part of the pipe, the target area can be limited to the lower half of the pipe, and the corrected area and volume can be calculated within that range.

[0064] In determining the abnormality level of sediment, for each abnormality level, a predetermined threshold may be set for, for example, the corrected area or volume, and the abnormality level may be judged as higher if the calculated corrected area or volume is larger. Figure 10 shows an example of a table used to determine the abnormality level of the sediment. As shown in the table in Figure 10, if the relationship between abnormality levels S, A, B, C, and D and the corrected area or volume (specific ranges based on thresholds) is set in advance, the abnormality level can be determined from the calculated corrected area or volume values.

[0065] In addition to methods using area and volume as indicators, other indicators may also be used, such as the size of the area where the sediment exists (circumscribed rectangle) or the number of clumps determined by labeling.

[0066] (3) Partial adhesion of foreign matter and deterioration of corrosion protection on the pipe surface A normal water pipe has a uniform color across its entire surface, so there are no edge-like features. On the other hand, when abnormalities such as partial adhesion of foreign matter or deterioration of the corrosion protection on the pipe surface (hereinafter referred to as foreign matter adhesion or internal deterioration) occur, edge-like features tend to appear in the image. Therefore, by utilizing the fact that edges are detected above a certain level through edge processing, it is possible to detect foreign matter adhesion or internal deterioration and determine the level of abnormality.

[0067] Figure 11 shows an example of a method for detecting foreign matter adhesion and internal deterioration from an image. As shown in the example image in Figure 11, if a predetermined number of pixels with edge intensity above a predetermined value are detected, the bounding rectangle of the connected range can be determined, thereby determining the bounding rectangle of the abnormal region containing foreign matter or deterioration, and the size of the abnormality (which can be expressed as the area of ​​the rectangle, etc.) can be confirmed.

[0068] In determining the abnormal level of foreign matter adhesion and internal deterioration, a predetermined threshold can be set for each abnormal level, for example, based on the area of ​​the circumscribed rectangle. The larger the area of ​​the calculated circumscribed rectangle, the higher the abnormal level can be determined. For example, similar to the table in Figure 10 mentioned above, if the relationship between abnormal levels S, A, B, C, and D and the area of ​​the circumscribed rectangle (specific range based on the threshold) is set in advance, the abnormal level can be determined from the value of the calculated area of ​​the circumscribed rectangle.

[0069] In addition to using the area of ​​the circumscribing rectangle as an indicator, other indicators may also be used, such as the number of anomalous regions or the size of the distribution range (the size of the circumscribing rectangle of the entire cluster of existing anomalous regions).

[0070] (4) Dirt adhering in a film-like manner to the entire inside of the pipe It is difficult to detect film-like deposits that adhere to the entire inside of a pipe (hereinafter referred to as "film-like deposits") using the method described in (3) above. Therefore, by calculating the average brightness and color of the inside surface of the pipe and determining the difference from the normal state, it is possible to detect film-like deposits and determine the level of abnormality. Since deposits often cause the surface to darken, abnormalities caused by film-like deposits can be detected by determining, for example, that an abnormality occurs when the average brightness value is low (when it is dark).

[0071] Furthermore, if the dirt forms a film, inserting camera 4 may scrape off the surface dirt, leaving linear marks. Therefore, detecting the presence of linear marks on the surface is also effective. In this case, it is possible to detect line segments using Hough transform after edge extraction and binarization processing, and an anomaly can be determined when a line of a predetermined length and width is found. If the locations where linear marks are made are limited in advance due to the position and shape of camera 4, accuracy can be improved by limiting the location of line segment detection.

[0072] The level of abnormality in film-like deposits can be determined based on the magnitude of the change in brightness compared to a normal state, or it can be determined using indicators such as the size of the lines traced by camera 4, for example, the thickness, length, area, or cumulative value of the edge intensity. In all cases, the value will be small in a normal state and will increase when there is an abnormality, thus determining the level of the abnormality.

