A method and device for detecting inclination of electric pole

Abstract

The present application discloses a method and device for detecting the tilt of a utility pole, the method comprising: obtaining an image to be tested whose edge is parallel to the ground plane; inputting the image to be tested into a trained utility pole detection model to obtain an optimal candidate frame containing the utility pole; performing straight line detection on the utility pole in the image of the optimal candidate frame to obtain the edge line segments of the utility pole; calculating two angles between the edge line segments on both sides of the utility pole and the ground plane, and if the two angles are equal, it is determined that the utility pole is not tilted. The present application solves the technical problems that the current inspection method is time-consuming and labor-intensive, inefficient, and the accuracy cannot be guaranteed.

Description

Technical Field

[0001] The present application relates to the field of image detection technology, and in particular to a method and device for detecting the inclination of a utility pole. Background Art

[0002] Power poles and towers are an important part of the power grid structure. In recent years, with the rapid development of power grid construction, the scale of the power grid has been expanding, and the mileage of transmission lines has also increased rapidly. Most transmission line corridors are distributed in the suburban wilderness, and are greatly affected by objective natural conditions such as bad weather and hillside geographical conditions. The reliability of their operation directly affects the stable operation of the power grid. Therefore, it is necessary to conduct transmission line inspections in a timely manner to promptly discover power safety defects.

[0003] At present, the inspection of power towers is still mainly done manually, which is boring. Moreover, the inspection results are easily affected by the status of the inspectors and are unreliable. Summary of the invention

[0004] The present application provides a method and device for detecting the inclination of a utility pole, which solves the technical problems that the current inspection method is time-consuming and labor-intensive, inefficient, and has insufficient accuracy.

[0005] In view of this, the first aspect of the present application provides a method for detecting the inclination of a utility pole, the method comprising:

[0006] Acquire an image to be tested whose edge is parallel to the ground plane;

[0007] Inputting the image to be tested into a trained electric pole detection model to obtain an optimal candidate frame containing an electric pole;

[0008] Performing straight line detection on the electric poles in the image of the optimal candidate frame to obtain edge line segments of the electric poles;

[0009] The two angles between the edge line segments on both sides of the electric pole and the ground plane are calculated. If the two angles are equal, it is determined that the electric pole is not tilted.

[0010] Optionally, after acquiring the image to be measured whose edge is parallel to the ground plane, the method further includes:

[0011] Preprocessing the image to be tested.

[0012] Optionally, before inputting the image to be tested into a trained utility pole detection model to obtain an optimal candidate box containing a utility pole, the method further includes:

[0013] The utility pole detection model is trained.

[0014] Optionally, the training of the utility pole detection model includes:

[0015] Obtain a large number of images of telephone poles with different shooting angles, different states, different light, different distances and different backgrounds as training images;

[0016] Preprocessing the training image;

[0017] The preprocessed training images are used to train the Faster R-CNN neural network model until the Faster R-CNN neural network model outputs the optimal candidate frame that meets the preset requirements, thereby obtaining a trained utility pole detection model.

[0018] Optionally, performing straight line detection on the electric poles in the image of the optimal candidate frame to obtain edge segments of the electric poles specifically includes:

[0019] Using a Gaussian filter to smooth the image of the optimal candidate frame;

[0020] Calculate the gradient value and direction of each pixel in the image of the optimal candidate frame after smoothing;

[0021] Calculate the anchor pixel in the image of the optimal candidate frame according to the gradient value and direction of each pixel;

[0022] The anchor pixels are connected to draw edge segments in the image of the optimal candidate frame, where the edge segments are the outlines of the electric poles.

[0023] A second aspect of the present application provides a pole tilt detection device, the device comprising:

[0024] An image acquisition unit, used for acquiring an image to be measured whose edge is parallel to the ground plane;

[0025] A utility pole detection unit, used to input the image to be tested into a trained utility pole detection model to obtain an optimal candidate frame containing a utility pole;

[0026] A straight line detection unit, used to perform straight line detection on the electric poles in the image of the optimal candidate frame to obtain edge segments of the electric poles;

[0027] The angle calculation unit is used to calculate the two angles between the edge line segments on both sides of the electric pole and the ground plane. If the two angles are equal, it is determined that the electric pole is not tilted.

