Intelligent power grid emergency rescue system based on unmanned aerial vehicle

Abstract

The invention belongs to the field of electric power emergency rescue, and discloses an unmanned aerial vehicle-based intelligent power grid emergency rescue system, which comprises an unmanned aerial vehicle, a control terminal, a route planning terminal and a loss evaluation terminal, the control terminal is used for controlling the unmanned aerial vehicle to fly to the sky of a to-be-shot area and acquiring an overall image of the to-be-shot area; the route planning terminal is used for generating an aerial photography route according to the overall image; the control terminal is used for controlling the unmanned aerial vehicle to shoot the to-be-shot area according to the aerial photography route to obtain a local image; and the loss evaluation terminal is used for identifying the local image and acquiring the damage condition of the power equipment in the local image. According to the invention, the proportion of the repeated areas in the obtained local images with adjacent positions can be effectively reduced, and the flight times of the unmanned aerial vehicle for completing shooting of the whole to-be-shot area can be effectively reduced.

Description

Technical Field

[0001] The present invention relates to the field of electric power emergency rescue, and in particular to an intelligent power grid emergency rescue system based on unmanned aerial vehicle (UAV). Background Art

[0002] After a natural disaster occurs, it is usually necessary to use drones to assess the damage to the power equipment of the power grid, so that the power equipment that is easy to repair can be repaired first, and power supply can be restored to more areas in a timely manner. In the prior art, when the personnel responsible for aerial photography arrive at the area where the power equipment is located (such as a substation), they usually take off directly and then manually control the drone to shoot the area based on experience. This easily leads to a large number of repeated areas in the images taken by the drone, so that the drone needs more flights to complete the shooting of the entire area, which reduces the efficiency of emergency repairs to the power equipment of the power grid. Summary of the invention

[0003] The purpose of the present invention is to disclose a smart grid emergency rescue system based on unmanned aerial vehicles to solve the technical problems raised in the background technology.

[0004] In order to achieve the above object, the present invention adopts the following technical solution:

[0005] The present invention provides a smart grid emergency rescue system based on a drone, comprising a drone, a control terminal, a route planning terminal and a loss assessment terminal;

[0006] The control terminal is used to control the drone to fly over the area to be photographed and obtain the overall image of the area to be photographed;

[0007] The route planning terminal is used to generate an aerial route based on the overall image;

[0008] The control terminal is used to control the drone to shoot the area to be shot according to the aerial photography route to obtain a local image;

[0009] The loss assessment terminal is used to identify the local image and obtain the damage status of the power equipment in the local image.

[0010] Preferably, controlling the drone to fly over the area to be photographed to obtain an overall image of the area to be photographed includes:

[0011] The control terminal receives the control instructions input by the emergency repair personnel responsible for aerial photography, and controls the UAV to fly over the area to be photographed according to the control instructions to obtain the overall image of the area to be photographed.

[0012] Preferably, the overall image of the area to be photographed includes the entire area to be photographed.

[0013] Preferably, generating an aerial route according to the overall image includes:

[0014] Segment an image containing only the area to be photographed from the overall image;

[0015] Partitioning the image containing only the area to be photographed to obtain multiple local areas;

[0016] The center of each local area is taken as an aerial photography point;

[0017] The shortest path generation algorithm is used to calculate all aerial photography points to obtain the aerial photography route.

[0018] Preferably, segmenting an image containing only the area to be photographed from the entire image includes:

[0019] Show the overall image to the repair personnel responsible for aerial photography;

[0020] Obtaining a sequence egU of pixel points belonging to the edge of the area to be photographed selected by the repair personnel responsible for aerial photography on the overall image;

[0021] A closed area is obtained on the whole image according to the sequence egU, and an image composed of all pixels in the closed area is used as an image containing only the area to be photographed.

[0022] Preferably, obtaining a closed area on the whole image according to the sequence egU includes:

[0023] In the overall image, the pixels in the sequence egU are processed as follows:

[0024] Step 1: Initialize the value of integer i to 1;

[0025] The second step is to connect the i-th pixel and the i+1-th pixel in the sequence egU;

[0026] The third step is to determine whether the value of i is less than or equal to NegU-1. If so, proceed to the fourth step. If not, proceed to the fifth step. NegU is the total number of pixels in egU.

[0027] Step 4: Add 1 to the value of i and go to step 2;

[0028] The fifth step is to connect the NegUth pixel and the first pixel in the sequence egU to obtain a closed area.