[0073] In determining the abnormality level of film-like deposits, a predetermined threshold can be set for each abnormality level, for example, for line thickness or length. The greater the calculated line thickness or length, the higher the abnormality level can be determined. For example, similar to the table in Figure 10 mentioned above, if the relationship between abnormality levels S, A, B, C, and D and line thickness or length (specific ranges based on thresholds) is set in advance, the abnormality level can be determined from the calculated line thickness or length values.

[0074] In addition to using line thickness and length as indicators, other indicators such as area and cumulative edge strength may also be used.

[0075] (5) Floating objects Suspended particles are foreign objects contained in the water flowing through the pipe, and appear as shown in the example image in Figure 6(b). Since suspended particles are in motion, it is possible to detect their amount by determining the amount of optical flow, as explained in the motion detection of camera 4. Furthermore, as the amount of suspended particles increases, there are many short linear noises, so detection is also possible by integrating the amount of pixels with edges above a certain intensity or by detecting the amount of line segments.

[0076] The level of abnormality is determined by criteria based on the amount of suspended matter. A state with no suspended matter is considered normal (level S), and the level of abnormality increases (worsens) as the amount of suspended matter increases. Specifically, optical flow can be calculated across the entire screen, and indicators such as the integral value of the flow length or the integral value of the edge intensity of the edge image can be used to determine the level of abnormality.

[0077] In determining the abnormality level of floating objects, a predetermined threshold can be set for each abnormality level, for example, based on the integral value of the flow length. A higher integral value of the flow length indicates a higher abnormality level. For example, similar to the table in Figure 10 mentioned earlier, if the relationship between abnormality levels S, A, B, C, and D and the integral value of the flow length (a specific range determined by the threshold) is set in advance, the abnormality level can be determined from the calculated integral value of the flow length. Furthermore, in the case of floating objects, accuracy can be improved by averaging or maximizing values ​​across multiple frames.

[0078] Alternatively, a method can be used in which a large number of images of a normal water pipe 3 are prepared and used for training, allowing the model to learn the range of fluctuations that occur within the normal range, and detecting anomalies by determining whether the input image deviates from the distribution of normal conditions. Specifically, the method using an AutoEncoder as described in "Sensing Technology Required in Manufacturing Sites Advancing IoT," Toshiba Review, Vol. 73, No. 1, January 2018, https: / / www.global.toshiba / content / dam / toshiba / migration / corp / techReviewAssets / tech / review / 2018 / 01 / 73_01pdf / a05.pdf (hereinafter, "Reference 5") may be used.

[0079] Alternatively, in determining the abnormality level, the range of each abnormality level may be defined using the magnitude of the integral value of the heatmap of the difference value from the normal state as an indicator, and the abnormality level may be determined based on this.

[0080] The various functions provided by the anomaly detection unit 30 are shown below.

[0081] (A function to select information and perform level determination in a way that reduces the influence of camera movement) When motion blur occurs due to the movement of camera 4, it becomes difficult to determine the abnormal level. Therefore, it is conceivable to accurately determine the abnormal level by selecting the image (frame) or level determination result to be used for determining the abnormal level in a way that reduces the influence of camera 4's movement. To achieve this, it is effective to equip the abnormality detection unit 30 with the following functions. (a) A function that selects the image with the highest abnormality level for each predetermined number of consecutive images (frames) and determines the abnormality level. (b) A function that uses the abnormality level obtained from the image (frame) when camera 4 is not moving to determine the abnormality level. (c) A function that determines the abnormality level by selecting the abnormality level obtained from the image with the least movement of the camera 4 for each predetermined number of consecutive images (frames).

