A method and system for identifying tree obstacles in high-altitude wire erection

Abstract

The present invention relates to the technical field of power operation and maintenance, and discloses a method and system for identifying tree obstacles in high-altitude overhead lines. The method performs object recognition on overhead line images collected at multiple angles of a target overhead line through a trained fully convolutional neural network, identifies the areas where the wires and tree obstacles are located, marks the wires and tree obstacles with detection frames, calculates the intersection over union between the detection frames of the wires and the detection frames of the tree obstacles corresponding to each overhead line image to obtain the overlap degree, and performs weighted calculation on the overlap degrees of all overhead line images to obtain the weighted overlap degree. By comparing the size relationship between the weighted overlap degree and a preset overlap degree threshold, it is determined whether the tree obstacle overlaps with the target overhead line, thereby improving the working efficiency and accuracy of identifying tree obstacles in high-altitude overhead lines.

Description

Technical Field

[0001] The present invention relates to the technical field of power operation and maintenance, and particularly to a method and system for identifying tree obstacles in overhead power lines. Background Art

[0002] Overhead power line erection is an important part of power grid infrastructure construction and the basis for ensuring the safe and stable operation of the distribution network. However, in the natural environment, overhead power line erection is affected by many uncertain factors, such as the growth of vegetation and sudden natural disasters. Once such an accident occurs, it will not only cause a large amount of property losses but also may threaten the safety of people's electricity use. The traditional method is to invite experienced power grid personnel to conduct on-site inspections to determine whether the surrounding environment of the overhead power lines poses a threat to the power grid. However, with the large-scale construction and development of the distribution network, the number of overhead power lines is increasing, and they are also distributed in places with complex terrains such as mountains and forests. The task of personnel inspections is becoming increasingly heavy, and even due to the complex terrain, it may cause injuries to personnel and damage to tools. This poses a quite severe safety challenge to the identification of tree obstacles in overhead power lines and greatly affects the work efficiency and accuracy. Summary of the Invention

[0003] The present invention provides a method and system for identifying tree obstacles in overhead power lines, which solves the technical problem of low work efficiency and accuracy in identifying tree obstacles in overhead power lines.

[0004] In view of this, the first aspect of the present invention provides a method for identifying tree obstacles in overhead power lines, including the following steps:

[0005] Obtain overhead line images at multiple angles of a target overhead line and construct an overhead line image set;

[0006] Based on a pre-trained fully convolutional neural network, perform object recognition on each overhead line image in the overhead line image set, identify the regions where the wires and tree obstacles are located, and use detection frames to label the wires and the tree obstacles respectively;

[0007] Calculate the intersection over union of the detection frame of the wire and the detection frame of the tree obstacle corresponding to each overhead line image to obtain the overlap degree;

[0008] Assign corresponding weight values to the overlap degrees according to the angles corresponding to each overhead line image, and perform weighted calculation on the overlap degrees according to the weight values corresponding to the overhead line images to obtain the weighted overlap degrees;

[0009] Judge whether the weighted overlap degree is greater than a preset overlap degree threshold. If the above determination is yes, generate a warning signal for feedback.

[0010] Preferably, the step of obtaining overhead line images at multiple angles of a target overhead line and constructing an overhead line image set specifically includes:

[0011] Obtain the overhead line images at multiple angles of the target overhead line based on multiple cameras pre-installed on the target overhead line;

[0012] Perform image enhancement processing on the overhead line images, and construct an overhead line image set from the enhanced overhead line images.

[0013] Preferably, the pre-trained fully convolutional neural network is trained using the YOLOv3 neural network.

[0014] Preferably, the step of assigning corresponding weights to the overlap degree according to the angle corresponding to each overhead line image, and performing weighted calculation on the overlap degree according to the weight corresponding to the overhead line image specifically includes:

[0015] Taking the running direction of the target overhead line as the zero-axis reference, determining the deviation angle of the angle corresponding to each overhead line image relative to the zero-axis reference, and determining the corresponding weight according to the ratio of the deviation angles corresponding to each overhead line image;

[0016] Perform weighted calculation on the overlap degree according to the weight corresponding to each overhead line image to obtain the weighted overlap degree.