[0028] Optionally, also include:

[0029] The first preprocessing unit is used to preprocess the image to be tested.

[0030] Optionally, also include:

[0031] A training unit is used to train the utility pole detection model.

[0032] Optionally, the training unit further includes:

[0033] A training image acquisition unit, used to acquire a large number of images of electric poles with different shooting angles, different states, different light, different distances and different backgrounds as training images;

[0034] A second preprocessing unit, used for preprocessing the training image;

[0035] The training unit is used to train the Faster R-CNN neural network model with the preprocessed training image until the Faster R-CNN neural network model outputs the optimal candidate frame that meets the preset requirements, that is, the utility pole detection model is obtained.

[0036] Optionally, the straight line detection unit further includes:

[0037] A filtering unit, used for smoothing the image of the optimal candidate frame using a Gaussian filter;

[0038] A calculation unit, used to calculate the gradient value and direction of each pixel in the image of the optimal candidate frame after smoothing;

[0039] An anchor point calculation unit, used to calculate the anchor point pixel in the image of the optimal candidate frame according to the gradient value and direction of each pixel;

[0040] The edge line segment acquisition unit is used to connect the anchor point pixels and draw edge line segments in the image of the optimal candidate frame, wherein the edge line segments are the outlines of the electric poles.

[0041] It can be seen from the above technical solutions that this application has the following advantages:

[0042] In the present application, a method and device for detecting the inclination of a utility pole are provided, which obtain an image to be tested whose edge is parallel to the ground plane; input the image to be tested into a trained utility pole detection model to obtain an optimal candidate frame containing the utility pole; perform straight line detection on the utility pole in the image of the optimal candidate frame to obtain the edge line segments of the utility pole; calculate the two angles between the edge line segments on both sides of the utility pole and the ground plane, and if the two angles are equal, it is determined that the utility pole is not tilted. The present application uses a neural network algorithm to obtain an optimal candidate frame containing the utility pole, and only performs straight line detection on the image within the candidate frame to avoid interference from other objects in the image to be tested; in addition, the edge line segments of the utility pole are obtained through straight line detection, the angles between the edge line segments on both sides of the utility pole and the horizontal plane are calculated, and the two angles are judged, making the calculation process simple and quick. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A method flow chart of an embodiment of a method for detecting the inclination of a utility pole according to the present application;

[0044] Figure 2 A method flow chart of another embodiment of a method for detecting the inclination of a utility pole according to the present application;

[0045] Figure 3 This is a schematic structural diagram of an embodiment of a utility pole tilt detection device of the present application;

[0046] Figure 4 This is a detection flow chart of the Faster R-CNN neural network model used in this application;

[0047] Figure 5 This is a schematic diagram of an embodiment of the present application for comparing the angles between utility poles and a horizontal plane. DETAILED DESCRIPTION

[0048] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0049] Figure 1 A method flow chart of an embodiment of a method for detecting the inclination of a utility pole according to the present application is shown as follows: Figure 1 As shown, Figure 1 Included:

[0050] 101. Obtain an image to be tested whose edge is parallel to the ground plane.

[0051] It should be noted that the image to be tested should be taken by a camera in a horizontal state. For example, a drone with a gyroscope sensor can be used to take the image to be tested, so as to obtain an image to be tested whose edge is parallel to the horizon. In addition, the image to be tested should contain a complete electric pole.

[0052] 102. Input the image to be tested into the trained electric pole detection model to obtain the optimal candidate box containing the electric pole.

[0053] It should be noted that the utility pole monitoring model used in this application is based on a neural network. The utility pole monitoring model is trained by using a large number of images containing utility poles as training images until a candidate frame including a complete utility pole can be obtained, thereby obtaining a trained utility pole monitoring model. By inputting the image to be tested into the utility pole monitoring model, the optimal candidate frame containing the utility pole is obtained to facilitate subsequent straight line detection. The edge of the optimal candidate frame should also be parallel to the horizontal plane.

[0054] 103. Perform straight line detection on the electric poles in the image of the optimal candidate frame to obtain edge segments of the electric poles.