[0082] Specifically, the function described in (a) above first determines the anomaly level of each frame regardless of the movement of camera 4, and then adopts the highest anomaly level among a predetermined number of frames as the determination result. Generally, the greater the movement of camera 4, the more the image blurs, making it difficult to detect anomaly areas and leading to a lower determination of the anomaly level. Therefore, the highest anomaly level (worst-case value) is sought. For example, the anomaly level is determined for each frame in the video data. Then, as shown in the example graph in Figure 12, the video data is divided into predetermined number of frames to set up intervals consisting of multiple frames (intervals separated by dashed lines in Figure 12), and for each interval, the highest anomaly level (the position marked with a circle on the graph in Figure 12) is determined as the anomaly level for that interval. The intervals to be divided may be at fixed intervals, or the length of the interval (number of frames) may be changed depending on the location.

[0083] Specifically, the function described in (b) above performs level determination only for frames in which the camera position / motion detection unit 20 determines that the camera 4 is not moving. Figure 13 shows an example of a graph in which the amount of camera movement in each frame of the video data is represented by edge strength (the number of pixels with edge strength above a predetermined level, or the value obtained by calculating the gradient of the edge strength for each pixel and accumulating these values). When the amount of camera movement is represented by the amount of edges, blur (shake) decreases and the amount of edges increases when the camera 4 is not moving. Therefore, when the edge strength exceeds a predetermined threshold, it is determined that the camera 4 is moving, and level determination is performed within that range. Furthermore, since the level determination result may be affected by the state of the pipe, the amount of floating objects, etc., a different threshold may be applied to each section rather than applying a fixed threshold to all frames.

[0084] Specifically, the function described in (c) above utilizes the level determination result for the frame with the least amount of movement of camera 4 in each section. The level determination can be performed based on the edge strength obtained by the camera position / motion detection unit 20, similar to the function described in (b) above. Also, similar to the function described in (a) above, the section is divided into predetermined numbers of frames, and as shown in the example graph in Figure 14, for each section, the frame with the greatest edge strength in the progression of edge magnitude is adopted as the "frame with the least amount of movement of camera 4," and the level determination is performed for that frame. The sections to be divided may be at fixed intervals, or the length of the section (number of frames) may be changed depending on the location.

[0085] By combining one or more of the above three functions, it is possible to detect anomalies and determine the level of anomalies with high accuracy, while being less affected by image changes caused by the movement of camera 4. In particular, since floating objects move even when camera 4 is stationary, it is possible to easily detect anomalies and determine the level of anomalies.

[0086] (A function that changes the detection target according to the direction of camera movement) In this embodiment, a water pipe through which tap water flows is assumed, and since water is flowing even when the camera 4 is stationary, movement may be included in the captured video. For example, if there are foreign objects (floating objects) mixed in with the water, movement will occur. Also, if there are any sediments in the water, moving the camera 4 will stir up the sediments, and the state of the sediments moving in the water will remain in the field of view for a certain period of time. Although this embodiment can also be applied to pipes in which water is not flowing, a challenge unique to underwater inspection is that detection accuracy can be improved by performing detection processing that matches the movement and position in order to mitigate the effects of such movement of foreign objects in the water. To achieve this, it is effective to equip the anomaly detection unit 30 with the following functions. (d) A function that determines which image to use for the determination, depending on the type of object to be detected: the image corresponding to when camera 4 is moving forward or the image corresponding to when camera 4 is moving backward. (e) A function that performs the above determination using only the corresponding image when the camera 4 is moving forward for a predetermined detection target. (f) A function that performs the above determination using only the corresponding image when the camera 4 is retracting for a predetermined detection target. (g) A function that performs the above determination for a predetermined detection target using the corresponding images when the camera 4 is moving forward and when it is moving backward. (h) A function that changes the processing content depending on whether the camera 4 is moving forward or backward for a predetermined detection target.

[0087] Regarding the direction of movement of camera 4, since the camera position / motion detection unit 20 can acquire position information or motion information of camera 4, it is easy to determine whether camera 4 is moving forward by pushing cable 6 in or backward by pulling cable 6 back.

[0088] Figure 15 shows an example of settings that define the movement and direction of camera 4 that can be applied to each of the five types of detection targets mentioned above. In Figure 15, a circle indicates that it is enabled, and an "x" indicates that it is disabled. This is merely an example, and other combinations of circles and "x"s are also possible. For example, although not shown in Figure 15, there may be a detection target where forward detection is set to "x" (disabled) and reverse detection is set to a circle (enabled).