[0017] Preferably, the method further includes:

[0018] Obtain the geographical location of the target overhead line;

[0019] When the weighted overlap degree is greater than the preset overlap degree threshold, report the geographical location of the target overhead line and the overhead line images to the master station.

[0020] In a second aspect, the present invention also provides an overhead line tree obstacle recognition system, including:

[0021] An image acquisition module, configured to obtain overhead line images at multiple angles of a target overhead line and construct an overhead line image set;

[0022] A recognition module, configured to perform target recognition on each overhead line image in the overhead line image set based on a pre-trained fully convolutional neural network, identify the areas where the wires and tree obstacles are located, and label the wires and the tree obstacles respectively using detection frames;

[0023] A first calculation module, configured to calculate the intersection over union of the detection frames of the wires and the detection frames of the tree obstacles corresponding to each overhead line image to obtain the overlap degree;

[0024] A second calculation module, configured to assign corresponding weights to the overlap degree according to the angle corresponding to each overhead line image, and perform weighted calculation on the overlap degree according to the weight corresponding to the overhead line image to obtain the weighted overlap degree;

[0025] A judgment module, configured to judge whether the weighted overlap degree is greater than a preset overlap degree threshold. If the above judgment is yes, a warning signal is generated for feedback.

[0026] Preferably, the image acquisition module specifically includes:

[0027] A plurality of cameras, which are installed on the target overhead line and are used to acquire overhead line images at multiple angles of the target overhead line;

[0028] An image processing module, configured to perform image enhancement processing on the overhead line image, and construct an overhead line image set from the enhanced overhead line image.

[0029] Preferably, the pre-trained fully convolutional neural network is trained using the YOLOv3 neural network.

[0030] Preferably, the second calculation module specifically includes:

[0031] A weight assignment module, configured to use the routing direction of the target overhead line as the zero-axis reference, determine the deviation angle of the angle corresponding to each overhead line image relative to the zero-axis reference, and determine the corresponding weight according to the ratio of the deviation angles corresponding to each overhead line image;

[0032] A weighted calculation sub-module, configured to perform weighted calculation on the overlap degree according to the weight corresponding to each overhead line image to obtain the weighted overlap degree.

[0033] Preferably, the system further includes:

[0034] A position acquisition module, configured to acquire the geographical location of the target overhead line;

[0035] A reporting module, configured to report the geographical location of the target overhead line and the overhead line image to the main station when the weighted overlap degree is greater than the preset overlap degree threshold.

[0036] It can be seen from the above technical solutions that the present invention has the following advantages:

[0037] The present invention performs target recognition on the overhead line images at multiple angles of the collected target overhead line through a trained fully convolutional neural network, identifies the areas where the wires and tree obstacles are located, marks the wires and tree obstacles with detection frames, calculates the intersection over union of the detection frames of the wires and the detection frames of the tree obstacles corresponding to each overhead line image to obtain the overlap degree, and performs weighted calculation on the overlap degrees of all overhead line images to obtain the weighted overlap degree, and compares the size relationship between the weighted overlap degree and the preset overlap degree threshold, so as to determine whether the tree obstacle overlaps with the target overhead line, thereby improving the working efficiency and accuracy of high-altitude overhead line tree obstacle recognition. Description of the Drawings

[0038] Figure 1 The flowchart of a method for identifying tree obstacles in overhead power lines provided by an embodiment of the present invention;

[0039] Figure 2 The structural schematic diagram of a system for identifying tree obstacles in overhead power lines provided by an embodiment of the present invention. Detailed implementation manners

[0040] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0041] For ease of understanding, please refer to Figure 1 , a method for identifying tree obstacles in overhead power lines provided by the present invention includes the following steps:

[0042] S1. Obtain overhead power line images at multiple angles of a target overhead power line and construct an overhead power line image set.