[0055] It should be noted that after obtaining the optimal candidate frame, straight line detection can be performed on the image in the optimal candidate frame to measure the edge segments of the electric pole in the image, thereby determining the outline of the electric pole.

[0056] 104. Calculate the two angles between the edge segments on both sides of the electric pole and the ground plane. If the two angles are equal, it is determined that the electric pole is not tilted.

[0057] It should be noted that after obtaining the edge segments of the electric pole, the angles between the edge segments on both sides of the electric pole and the horizontal plane can be calculated, and the two angles can be judged. Specifically, if the two angles are equal, the electric pole is normal; if the two angles are not equal, the electric pole is tilted. This judgment method is applicable to different types of electric poles.

[0058] The present application obtains the optimal candidate frame containing the utility pole by adopting a neural network algorithm, and only performs straight line detection on the image within the candidate frame to avoid interference from other objects in the image to be tested; in addition, the edge segments of the utility pole are obtained through straight line detection, the angles between the edge segments on both sides of the utility pole and the horizontal plane are calculated, and the two angles are judged, making the calculation process simple and quick.

[0059] The present application also provides another embodiment of a method for detecting the inclination of a utility pole. Figure 2 As shown, Figure 2 Included:

[0060] 201. Obtain a large number of telephone pole images with different shooting angles, different states, different light, different distances and different backgrounds as training images.

[0061] It should be noted that in order to make the detection results of the obtained pole detection model as accurate as possible, a large number of pole images with different shooting angles, different states, different light, different distances and different backgrounds are used as training images.

[0062] 202. Preprocess the training images.

[0063] It should be noted that since the input size of the neural network model is of any size. But it is often limited by hardware (i.e., the device running the algorithm), therefore, the image input size of the neural network model needs to be set to a fixed length and width, but the image data to be processed varies in size, so the training images need to be scaled so that the training image size becomes a uniform aspect ratio. In addition, since this application is not interested in the brightness of the image, but more in its content, it makes sense to remove the mean of the pixels for each data point at this time, therefore, the training images can be normalized so that the average brightness value of the image is removed.

[0064] 203. The preprocessed training images are used to train the Faster R-CNN neural network model until the Faster R-CNN neural network model outputs an optimal candidate box that meets the preset requirements, thereby obtaining a trained utility pole detection model.

[0065] It should be noted that the present application can use a deep learning neural network based on Faster R-CNN for detection and identification. Faster R-CNN is a two-stage target detection algorithm. Compared with the common single-stage detection algorithms SSD and YOLO, although the detection speed of Faster R-CNN is not as good as SSD and YOLO, Faster R-CNN has a higher detection accuracy. In order to ensure the accuracy of subsequent recognition of the status of utility poles, the recognition and detection of utility poles is the basis for subsequent processing. The accuracy of detection is more important than the detection speed, so Faster R-CNN is used as the deep learning algorithm for detection and identification of utility poles.

[0066] Specifically, the detection process of Faster R-CNN is as follows Figure 4 As shown in the figure: the training image is input into VGG16 for feature extraction, and the extracted feature map is shared with RPN (Region Proposal Network) to generate the original target candidate box. Then, the classifier is used to determine the category of the features extracted from the original candidate box. For the candidate box belonging to the pole feature, the regressor is used to further adjust its position until a candidate box containing the complete pole is obtained.

[0067] 204. Obtain an image to be tested whose edge is parallel to the ground plane.

[0068] It should be noted that the image to be tested should be taken by a camera in a horizontal state. For example, a drone with a gyroscope sensor can be used to take the image to be tested, so as to obtain an image to be tested whose edge is parallel to the horizon. In addition, the image to be tested should contain a complete electric pole.

[0069] 205. Preprocess the image to be tested.

[0070] It should be noted that the image input size of the neural network model can be set to a fixed length and width, that is, the image to be tested is scaled so that the size of the image to be tested becomes a uniform aspect ratio. In addition, since this application is not interested in the brightness of the image, but is more concerned with the content of the image to be tested, it is meaningful to remove the mean of the pixels for each data point, and the image to be tested can be normalized to remove the average brightness value of the image.