[0089] When camera 4 is moving forward, any foreign matter that has accumulated is less likely to enter camera 4's field of view even if it is stirred up by the movement of camera 4. Therefore, detection processing is basically performed only when camera 4 is moving forward. On the other hand, for film-like deposits, the trace left by camera 4's movement may be detected, so anomaly detection and level determination are performed using images taken when camera 4 is moving backward. This is because when camera 4 is moving forward, any trace left by camera 4 is not in camera 4's field of view. Also, accumulated matter is more likely to be stirred up when camera 4 moves backward, so it is also possible to detect accumulated matter waiting when camera 4 is moving backward. Thus, when camera 4 recognizes that it is moving backward, anomaly detection and level determination of the corresponding detection target may be performed.

[0090] Sediment can be detected at the pixel level using deep learning, and film-like deposits on the inner surface of pipes can also be detected using a color-based method while moving forward. These processes are easier to perform when moving forward. Therefore, it is possible to improve detection accuracy by changing the detection process depending on the direction of movement of camera 4, or by switching the processing between forward and backward movement and combining the results of the detections to detect anomalies.

[0091] (A function that allows the unit of anomaly detection to be changed depending on the target of detection.) For the five types of detection targets mentioned above, the unit for anomaly detection and level determination may be changed depending on the type of detection target. In this case, it is effective to provide the anomaly detection unit 30 with the following functions. (i) A function that performs the above determination for each unit, treating one frame of the image as one unit for a predetermined detection target. (j) A function that performs the above determination for each unit, treating multiple frames of images as one unit for a predetermined detection target. (k) A function that performs the above determination for a predetermined detection target, treating all images as one unit.

[0092] Figure 16 shows an example of settings that define the judgment units (frame, interval, or whole) that can be applied to each of the five types of detection targets. In Figure 16, a circle indicates that it is valid, and a dash (-) indicates that it is not applicable.

[0093] For all detection targets, it is basically possible to perform anomaly detection and level determination on a per-frame basis. In this case, by associating the camera 4's position information obtained from the attribute information input unit 11 with the anomaly type and level, it becomes possible to manage the anomaly type and level in correspondence with the position. When these results are output by the output processing unit 50, it is possible to output the anomaly detection result and anomaly level in correspondence with the camera 4's position.

[0094] Furthermore, it is possible to obtain anomaly detection results in units of intervals consisting of multiple frames and output them together with area information (a unit of area larger than the position specified by the frame). For example, this applies when finding traces of camera 4 in internal adhesion or when making a judgment after examining a certain area, such as with floating objects. In this case, the area may be set by grouping frames at regular intervals, or each area may be set to an arbitrary length. When outputting the results in the output processing unit 50, the position of camera 4 is output in units of the grouped area. This can also be achieved by selecting representative detection results such as the worst value, best value, average value, and median value from the group of frame-level results obtained from multiple frames included in that area, based on the results obtained for each frame.

[0095] Alternatively, the system may be configured to output a single anomaly detection result for the entire video. For example, since the state of floating objects does not change much depending on the location, the system may output a single final result for the entire video. This can be achieved by selecting a representative detection result from the set of results obtained for all frames of the entire video, such as the worst value, best value, average value, or median value.

[0096] To enable switching the unit of judgment for each detection target, this can be done by pre-setting a fixed value, or by providing a means to change the setting using a separate configuration file or configuration screen.

[0097] • Integrated determination unit 40 The integrated determination unit 40 detects abnormalities individually for the five types of detection targets mentioned above, and then outputs the final result using the interrelationships between those results. However, it is also acceptable to output the final result directly without using the interrelationships.

