Satellite remote sensing image transmission line tower detection positioning method and device

Abstract

The application relates to a satellite remote sensing image transmission line tower detection positioning method and device, wherein the method comprises the following steps: cutting a satellite remote sensing image of a to-be-predicted research area according to a preset overlap degree, numbering and recording position information of each image in the satellite remote sensing image of the to-be-predicted research area after cutting, and obtaining image data after cutting according to the preset overlap degree; inputting the image data after cutting according to the preset overlap degree into a pre-trained target detection model, and obtaining prediction information of each picture in the image data after cutting according to the preset overlap degree; the prediction information of each picture comprises a transmission line tower center point normalized coordinate and a confidence degree; and the trained target detection model is a model obtained by training and testing a yolov5 target detection model by using a training set and a test set of satellite remote sensing images in advance.

Description

A method and device for detecting and locating power transmission line towers using satellite remote sensing images Technical Field

[0001] This invention relates to the field of power transmission line operation and maintenance technology, and in particular to a method and device for detecting and locating power transmission line towers using satellite remote sensing images. Background Technology

[0002] Electricity is a major component of my country's energy structure and a crucial pillar of economic development. Transmission line towers are facilities in the power grid that carry and guide high-voltage overhead lines. With the widespread adoption of electricity and the increasing complexity of the power grid, the number of transmission line towers is increasing year by year, characterized by wide distribution, large spans, and complex surrounding terrain. While traditional manual on-site measurement for obtaining the location of transmission line towers offers high accuracy, it requires a significant investment of time and manpower, resulting in high costs and low efficiency. Remote sensing imagery contains rich spatial information and is of great significance in resource surveys, environmental monitoring, geological disaster investigation, regional analysis, and construction planning. However, using visual interpretation of remote sensing imagery to obtain the location of transmission line towers still requires substantial manpower and time costs, and accuracy depends heavily on the interpreter's professional skills and attention to detail. Furthermore, visual fatigue significantly reduces the efficiency and accuracy of manual visual interpretation work. Summary of the Invention

[0003] (a) Technical problems to be solved

[0004] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method and device for detecting and locating transmission line towers using satellite remote sensing images. It solves the technical problems of high cost and low efficiency caused by the traditional method of obtaining the location of transmission line towers through manual on-site measurement, which, although highly accurate, requires a large investment of time and manpower.

[0005] (II) Technical Solution

[0006] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0007] In a first aspect, embodiments of the present invention provide a method for detecting and locating transmission line towers using satellite remote sensing imagery, comprising:

[0008] The satellite remote sensing images of the study area to be predicted are cropped according to a pre-set overlap degree, and the position information of each cropped image in the satellite remote sensing images of the study area to be predicted is recorded by number, so as to obtain the image data after cropping according to the pre-set overlap degree.

[0009] The image data cropped according to a pre-set overlap degree is input into a pre-trained target detection model, and prediction information for each image in the image data cropped according to a pre-set overlap degree is obtained.

[0010] The prediction information for each image includes the normalized coordinates of the center point of the transmission line tower and the confidence level.

[0011] The trained target detection model is the model that was trained and tested using a training set and a test set of satellite remote sensing images.

[0012] Preferably, the method further includes:

[0013] The predicted information of each image in the image data after being cropped according to a pre-set overlap degree is analyzed and processed to determine the independent and complete transmission line towers, the number of transmission line towers, and the coordinates of the transmission line towers in the image data after being cropped according to a certain overlap degree.

[0014] Preferably, the predicted information of each image in the image data after being cropped according to a pre-set overlap degree is analyzed and processed to determine the independent and complete transmission line towers, the number of transmission line towers, and the coordinates of the transmission line towers in the image data after being cropped according to a certain overlap degree. Specifically, this includes:

[0015] The prediction information for each image includes the normalized coordinates of the center point of the transmission line tower and the confidence level.

[0016] The normalized coordinates of the center point of the transmission line tower in each image are processed by the first coordinate transformation to obtain the corresponding coordinates in the image;

[0017] Based on the location information of each image in the satellite remote sensing image of the study area to be predicted, and the coordinates in each image, a second coordinate transformation process is performed to obtain the coordinates in the satellite remote sensing image of the study area to be predicted.