[0029] Preferably, the image containing only the area to be photographed is partitioned to obtain multiple local areas, including:

[0030] Obtaining an extended image corresponding to the image of the area to be photographed;

[0031] Divide the expanded image into multiple local regions of equal area;

[0032] Delete the local area that does not include the area to be captured.

[0033] Preferably, obtaining an extended image corresponding to the image of the area to be photographed includes:

[0034] Establishing a rectangular coordinate system in the image of the area to be photographed;

[0035] Obtaining the minimum value X1 and the maximum value X2 of the coordinates of the pixel points in the image of the shooting area on the X-axis of the rectangular coordinate system;

[0036] Obtaining the minimum value Y1 and the maximum value Y2 of the coordinates of the pixel points in the image of the shooting area on the Y axis of the rectangular coordinate system;

[0037] Then the pixels in the extended image satisfy:

[0038] X1≤x≤X2 and Y1≤y≤Y2;

[0039] x and y are the coordinates of the pixel on the X-axis and Y-axis respectively.

[0040] Preferably, according to the aerial photography route, the drone is controlled to photograph the area to be photographed to obtain a local image, including:

[0041] The drone is controlled to fly over each aerial photography point in turn according to the aerial photography route to take pictures, and a local image of each aerial photography point is obtained.

[0042] Preferably, identifying the local image and obtaining the damage condition of the electric power equipment in the local image includes:

[0043] Perform filtering on the local image to obtain an input image;

[0044] The input image is input into the pre-trained evaluation model for recognition to obtain the damage status of the power equipment in the local image.

[0045] Beneficial effects:

[0046] Compared with the prior art, the present invention can automatically generate an aerial photography route based on the area selected by the emergency repair personnel responsible for aerial photography in the overall image, thereby avoiding the emergency repair personnel responsible for aerial photography manually controlling the drone to shoot the area to be shot based on experience, and can effectively reduce the proportion of repeated areas in the obtained local images with adjacent positions, and effectively reduce the number of flights required for the drone to complete the shooting of the entire area to be shot. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 This is a schematic diagram of a smart grid emergency rescue system based on drones according to the present invention.

[0049] Figure 2 It is a schematic diagram of a method for partitioning an image containing only the area to be photographed according to the present invention. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the 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 the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0051] like Figure 1 In one embodiment shown, the present invention provides a smart grid emergency rescue system based on a drone, including a drone, a control terminal, a route planning terminal and a loss assessment terminal;

[0052] The control terminal is used to control the drone to fly over the area to be photographed and obtain the overall image of the area to be photographed;

[0053] The route planning terminal is used to generate an aerial route based on the overall image;

[0054] The control terminal is used to control the drone to shoot the area to be shot according to the aerial photography route to obtain a local image;

[0055] The loss assessment terminal is used to identify the local image and obtain the damage status of the power equipment in the local image.

[0056] Specifically, the area to be photographed refers to an area containing power equipment, such as an area where a substation is located.

[0057] Preferably, controlling the drone to fly over the area to be photographed to obtain an overall image of the area to be photographed includes:

[0058] The control terminal receives the control instructions input by the emergency repair personnel responsible for aerial photography, and controls the UAV to fly over the area to be photographed according to the control instructions to obtain the overall image of the area to be photographed.

[0059] Specifically, the control instructions include speed control instructions, direction control instructions, and shooting control instructions.

[0060] Speed ​​control instructions can be divided into acceleration instructions and deceleration instructions. The acceleration instructions and deceleration instructions can correspond to the value of the speed change, such as acceleration of 5m / s.

[0061] Direction control instructions include forward instructions, backward instructions, left turn instructions, right turn instructions, up instructions and down instructions, etc.

[0062] The shooting control command is used to control the drone's camera to shoot images.

[0063] Preferably, the overall image of the area to be photographed includes the entire area to be photographed.

[0064] When taking the overall image, the drone is controlled by the emergency repair personnel responsible for aerial photography and flies to a position above the center of the area to be photographed for shooting.

[0065] Preferably, generating an aerial route according to the overall image includes:

[0066] Segment an image containing only the area to be photographed from the overall image;

[0067] Partitioning the image containing only the area to be photographed to obtain multiple local areas;

[0068] The center of each local area is taken as an aerial photography point;

[0069] The shortest path generation algorithm is used to calculate all aerial photography points to obtain the aerial photography route.

[0070] In order to effectively identify the damage of power equipment and carry out emergency power repairs in time, the flight altitude cannot be too high, which requires the drone to shoot at multiple locations to achieve coverage of the entire area to be shot.