[0071] 206. Input the image to be tested into the trained electric pole detection model to obtain the optimal candidate box containing the electric pole.

[0072] It should be noted that the utility pole monitoring model used in this application is based on a neural network. The utility pole monitoring model is trained by using a large number of images containing utility poles as training images until a candidate frame including a complete utility pole can be obtained, thereby obtaining a trained utility pole monitoring model. By inputting the image to be tested into the utility pole monitoring model, the optimal candidate frame containing the utility pole is obtained to facilitate subsequent straight line detection. The edge of the optimal candidate frame should also be parallel to the horizontal plane.

[0073] 207. Use a Gaussian filter to smooth the image of the optimal candidate frame.

[0074] It should be noted that in actual application, the camera is affected by factors such as lighting and climate when collecting images, which may cause noise in the collected images. This application uses a comprehensive filtering method that combines traditional filtering and fast guided filtering to perform image filtering, which has a strong edge preservation effect and a good smoothing effect.

[0075] 208. Calculate the gradient size and direction of each pixel in the image of the optimal candidate frame after smoothing.

[0076] It should be noted that the formula for calculating the gradient size and direction of each pixel is as follows:

[0077]

[0078]

[0079]

[0080]

[0081] Where I(x,y) represents the output image intensity at the pixel point (x,y), g(x,y) represents the gradient value of the pixel point, and angle(x,y) represents the pixel angle, that is, the gradient direction.

[0082] 209. Calculate the anchor pixel in the image of the optimal candidate box according to the gradient size and direction of each pixel.

[0083] It should be noted that if there is a pixel whose gradient value is greater than or equal to the gradient values ​​of two adjacent pixels in the gradient direction of the pixel, then this pixel is regarded as an anchor pixel.

[0084] 210 connects the anchor pixels and draws edge segments in the image of the optimal candidate frame, where the edge segments are the outlines of the electric poles.

[0085] 211. Calculate the two angles between the edge segments on both sides of the electric pole and the ground plane. If the two angles are equal, it is determined that the electric pole is not tilted.

[0086] It should be noted that due to the possibility of certain errors in shooting, the angles a and b calculated based on the horizon, if the difference between a and b is not greater than 10 degrees, the posture is judged to be normal, and if the difference between a and b is greater than 10 degrees, it is judged to be tilted. For details, please refer to Figure 5 , Figure 5 A schematic diagram comparing the angles between utility poles and the horizontal plane.

[0087] The above is an embodiment of the method of the present application. The present application also includes an embodiment of a pole tilt detection device, such as Figure 3 As shown, Figure 3 Included:

[0088] The image acquisition unit 301 is used to acquire an image to be measured whose edge is parallel to the ground plane.

[0089] The utility pole detection unit 302 is used to input the image to be detected into the trained utility pole detection model to obtain the optimal candidate box containing the utility pole.

[0090] The line detection unit 303 is used to perform line detection on the electric poles in the image of the optimal candidate frame to obtain edge segments of the electric poles.

[0091] The angle calculation unit 304 is used to calculate two angles between the edge segments on both sides of the electric pole and the ground plane. If the two angles are equal, it is determined that the electric pole is not tilted.

[0092] In a specific embodiment, it also includes:

[0093] The first preprocessing unit 305 is used to preprocess the image to be tested.

[0094] The training unit 306 is used to train the utility pole detection model.

[0095] The training unit 306 also includes:

[0096] The training image acquisition unit 3061 is used to acquire a large number of utility pole images with different shooting angles, different states, different light, different distances and different backgrounds as training images.

[0097] The second preprocessing unit 3062 is used to preprocess the training image.

[0098] The training unit 3063 is used to train the Faster R-CNN neural network model with the preprocessed training image until the Faster R-CNN neural network model outputs the optimal candidate box that meets the preset requirements, that is, the utility pole detection model is obtained.

[0099] The straight line detection unit 303 also includes:

[0100] The filtering unit 3031 is used to smooth the image of the optimal candidate frame using a Gaussian filter.

[0101] The calculation unit 3032 is used to calculate the gradient value and direction of each pixel in the image of the optimal candidate frame after smoothing.