[0018] For each image, determine whether the confidence level of the center point of the transmission line tower in the prediction information is lower than a preset threshold, obtain the first judgment result, and delete the transmission line towers whose first judgment result is lower than the preset threshold as targets of the YOLOv5 target detection model, and retain the transmission line towers whose first judgment result is greater than or equal to the preset threshold as targets of the YOLOv5 target detection model.

[0019] Determine whether there is overlap in the predicted bounding boxes in the YOLOv5 target detection model, obtain the second judgment result, and determine the transmission line tower as the final detection target of the YOLOv5 target detection model based on the second judgment result.

[0020] The GDAL six-parameter transformation model is used to convert the coordinates of the transmission line towers corresponding to the final detected targets in the YOLOv5 target detection model in the satellite remote sensing image of the study area to be predicted into the corresponding geographic coordinates, and outputs the number of transmission line towers in the image area and the geographic coordinates of the transmission line towers.

[0021] Preferably, based on the second judgment result, the transmission line towers that are determined as the final detection targets of the YOLOv5 target detection model are included, specifically:

[0022] If the second judgment result is that there is no overlap in the prediction box in the YOLOv5 target detection model, then the transmission line tower in the prediction box in the YOLOv5 target detection model is taken as the transmission line tower of the final detection target of the YOLOv5 target detection model.

[0023] If the second judgment result is that there is overlap in the predicted bounding boxes in the YOLOv5 object detection model, then it is determined whether the predicted bounding boxes with overlapping areas are located in the same image;

[0024] If the prediction boxes containing overlapping areas are in the same image, then all transmission line towers in the prediction boxes are considered as transmission line towers as the final detection targets of the YOLOv5 target detection model.

[0025] If the predicted bounding boxes containing overlapping areas are not in the same image, then the transmission line towers with the highest confidence in the predicted bounding boxes will be used as the final transmission line towers of the YOLOv5 target detection model.

[0026] Preferred,

[0027] The first coordinate transformation is performed using formula (1);

[0028] The formula (1) is:

[0029]

[0030]

[0031] Among them, X min The x-coordinate of the top-left corner of the image in the image data after cropping according to a preset overlap.

[0032] Y min The ordinate of the top left corner of the image in the image data after cropping according to a preset overlap.

[0033] X max The x-coordinate of the bottom right corner of the image in the image data after cropping according to a preset overlap.

[0034] Y max The ordinate of the bottom right corner of the image in the image data after cropping according to a preset overlap.

[0035] W represents the width of each image in the image data after cropping according to a pre-defined overlap.

[0036] H represents the height of each image in the image data after cropping according to a pre-defined overlap.

[0037] x is the normalized x-coordinate of the predicted box center;

[0038] y is the normalized center ordinate of the predicted bounding box;

[0039] w is the normalized width of the prediction box;

[0040] h is the normalized height of the predicted bounding box;

[0041] The coordinates of the study area to be predicted in the satellite remote sensing image are obtained by performing the second coordinate transformation process using formula (2);

[0042] Formula (2) is:

[0043]

[0044]

[0045] Where x1 is the x-coordinate of the upper left corner point in the satellite remote sensing image of the study area to be predicted;

[0046] y1 is the ordinate of the top left corner point in the satellite remote sensing image of the study area to be predicted;

[0047] x2 is the x-coordinate of the lower right corner point in the satellite remote sensing image of the study area to be predicted;

[0048] y2 is the ordinate of the lower right corner point in the satellite remote sensing image of the study area to be predicted;

[0049] i represents the row number of the image's location in the satellite remote sensing image of the study area to be predicted;

[0050] j is the column number of the image's location in the satellite remote sensing image of the study area to be predicted;

[0051] q is the pre-defined degree of overlap.