[0071] Specifically, the shortest path generation algorithm used may be the Dijkstra algorithm. The algorithm is applicable to directed and undirected graphs, can process positive weight edges, and selects the next node to be visited by maintaining a priority queue until the target node is found or all nodes are visited, thereby achieving the acquisition of the shortest path.

[0072] Preferably, segmenting an image containing only the area to be photographed from the entire image includes:

[0073] Show the overall image to the repair personnel responsible for aerial photography;

[0074] Obtaining a sequence egU of pixel points belonging to the edge of the area to be photographed selected by the repair personnel responsible for aerial photography on the overall image;

[0075] A closed area is obtained on the whole image according to the sequence egU, and an image composed of all pixels in the closed area is used as an image containing only the area to be photographed.

[0076] Specifically, the overall image can be displayed to the emergency repair personnel responsible for aerial photography through the display screen.

[0077] Specifically, the sequence egU of pixel points belonging to the edge of the area to be photographed selected by the repair personnel responsible for aerial photography on the overall image is obtained, including:

[0078] According to the order in which the repair personnel responsible for aerial photography click on the pixels, the pixel points belonging to the edge of the area to be photographed selected by the repair personnel on the overall image are stored in the sequence egU. The earlier the click order, the earlier the position of the pixel point in the sequence egU.

[0079] When obtaining the selected pixel point, the position of the mouse can be located and obtained. If the display screen is a touch screen, the corresponding pixel point is determined by obtaining the position where the repair personnel responsible for the aerial photography click on the screen.

[0080] Preferably, obtaining a closed area on the whole image according to the sequence egU includes:

[0081] In the overall image, the pixels in the sequence egU are processed as follows:

[0082] Step 1: Initialize the value of integer i to 1;

[0083] The second step is to connect the i-th pixel and the i+1-th pixel in the sequence egU;

[0084] The third step is to determine whether the value of i is less than or equal to NegU-1. If so, proceed to the fourth step. If not, proceed to the fifth step. NegU is the total number of pixels in egU.

[0085] Step 4: Add 1 to the value of i and go to step 2;

[0086] The fifth step is to connect the NegUth pixel and the first pixel in the sequence egU to obtain a closed area.

[0087] By connecting the pixels in egU in sequence, a closed area can be formed on the overall image.

[0088] Preferably, Figure 2 , partition the image containing only the area to be photographed to obtain multiple local areas, including:

[0089] Obtaining an extended image corresponding to the image of the area to be photographed;

[0090] Divide the expanded image into multiple local regions of equal area;

[0091] Delete the local area that does not include the area to be captured.

[0092] Since the area to be photographed may not be regular, the regular area is acquired by acquiring an extended image, which facilitates the subsequent acquisition of the local area. After acquiring the extended image, there may be pixels in the extended image that do not belong to the area to be photographed. Therefore, after acquiring the local area, the pixels in the local area may not be the pixels of the area to be photographed. Therefore, deleting such local areas can effectively reduce the number of aerial photography points while ensuring complete coverage of aerial photography.

[0093] Preferably, the extended image is divided into a plurality of local regions of equal area, including:

[0094] H represents the aerial photography height of the drone, h represents the horizontal viewing angle of the drone's camera during shooting, v represents the vertical viewing angle of the drone's camera during shooting, R represents the overlap rate of the two images taken at the center of two adjacent local areas with a height of H, and the width of the local area is:

[0095]

[0096] The length of the local area is:

[0097]

[0098] Convert width and length to the lengths widthp and lengthp in the image of the area to be captured;

[0099] The expanded image is divided into multiple local regions with an area of ​​widthp×lengthp.

[0100] The area of ​​the local area of ​​the present invention is not determined arbitrarily, but is obtained based on the aerial photography altitude, horizontal viewing angle and vertical vision. By setting a reasonable R, the overlap rate can be effectively controlled and the number of flights of the drone can be reduced. Because the battery life of a drone is generally short, it may not be possible to shoot all aerial photography points on a single charge.

[0101] Specifically, the value of R may be 10.

[0102] Specifically, converting width and length into the lengths widthp and length in the image of the area to be captured includes:

[0103] Len is used to represent the length of one pixel in the image of the area to be captured in real space;

[0104] Then widthp=width / len, lengthp=length / len.