[0102] An anchor point calculation unit 3033, used to calculate anchor point pixels in the image of the optimal candidate frame, where the anchor point pixels are pixels with a high probability among the edge pixels of the image of the optimal candidate frame;

[0103] The edge segment acquisition unit 3034 is used to connect the anchor point pixels and draw the edge segments in the image of the optimal candidate frame, where the edge segments are the outlines of the electric poles.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0105] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0106] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0107] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0108] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, English abbreviation: ROM), random access memory (full name in English: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program codes.

[0109] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for detecting the inclination of a utility pole, characterized in that: include: Acquire an image to be tested whose edge is parallel to the ground plane; Inputting the image to be tested into a trained electric pole detection model to obtain an optimal candidate frame containing an electric pole; Performing straight line detection on the electric poles in the image of the optimal candidate frame to obtain edge line segments of the electric poles; Calculate two angles between the edge line segments on both sides of the electric pole and the ground plane. If the two angles are equal, it is determined that the electric pole is not tilted. Inputting the image to be tested into the trained electric pole detection model to obtain the optimal candidate box containing the electric pole also includes: Training the utility pole detection model; The performing straight line detection on the electric poles in the image of the optimal candidate frame to obtain edge line segments of the electric poles specifically includes: Using a Gaussian filter to smooth the image of the optimal candidate frame; Calculate the gradient value and direction of each pixel in the image of the optimal candidate frame after smoothing; Calculate the anchor pixel in the image of the optimal candidate frame according to the gradient value and direction of each pixel; The anchor pixels are connected to draw edge segments in the image of the optimal candidate frame, where the edge segments are the outlines of the electric poles.

2. The method for detecting the inclination of a utility pole according to claim 1, characterized in that: After acquiring the image to be measured whose edge is parallel to the ground plane, the method further includes: Preprocessing the image to be tested.

3. The method for detecting the inclination of a utility pole according to claim 1, characterized in that: The training of the utility pole detection model comprises: Obtain a large number of images of telephone poles with different shooting angles, different states, different light, different distances and different backgrounds as training images; Preprocessing the training image; The preprocessed training images are used to train the Faster R-CNN neural network model until the Faster R-CNN neural network model outputs the optimal candidate frame that meets the preset requirements, thereby obtaining a trained utility pole detection model.

4. A utility pole tilt detection device, characterized in that: include: An image acquisition unit, used for acquiring an image to be measured whose edge is parallel to the ground plane; A utility pole detection unit, used to input the image to be tested into a trained utility pole detection model to obtain an optimal candidate frame containing a utility pole; A straight line detection unit, used to perform straight line detection on the electric poles in the image of the optimal candidate frame to obtain edge segments of the electric poles; An angle calculation unit, used to calculate two angles between the edge line segments on both sides of the electric pole and the ground plane, and if the two angles are equal, it is determined that the electric pole is not tilted; A training unit, used for training the utility pole detection model; The straight line detection unit also includes: A filtering unit, used for smoothing the image of the optimal candidate frame using a Gaussian filter; A calculation unit, used to calculate the gradient value and direction of each pixel in the image of the optimal candidate frame after smoothing; An anchor point calculation unit, used to calculate the anchor point pixel in the image of the optimal candidate frame according to the gradient value and direction of each pixel; The edge line segment acquisition unit is used to connect the anchor point pixels and draw edge line segments in the image of the optimal candidate frame, wherein the edge line segments are the outlines of the electric poles.

5. The utility pole tilt detection device according to claim 4, characterized in that: Also includes: The first preprocessing unit is used to preprocess the image to be tested.

6. The utility pole tilt detection device according to claim 4, characterized in that: The training unit also includes: A training image acquisition unit, used to acquire a large number of images of electric poles with different shooting angles, different states, different light, different distances and different backgrounds as training images; A second preprocessing unit, used for preprocessing the training image; The training unit is used to train the Faster R-CNN neural network model with the preprocessed training image until the Faster R-CNN neural network model outputs the optimal candidate frame that meets the preset requirements, that is, a trained utility pole detection model is obtained.

Images