[0098] When determining the integrated abnormality judgment result, the calculation method (such as a weighted addition method) used to generate the integrated judgment result may be changed (for example, by merging or excluding) depending on the interrelationship of the abnormality detection results for each detected object. For example, when judging rust, only rust may be used, but when judging sediment, both rust and sediment may be used. In this case, the integrated judgment unit 40 may have the following functions. (l) A function that changes the calculation method for generating the integrated judgment result according to the interrelationship of the anomaly detection results for each detected target, and generates the integrated judgment result.

[0099] As for how to obtain the integrated anomaly judgment result, for example, if Y(1) to Y(5) are the final results for each of the five types of anomalies, X1 to X5 are the individually obtained anomaly detection results, and a to e are the respective weight coefficients, then the final result may be obtained by weighting each anomaly detection result by multiplying it by a predetermined coefficient as shown in the following equation.

[0100] Y(i) = a*X1 + b*X2 + c*X3 + d*X4 + e*X5 This formula is applicable to any unit, whether it's a frame, a segment, or the whole, as shown in Figure 16.

[0101] By adjusting each coefficient in the formula, it becomes possible to respond appropriately when individual dependencies are clear. For example, when determining the abnormal level of sediment, it is possible to determine the result of the OR (logical disjunction) of the abnormality detection results of rust and sediment as the abnormality detection result (this can also be achieved by the coefficient of the Y formula above), or to exclude a certain abnormality from other abnormality detection results using the above formula.

[0102] Alternatively, a predetermined anomaly may be determined first, and the result may be used to adjust the result of another anomaly detection. For example, the anomaly level of a predetermined detection target may be adjusted depending on whether or not there is influence between the detection targets (for example, since sediment can be easily confused with floating matter, the anomaly level may be adjusted depending on the presence or absence of floating matter). In this case, the integrated determination unit 40 may have the following functions. (j) A function that adjusts the abnormality level of a predetermined detection target and generates an integrated judgment result depending on whether or not there is an influence between the detection targets.

[0103] For example, since floating objects occur across the entire area of ​​the image, it may be possible to first determine the presence or level of abnormalities caused by floating objects, and then adjust the detection criteria for other abnormality detection targets if there are many floating objects. When there are many floating objects, abnormalities in the image become difficult to see, which can cause existing abnormalities to be missed or, conversely, make it easier to identify abnormalities when there are none. It is also possible to lower or raise the level of abnormality to a predetermined level, and the length of the area (interval) for determining abnormalities may be changed according to the amount of floating objects. For example, when there are many floating objects, the influence of noise can be reduced by performing level determination from abnormality detection results obtained over a long range.

[0104] Furthermore, instead of limiting the system to detecting five types of anomalies, it may be possible to output a single overall judgment result for the entire pipe, regardless of the type of anomaly detected. This can be achieved by calculating the worst-case value, average value, weighted sum, etc., of the five types of anomaly level judgment results.

[0105] Output processing unit 50 The output processing unit 50 has a screen that allows the input video and anomaly detection results to be viewed in comparison. This output processing unit 50 may be composed of, for example, a personal computer (PC) and a display device such as an LCD display. In this case, the PC stores and manages the video data input by the video input unit 10, the attribute information input by the attribute information input unit 11, and the results detected and judged by the anomaly detection unit 30 and the integrated judgment unit 40. An example of the screen displayed on the display is shown in Figure 17.

[0106] The screen 60 shown in Figure 17 displays the input image 61, the anomaly detection result 62, a bar 63 indicating the rust anomaly judgment result, a bar 64 indicating the sediment anomaly judgment result, a bar 65 for seeking the video and a marker 66 indicating the current position of the displayed video, the anomaly judgment result at the current rust location 67, the anomaly judgment result at the current sediment location 68, and the overall judgment result 69 showing the anomaly judgment result for the entire video.

[0107] The bar 65 in the section labeled "Video" at the bottom of the screen is a seek bar for the video. By moving the marker 66 in the section labeled "Current Video Position," it is possible to view the image of any frame in the video. Although not shown in the example in Figure 17, it is also possible to create play, pause, and frame-by-frame buttons to automatically play the video.