[0052] Preferably, before acquiring the image data cropped according to a preset overlap, the method further includes:

[0053] Preprocessing is performed on the satellite remote sensing image data of the study area to obtain the preprocessed satellite remote sensing image data of the study area;

[0054] The preprocessing involves initially cropping the areas containing power transmission line towers in the entire study area from the satellite remote sensing image data of the study area, discarding the areas that do not contain power transmission line towers;

[0055] The preprocessed satellite remote sensing image data of the study area is cropped according to a pre-set format to obtain the first image set;

[0056] The first image set includes multiple satellite remote sensing images of the study area that meet a pre-defined format;

[0057] For each satellite remote sensing image of the study area in the first image set that meets the preset format, image processing is performed to obtain the second image set;

[0058] The image processing involves processing each satellite remote sensing image of the study area in the first image set that meets a pre-defined format using either grayscale world mode, automatic white balance mode, or histogram equalization mode.

[0059] Use 60% of the images in the second image set as the training set, 20% as the validation set, and 20% as the test set.

[0060] The YOLOv5 object detection model is trained, tested, and validated using the training set, test set, and validation set to obtain the trained YOLOv5 object detection model.

[0061] The trained YOLOv5 object detection model is the model when the loss function of the YOLOv5 object detection model converges after training with the help of the training set.

[0062] Preferred,

[0063] The specific methods for implementing the grayscale world approach include:

[0064] Adjust remote sensing image channel B using formula (A) n Band component C(B) n ′);

[0065] Formula (A) is:

[0066]

[0067] in,

[0068] in,

[0069] This is the average gray value of all pixels in band n;

[0070] C(B n ) represents the gray value of each pixel in band n.

[0071] Preferred,

[0072] The implementation method of the automatic white balance algorithm specifically includes:

[0073] The remote sensing image is converted to the Lab color space, color cast is detected, the color cast value is calculated, and after color correction, it is converted to the RGB color space.

[0074] Preferred,

[0075] The bounding box loss function in the loss function of the YOLOv5 object detection model is CIoU;

[0076] The CIoU is:

[0077]

[0078]

[0079] Where IoU is the intersection-union ratio of the ground truth bounding box and the predicted bounding box;

[0080] c is the length of the diagonal of the prediction box;

[0081] ρ 2 (b,b gt () is the Euclidean distance used to measure the center point of the predicted bounding box and the ground truth bounding box;

[0082] α is the weighting coefficient;

[0083] v is used to measure the similarity of aspect ratio.

[0084] On the other hand, this embodiment also provides a device for detecting and locating transmission line towers using satellite remote sensing images, the device comprising:

[0085] At least one processor; and

[0086] At least one memory communicatively connected to the processor, wherein the memory stores program instructions executable by the processor, and the processor invokes the program instructions to execute the satellite remote sensing image transmission line tower detection and positioning method as described above.

[0087] (III) Beneficial Effects

[0088] The beneficial effects of this invention are as follows: The satellite remote sensing image transmission line tower detection and positioning method and device of this invention, by employing a pre-trained target detection model and acquiring the prediction information of each image in the image data after being cropped according to a pre-set overlap degree, further analyzes and processes the prediction information of each image in the image data after being cropped according to a pre-set overlap degree to determine the independent and complete transmission line towers, the number of transmission line towers, and the coordinates of the transmission line towers in the image data after being cropped according to a certain overlap degree, relative to the prior art, thereby realizing the detection and positioning of transmission line towers. Compared with traditional methods of manual on-site measurement and visual interpretation, this makes large-scale simultaneous detection possible, improves efficiency, and reduces costs.

[0089] On the other hand, the satellite remote sensing image transmission line tower detection and positioning method and apparatus of the present invention applies the gray-world algorithm, automatic white balance algorithm, or histogram equalization algorithm to the processing of satellite remote sensing images of the study area. Compared with traditional deep learning target detection algorithms, it significantly improves precision and recall, enhances the model's generalization ability, and has higher robustness. Attached Figure Description

[0090] Figure 1 is a flowchart of a satellite remote sensing image transmission line tower detection and positioning method according to the present invention.