[0043] Among them, the overhead power line images at multiple angles can be obtained by a drone, and the shooting is centered on the overhead power line. In one example, the overhead power line images at multiple angles of the target overhead power line are obtained by multiple cameras pre-installed on the target overhead power line.

[0044] In practical applications, the shooting angles need to include the angles on both pairs of sides of the target overhead power line.

[0045] S2. Based on a pre-trained fully convolutional neural network, perform target recognition on each overhead power line image in the overhead power line image set, identify the areas where the wires and tree obstacles are located, and use detection frames to label the wires and tree obstacles respectively.

[0046] Among them, the pre-trained fully convolutional neural network is trained by using the YOLOv3 neural network.

[0047] During the training process, a sample library is constructed with the mapping relationships between a large number of wire images and the corresponding detection frames, and another sample library is constructed with the mapping relationships between a large number of tree obstacle images and the corresponding detection frames; the YOLOv3 neural network is used to train the two sample libraries respectively, and the wire images and tree obstacle images are used as inputs, and the corresponding detection frames are used as output quantities. After optimizing the parameters of the trained YOLOv3 neural network, the fully convolutional neural network in this embodiment is obtained.

[0048] S3. Calculate the intersection over union (IOU) between the detection bounding boxes of the wires and the detection bounding boxes of the tree obstacles corresponding to each overhead line image to obtain the overlap degree.

[0049] It can be understood that the intersection over union, that is, IOU, is the ratio of the intersection to the union between the detection bounding box of the wire and the detection bounding box of the tree obstacle. By obtaining the IOU of the two detection bounding boxes, the overlap degree is obtained.

[0050] S4. Assign corresponding weight values to the overlap degrees according to the angles corresponding to each overhead line image, and perform weighted calculation on the overlap degrees according to the weight values corresponding to the overhead line images to obtain the weighted overlap degrees.

[0051] S5. Determine whether the weighted overlap degree is greater than a preset overlap degree threshold. If the above determination is yes, generate a warning signal for feedback.

[0052] It should be noted that a method for identifying tree obstacles in overhead lines provided in this embodiment performs target recognition on overhead line images at multiple angles of a target overhead line collected by a trained fully convolutional neural network, identifies the areas where the wires and tree obstacles are located, marks the wires and tree obstacles with detection bounding boxes, calculates the intersection over union between the detection bounding boxes of the wires and the detection bounding boxes of the tree obstacles corresponding to each overhead line image to obtain the overlap degree, performs weighted calculation on the overlap degrees of all overhead line images to obtain the weighted overlap degrees, and compares the size relationship between the weighted overlap degrees and the preset overlap degree threshold, so as to determine whether the tree obstacles overlap with the target overhead line, thereby improving the working efficiency and accuracy of identifying tree obstacles in overhead lines.

[0053] In a specific embodiment, step S1 specifically includes:

[0054] S101. Obtain overhead line images at multiple angles of the target overhead line based on multiple cameras pre-installed on the target overhead line.

[0055] S102. Perform image enhancement processing on the overhead line images, and construct an overhead line image set with the enhanced overhead line images.

[0056] Among them, the methods of image enhancement processing can be grayscale processing, histogram equalization processing, brightness adjustment, and denoising processing.

[0057] In a specific embodiment, step S4 specifically includes:

[0058] S401. Take the running direction of the target overhead line as the zero-axis reference, determine the deviation angle of the angle corresponding to each overhead line image relative to the zero-axis reference, and determine the corresponding weight value according to the ratio of the deviation angles corresponding to each overhead line image.

[0059] It can be understood that, taking the running direction of the target overhead line as the zero-axis reference, the larger the deviation from the running direction of the overhead line, the larger the shooting space and the greater the importance of shooting. Therefore, the corresponding weight can be determined according to the ratio of the deviation angles corresponding to each overhead line image.