[0108] The left side of bar 65 corresponds to the beginning of the video, and the right side corresponds to the last frame of the video. Above this, bars 63 and 64 display the corresponding locations of the anomaly detection items (in this example there are two types: rust and deposits, but you can increase the number of anomaly types). The vertical stripes indicate the location of images with an anomaly level of B, and the diagonal shading indicates the location of images with an anomaly level of D.

[0109] The input image 61 is the original image of the input image corresponding to the "current position of the video," and the anomaly detection result 62 is an image in which the results detected by the anomaly detection unit 30 and the integrated judgment unit 40 are superimposed on it. In this example image, sediment accumulated at the bottom of the pipe has been detected, and the area of ​​sediment in the anomaly detection result 62 image is enclosed in a frame to clearly indicate the location of the anomaly.

[0110] In addition, symbols, rectangles, or the superposition of transparent colors may be used on the image of the anomaly detection result 62 to indicate the location and type of anomalies detected by the anomaly detection unit 30 or the integrated judgment unit 40, making them easier to understand.

[0111] The rust anomaly detection results 67 and the sediment anomaly detection results 68 show the results of the anomaly level determination for the anomalies in the image at the currently displayed location. In this example image, there are no rust anomalies, so the anomaly level determination result is S, and sediment is present, so the anomaly level determination result is B.

[0112] The overall judgment result 69 displays the overall judgment result, which is the result of the anomaly judgment of the entire video. Basically, it displays the worst value of each anomaly level, but it is not limited to this, and the screen output content may also use the average value or a specific anomaly. In addition, attribute information input by the attribute information input unit 11 may also be displayed.

[0113] (operation) Next, an example of the operation of the pipe inspection device according to the first embodiment will be described with reference to the flowchart in Figure 18. Here, the explanation is based on the configuration example in Figure 2, but it is also possible to modify and implement the operation based on the configurations in Figures 3 and 4. Furthermore, the individual processes shown in Figure 18 are not necessarily performed in the order of the step numbers and can be changed as appropriate. For example, the order of steps S1 and S2 may be reversed, or they may be performed simultaneously.

[0114] First, the video input unit 10 receives video data from camera 4, which has been filmed while moving inside the water pipe 3 (step S1). Additionally, the attribute information input unit 11 receives attribute information indicating the attributes associated with camera 4 (step S2).

[0115] Next, the camera position / motion detection unit 20 uses the video data input by the video input unit 10 and the attribute information input by the attribute information input unit 11 to detect the position or movement of the camera 4 inside the pipe (step S3).

[0116] Next, the anomaly detection unit 30 selects one or more images from the video data of the detection target captured by the camera 4 according to the position or movement of the camera 4 detected by the camera position / motion detection unit 20, uses the selected images to determine whether there is an anomaly in each detection target inside the pipe and the level of the anomaly, and generates the determined result as an anomaly detection result for each detection target (step S4).

[0117] Next, the integrated determination unit 40 integrates the abnormality detection results generated by the abnormality detection unit 30 for each detection target to generate an integrated determination result indicating the level of abnormality in the pipe (step S5).

[0118] Finally, the output processing unit 50 outputs information including the anomaly detection result generated by the anomaly detection unit 30 for each detected target and the integrated judgment result generated by the integrated judgment unit 40 to a display device or storage device (step S6).

[0119] According to the first embodiment, by utilizing information indicating the position or movement of the camera 4, images (frames) to be used for inspection inside the pipe can be appropriately selected, thereby improving the accuracy of anomaly detection. This also reduces variability in diagnostic results and lowers the processing load in image diagnosis. Furthermore, the device and system can be realized without the addition of special sensors other than visible light sensors.

[0120] <Second Embodiment> Next, a second embodiment will be described. The following description will focus on the differences from the first embodiment.

[0121] (System Configuration) The overall system configuration is the same as shown in Figure 1, so its explanation will be omitted here.

[0122] (Configuration of pipe inspection device 1) Figure 19 shows an example of the configuration of the pipe inspection device 1 according to the second embodiment.