[0105] Preferably, obtaining an extended image corresponding to the image of the area to be photographed includes:

[0106] Establishing a rectangular coordinate system in the image of the area to be photographed;

[0107] Obtaining the minimum value X1 and the maximum value X2 of the coordinates of the pixel points in the image of the shooting area on the X-axis of the rectangular coordinate system;

[0108] Obtaining the minimum value Y1 and the maximum value Y2 of the coordinates of the pixel points in the image of the shooting area on the Y axis of the rectangular coordinate system;

[0109] Then the pixels in the extended image satisfy:

[0110] X1≤x≤X2 and Y1≤y≤Y2;

[0111] x and y are the coordinates of the pixel on the X-axis and Y-axis respectively.

[0112] Preferably, according to the aerial photography route, the drone is controlled to photograph the area to be photographed to obtain a local image, including:

[0113] The drone is controlled to fly over each aerial photography point in turn according to the aerial photography route to take pictures, and a local image of each aerial photography point is obtained.

[0114] Specifically, the emergency repair personnel responsible for aerial photography can control the drone through the control terminal to fly over each aerial photography point in turn for shooting.

[0115] Preferably, identifying the local image and obtaining the damage condition of the electric power equipment in the local image includes:

[0116] Perform filtering on the local image to obtain an input image;

[0117] The input image is input into the pre-trained evaluation model for recognition to obtain the damage status of the power equipment in the local image.

[0118] Specifically, filtering is performed on the local image to obtain an input image, including:

[0119] For a pixel point k in the local image, the following formula is used to calculate the pixel value of k after filtering:

[0120]

[0121] gk represents the pixel value of k after k is filtered, hk represents the pixel value of k before k is filtered, ku represents the set of pixel points in a square with a side length of F centered on k, hz represents the pixel value of pixel point z, preh is a preset pixel value, ktz is the length of the straight line segment connecting k and z; pret is a preset length; numg is the number of pixels in ku that meet the detection conditions, aveh is the mean pixel value of the pixels in ku, and w is a preset weight.

[0122] The noise of the image input to the model can be reduced by filtering, and the accuracy of the recognition result can be improved. In the filtering process, in addition to considering the difference between k and the surrounding pixels in pixel value and the length of the straight line segment, the number of pixels that meet the detection conditions is also considered, so that when the probability that k belongs to the edge of the image is higher, the pixel value of k is larger, so that the clarity of the edge in the input image is less affected by the interference of filtering, and the accuracy of the recognition result is improved.

[0123] Specifically, the pixel points meeting the detection condition are the pixel points belonging to strong edges detected by using the Canny detection algorithm to detect the local image.

[0124] Specifically, the preset pixel value may be 180.

[0125] Specifically, the preset length may be half of the side length of the local image.

[0126] Specifically, the pixel value of the present invention may be a grayscale value.

[0127] Preferably, the pre-trained evaluation model is a ResNet model, and the trained ResNet model is used to classify the image of the area where each power equipment in the local image is located, to obtain the corresponding damage level of the image, and the damage level is used as the damage condition.

[0128] When training the ResNet model, images labeled with damage levels are used for training. The more serious the damage to the power equipment, the higher the damage level.

[0129] The training process is as follows:

[0130] (1) Data collection and annotation:

[0131] Image acquisition: Collect images containing damage conditions of power equipment (such as normal, slightly damaged, severely damaged, etc.).

[0132] Data annotation:

[0133] Label each image with its damage level (e.g. "Level 1", "Level 2", "Level 3", ...).

[0134] You can use tools such as LabelImg or directly organize them into folder format: one folder for each category.

[0135] (2) Data division:

[0136] Divide the data into training, validation, and test sets:

[0137] Training set: used for model training (70%-80%).

[0138] Validation set: used to tune hyperparameters (10%-15%).

[0139] Test set: Evaluate model performance (10%-15%).

[0140] (3) Data enhancement:

[0141] Use data augmentation techniques to improve the robustness of the model:

[0142] Geometric transformations: rotation, flipping, cropping.

[0143] Color adjustment: brightness, contrast, saturation.

[0144] Noise Addition: Simulates image degradation.

[0145] You can use tools such as torchvision.transforms or Albumentations.

[0146] (4) Select ResNet version:

[0147] ResNet-18 or ResNet-34: Lightweight and suitable for small data sets or edge computing devices.

[0148] ResNet-50 or ResNet-101: Stronger feature extraction capabilities, suitable for large data sets.

[0149] (5) Pre-training weights:

[0150] Load ImageNet pre-trained weights (reduces training time and improves initial performance).

[0151] (6) Modify the classification header

[0152] Modify the last fully connected layer (FC) of ResNet to adapt to the target classification task.

[0153] (7) Use the cross-entropy loss function CrossEntropyLoss to train ResNet.