[0060] In one example, based on the running direction of the target overhead line as the zero-axis reference, overhead line images of the target overhead line are obtained at four shooting angles. The first shooting angle is 0°, the second shooting angle is 30°, the third shooting angle is 60°, and the fourth shooting angle is 90°. Then the deviation angles of the four shooting angles relative to the zero-axis reference are 0, 30, 60, and 90 respectively, and the corresponding weights are 1:3:6:9.

[0061] S402. Perform weighted calculation on the overlap degree according to the weight corresponding to each overhead line image to obtain the weighted overlap degree.

[0062] In a specific embodiment, the method further includes:

[0063] S6. Obtain the geographical location of the target overhead line.

[0064] It can be understood that the geographical location of the target overhead line can be obtained by GPS technology.

[0065] S7. When the weighted overlap degree is greater than a preset overlap degree threshold, report the geographical location of the target overhead line and the overhead line image to the master station.

[0066] It can be understood that reporting the geographical location of the target overhead line and the overhead line image to the master station enables the operation and maintenance personnel to timely maintain the target overhead line.

[0067] The above is a detailed description of an embodiment of a method for identifying tree obstacles on overhead lines provided by the present invention. The following is a detailed description of an embodiment of a system for identifying tree obstacles on overhead lines provided by the present invention.

[0068] For ease of understanding, please refer to Figure 2 , a system for identifying tree obstacles on overhead lines provided by the present invention includes:

[0069] An image acquisition module 100, configured to acquire overhead line images of the target overhead line at multiple angles and construct an overhead line image set;

[0070] An identification module 200, configured to perform target identification on each overhead line image in the overhead line image set based on a pre-trained fully convolutional neural network, identify the areas where the wires and tree obstacles are located, and use detection frames to label the wires and tree obstacles respectively;

[0071] A first calculation module 300, configured to calculate the intersection over union of the detection frames of the wires and the detection frames of the tree obstacles corresponding to each overhead line image to obtain the overlap degree;

[0072] A second calculation module 400, configured to assign weights to the overlapping degrees corresponding to each overhead line image according to the angles corresponding to the overhead line images, and perform weighted calculation on the overlapping degrees according to the weights corresponding to the overhead line images to obtain weighted overlapping degrees;

[0073] A judgment module 500, configured to judge whether the weighted overlapping degree is greater than a preset overlapping degree threshold. If the above judgment is yes, a warning signal is generated for feedback.

[0074] In a specific embodiment, the image acquisition module specifically includes:

[0075] A plurality of cameras, which are installed on a target overhead line and are used to acquire overhead line images at multiple angles of the target overhead line;

[0076] An image processing module, configured to perform image enhancement processing on the overhead line images, and construct an overhead line image set from the overhead line images after image enhancement.

[0077] In a specific embodiment, the pre-trained fully convolutional neural network is trained using the YOLOv3 neural network.

[0078] In a specific embodiment, the second calculation module specifically includes:

[0079] A weight assignment module, configured to use the routing direction of the target overhead line as the zero-axis reference, determine the deviation angle of the angle corresponding to each overhead line image relative to the zero-axis reference, and determine the corresponding weights according to the ratio of the deviation angles corresponding to each overhead line image;

[0080] A weighted calculation sub-module, configured to perform weighted calculation on the overlapping degrees according to the weights corresponding to each overhead line image to obtain weighted overlapping degrees.

[0081] In a specific embodiment, the system further includes:

[0082] A position acquisition module, configured to acquire the geographical location of the target overhead line;

[0083] A reporting module, configured to report the geographical location of the target overhead line and the overhead line images to a master station when the weighted overlapping degree is greater than a preset overlapping degree threshold.

[0084] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described system, device, and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0085] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, 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 displayed or discussed couplings or direct couplings or communication connections between each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical or other forms.