[0123] The configuration example in Figure 19 is the same as the configuration example in Figure 2, but with the integrated judgment unit 40 omitted. In this case, the output processing unit 50 outputs information including the anomaly detection result generated for each detection target by the anomaly detection unit 30 to a display device or storage device. Also, in this case, the overall judgment result 69 will not be displayed on the screen of Figure 17 mentioned above. The other configurations are the same as in the first embodiment.

[0124] (operation) Next, referring to the flowchart in Figure 20, an example of the operation of the pipe inspection device according to the second embodiment will be explained. Here, we will explain the differences from the flowchart in Figure 18 described above.

[0125] The operation example in Figure 20 is the same as the operation example in Figure 18 described above, but with the processing in step S5 omitted. In this case, in Figure 20, after the processing in step S4 is performed, in step S6', the output processing unit 50 outputs information including the abnormality detection result generated for each detection target by the abnormality detection unit 30 to a display device or storage device.

[0126] According to the second embodiment, the functional configuration of the pipe inspection device 1 can be simplified, and the integrated judgment process can be omitted, thereby reducing the burden on processing.

[0127] <Third Embodiment> Next, a third embodiment will be described. The following description will focus on the differences from the first embodiment.

[0128] (System Configuration) The overall system configuration is the same as shown in Figure 1, so its explanation will be omitted here.

[0129] (Configuration of pipe inspection device 1) Figure 21 shows an example of the configuration of the pipe inspection device 1 according to the third embodiment.

[0130] The configuration example in Figure 21 is a configuration in which a setting unit 12 is added to the configuration example in Figure 2. This setting unit 12 has the function of setting or changing information used in processing in at least one of the attribute information input unit 11, camera position / motion detection unit 20, anomaly detection unit 30, and integrated judgment unit 40, for example, in response to operations performed by the user.

[0131] For example, in the settings for the anomaly detection unit 30, it is possible to set and change the presence or absence of movement and the direction of movement of the camera 4. For example, desired settings and changes can be made through the screen shown in Figure 15. The settings and changes may be recorded in a settings file and loaded into a program for execution. Furthermore, regarding the units mentioned above, the contents shown in Figure 16 may be recorded in a settings file, and settings and changes may be made on the screen shown in Figure 15.

[0132] Furthermore, in the settings of the camera position and motion detection unit 20, it may be possible to switch between various thresholds and algorithms used when determining movement, compared to the method described in the first embodiment. In the settings of the integrated determination unit 40, it may be possible to set each coefficient of the weighted addition calculation formula for obtaining the overall result.

[0133] The other configurations are the same as in the first embodiment.

[0134] (operation) Next, with reference to the flowchart in Figure 22, an example of the operation related to the setting unit 12 of the pipe inspection device according to the third embodiment will be explained.

[0135] The setting unit 12 monitors user operations (step S11), and when any operation is detected (YES in step S12), it performs the corresponding setting or modification in accordance with that operation (step S13).

[0136] According to the third embodiment, the user can configure and change various settings, thereby improving the usability of the device.

[0137] As described in detail above, according to the embodiment, it becomes possible to detect abnormalities inside the pipe with high accuracy.

[0138] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. [Explanation of Symbols]

[0139] 1... Pipe inspection device, 2... Site, 3... Water pipe, 4... Camera, 5... Recording medium, 10... Video input unit, 11... Attribute information input unit, 12... Setting unit, 20... Camera position / motion detection unit, 30... Anomaly detection unit, 40... Integrated judgment unit, 50... Output processing unit.

Claims

[Claim 1] An in-pipe inspection device comprising an abnormality detection unit that, depending on the position or movement of a camera that photographs the inside of a pipe, selects one or more images from the video data of a target to be detected inside the pipe that the camera photographs, determines the presence or absence and level of abnormality of the target to be detected inside the pipe using the selected images, and generates an abnormality detection result for each target to be detected based on the determined result.

Images