[0154] The present invention can automatically generate an aerial photography route based on the area selected by the emergency repair personnel responsible for aerial photography in the overall image, thereby avoiding the emergency repair personnel responsible for aerial photography manually controlling the drone to shoot the area to be shot based on experience, and can effectively reduce the proportion of repeated areas in the obtained local images with adjacent positions, and effectively reduce the number of flights required for the drone to complete the shooting of the entire area to be shot.

[0155] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A smart grid emergency rescue system based on drones, characterized in that: Includes drones, control terminals, route planning terminals, and damage assessment terminals; The control terminal is used to control the drone to fly over the area to be photographed and obtain the overall image of the area to be photographed; The route planning terminal is used to generate an aerial route based on the overall image; The control terminal is used to control the drone to shoot the area to be shot according to the aerial photography route to obtain a local image; The loss assessment terminal is used to identify the local image and obtain the damage status of the power equipment in the local image.

2. The smart grid emergency rescue system based on drone according to claim 1 is characterized in that: Control the drone to fly over the area to be photographed and obtain the overall image of the area to be photographed, including: The control terminal receives the control instructions input by the emergency repair personnel responsible for aerial photography, and controls the UAV to fly over the area to be photographed according to the control instructions to obtain the overall image of the area to be photographed.

3. The smart grid emergency rescue system based on drone according to claim 1 is characterized in that: The overall image of the area to be photographed includes the entire area to be photographed.

4. The smart grid emergency rescue system based on drone according to claim 1, characterized in that: Generate an aerial route based on the overall image, including: Segment an image containing only the area to be photographed from the overall image; Partitioning the image containing only the area to be photographed to obtain multiple local areas; The center of each local area is taken as an aerial photography point; The shortest path generation algorithm is used to calculate all aerial photography points to obtain the aerial photography route.

5. The smart grid emergency rescue system based on drone according to claim 4 is characterized in that: Segment the image containing only the area to be captured from the overall image, including: Show the overall image to the repair personnel responsible for aerial photography; Obtaining a sequence egU of pixel points belonging to the edge of the area to be photographed selected by the repair personnel responsible for aerial photography on the overall image; A closed area is obtained on the whole image according to the sequence egU, and an image composed of all pixels in the closed area is used as an image containing only the area to be photographed.

6. The smart grid emergency rescue system based on drone according to claim 5, characterized in that: Obtain the closed area on the whole image according to the sequence egU, including: In the overall image, the pixels in the sequence egU are processed as follows: Step 1: Initialize the value of integer i to 1; The second step is to connect the i-th pixel and the i+1-th pixel in the sequence egU; The third step is to determine whether the value of i is less than or equal to NegU-1. If so, proceed to the fourth step. If not, proceed to the fifth step. NegU is the total number of pixels in egU. Step 4: Add 1 to the value of i and go to step 2; The fifth step is to connect the NegUth pixel and the first pixel in the sequence egU to obtain a closed area.

7. The smart grid emergency rescue system based on drone according to claim 4, characterized in that: Partition the image that only contains the area to be captured to obtain multiple local areas, including: Obtaining an extended image corresponding to the image of the area to be photographed; Divide the expanded image into multiple local regions of equal area; Delete the local area that does not include the area to be captured.

8. The smart grid emergency rescue system based on drone according to claim 7, characterized in that: Obtaining an extended image corresponding to the image of the area to be photographed, including: Establishing a rectangular coordinate system in the image of the area to be photographed; Obtaining the minimum value X1 and the maximum value X2 of the coordinates of the pixel points in the image of the shooting area on the X-axis of the rectangular coordinate system; Obtaining the minimum value Y1 and the maximum value Y2 of the coordinates of the pixel points in the image of the shooting area on the Y axis of the rectangular coordinate system; Then the pixels in the extended image satisfy: X1≤x≤X2 and Y1≤y≤Y2; x and y are the coordinates of the pixel on the X-axis and Y-axis respectively.

9. The smart grid emergency rescue system based on drone according to claim 4, characterized in that: According to the aerial photography route, the drone is controlled to shoot the area to be photographed to obtain local images, including: The drone is controlled to fly over each aerial photography point in turn according to the aerial photography route to take pictures, and a local image of each aerial photography point is obtained.

10. The smart grid emergency rescue system based on drone according to claim 1, characterized in that: Identify the local image and obtain the damage status of the power equipment in the local image, including: Perform filtering on the local image to obtain an input image; The input image is input into the pre-trained evaluation model for recognition to obtain the damage status of the power equipment in the local image.

Images