[0086] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0087] In addition, each functional unit in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0088] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for identifying tree obstacles in high-altitude wire erection, characterized in that, It includes the following steps: Obtain the overhead line images at multiple angles of the target overhead line and construct an overhead line image set; Based on a pre-trained fully convolutional neural network, perform target recognition on each overhead line image in the overhead line image set, identify the areas where the wires and tree obstacles are located, and use detection frames to label the wires and the tree obstacles respectively; Calculate the intersection over union of the detection frame of the wire and the detection frame of the tree obstacle corresponding to each overhead line image to obtain the overlap degree; Assign weights to the overlap degrees corresponding to the angles of each overhead line image, and perform weighted calculation on the overlap degrees according to the weights corresponding to the overhead line images, specifically including: Taking the routing direction of the target overhead line as the zero-axis reference, determine the deviation angle of the angle corresponding to each overhead line image relative to the zero-axis reference, and determine the corresponding weight according to the ratio of the deviation angles corresponding to each overhead line image; Perform weighted calculation on the overlap degrees according to the weights corresponding to each overhead line image to obtain the weighted overlap degree; Judge whether the weighted overlap degree is greater than a preset overlap degree threshold. If the above judgment is yes, generate a warning signal for feedback.

2. The method for identifying tree obstacles in high-altitude wire erection according to claim 1, wherein The step of obtaining the overhead line images at multiple angles of the target overhead line and constructing an overhead line image set specifically includes: Obtain the overhead line images at multiple angles of the target overhead line based on multiple cameras pre-installed on the target overhead line; Perform image enhancement processing on the overhead line images, and construct an overhead line image set with the enhanced overhead line images.

3. The method for identifying tree obstacles in high-altitude wire erection according to claim 1, characterized in that The pre-trained fully convolutional neural network is trained using the YOLOv3 neural network.

4. The method for identifying tree obstacles in high-altitude wire erection according to claim 1, wherein It also includes: Obtain the geographical location of the target overhead line; When the weighted overlap degree is greater than the preset overlap degree threshold, report the geographical location of the target overhead line and the overhead line images to the master station.

5. An overhead line tree obstacle recognition system, characterized in that It includes: An image acquisition module, which is used to obtain the overhead line images at multiple angles of the target overhead line and construct an overhead line image set; An identification module, which is used to perform target recognition on each overhead line image in the overhead line image set based on a pre-trained fully convolutional neural network, identify the areas where the wires and tree obstacles are located, and use detection frames to label the wires and the tree obstacles respectively; A first calculation module, which is used to calculate the intersection over union of the detection frame of the wire and the detection frame of the tree obstacle corresponding to each overhead line image to obtain the overlap degree; A second calculation module, which is used to assign weights to the overlap degrees corresponding to the angles of each overhead line image, and perform weighted calculation on the overlap degrees according to the weights corresponding to the overhead line images to obtain the weighted overlap degree; A judgment module, which is used to judge whether the weighted overlap degree is greater than a preset overlap degree threshold. If the above judgment is yes, generate a warning signal for feedback; The second calculation module specifically includes: A weight assignment module, which is used to take the routing direction of the target overhead line as the zero-axis reference, determine the deviation angle of the angle corresponding to each overhead line image relative to the zero-axis reference, and determine the corresponding weight according to the ratio of the deviation angles corresponding to each overhead line image; A weighted calculation sub-module, which is used to perform weighted calculation on the overlap degrees according to the weights corresponding to each overhead line image to obtain the weighted overlap degree.

6. The high-altitude overhead line tree obstacle recognition system according to claim 5, characterized in that The image acquisition module specifically includes: Multiple cameras, and multiple of the cameras are installed on the target overhead line to obtain overhead line images at multiple angles of the target overhead line; An image processing module, configured to perform image enhancement processing on the overhead line images and construct an overhead line image set from the overhead line images after image enhancement.

7. The high-altitude overhead line tree obstacle recognition system according to claim 5, characterized in that, The pre-trained fully convolutional neural network is trained using the YOLOv3 neural network.

8. The high-altitude wire erection tree obstacle recognition system according to claim 5, characterized in that It further includes: A position acquisition module, configured to acquire the geographical location of the target overhead line; A reporting module, configured to report the geographical location of the target overhead line and the overhead line images to the master station when the weighted overlap degree is greater than the preset overlap degree threshold.

Images