[0091] Figure 2 is a schematic diagram of the satellite remote sensing image transmission line tower detection and positioning method in an embodiment of the present invention;

[0092] Figure 3 is a schematic diagram of the analysis and processing flow in an embodiment of the present invention;

[0093] Figure 4 is a schematic diagram of the detection results using a satellite remote sensing image transmission line tower detection and positioning method of the present invention;

[0094] Figure 5 is a schematic diagram of the detection details in the detection results of a satellite remote sensing image transmission line tower detection and positioning method according to the present invention. Detailed Implementation

[0095] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0096] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0097] This invention provides a method for detecting and locating power transmission line towers based on deep learning satellite remote sensing images.

[0098] This invention innovatively employs a grayscale world algorithm, an automatic white balance algorithm, and a histogram equalization algorithm on remote sensing images to eliminate the influence of ambient light, obtain images of the original scene, reduce the impact of light sources on satellite camera sensors, and simulate the constancy of the human visual system. To preserve the spatial detail and coordinate information of the remote sensing images, a CPAS detection and localization method is designed to detect power transmission line towers in the images and output their location information. Compared with traditional methods such as manual field measurement and visual interpretation, this invention enables simultaneous detection over a large area, improving efficiency and reducing costs. Compared with traditional deep learning object detection algorithms, it improves precision and recall, enhances the model's generalization ability, and exhibits higher robustness.

[0099] Referring to Figure 1, this embodiment provides a method for detecting and locating transmission line towers using satellite remote sensing imagery, including:

[0100] The satellite remote sensing images of the study area to be predicted are cropped according to a pre-set overlap degree, and the position information of each cropped image in the satellite remote sensing images of the study area to be predicted is recorded by number, so as to obtain the image data after cropping according to the pre-set overlap degree.

[0101] The image data cropped according to a pre-set overlap is input into a pre-trained target detection model, and prediction information for each image in the image data cropped according to the pre-set overlap is obtained.

[0102] The prediction information for each image includes the normalized coordinates of the center point of the transmission line tower and the confidence level.

[0103] The trained target detection model is the model that was trained and tested using a training set and a test set of satellite remote sensing images.

[0104] In practical applications of this embodiment, the method further includes:

[0105] The predicted information of each image in the image data after being cropped according to a pre-set overlap degree is analyzed and processed to determine the independent and complete transmission line towers, the number of transmission line towers, and the coordinates of the transmission line towers in the image data after being cropped according to a certain overlap degree.

[0106] Referring to Figure 3, in the practical application of this embodiment, the predicted information of each image in the image data after being cropped according to a pre-set overlap degree is analyzed and processed to determine the independent and complete transmission line towers, the number of transmission line towers, and the coordinates of the transmission line towers in the image data after being cropped according to a certain overlap degree. Specifically, this includes:

[0107] The prediction information for each image includes the normalized coordinates of the center point of the transmission line tower and the confidence level.

[0108] The normalized coordinates of the center point of the transmission line tower in each image are processed by the first coordinate transformation to obtain the corresponding coordinates in the image.

[0109] Based on the location information of each image in the satellite remote sensing image of the study area to be predicted, and the coordinates in each image, a second coordinate transformation process is performed to obtain the coordinates in the satellite remote sensing image of the study area to be predicted.

[0110] For each image, determine whether the confidence level of the center point of the transmission line tower in the predicted information is lower than a preset threshold, obtain the first judgment result, and delete the transmission line towers whose first judgment result is lower than the preset threshold as targets of the YOLOv5 target detection model, and retain the transmission line towers whose first judgment result is greater than or equal to the preset threshold as targets of the YOLOv5 target detection model.

[0111] Determine whether there is overlap in the predicted bounding boxes in the YOLOv5 target detection model, obtain the second judgment result, and determine the transmission line tower as the final detection target of the YOLOv5 target detection model based on the second judgment result.

[0112] The GDAL six-parameter transformation model is used to convert the coordinates of the transmission line towers corresponding to the final detected targets in the YOLOv5 target detection model in the satellite remote sensing image of the study area to be predicted into the corresponding geographic coordinates, and outputs the number of transmission line towers in the image area and the geographic coordinates of the transmission line towers.

[0113] In this specific application, the transmission line towers of the final detected target of the YOLOv5 target detection model are displayed in the input satellite remote sensing image of the study area to be predicted, as shown in Figures 4 and 5. The number of transmission line towers and the coordinates of the transmission line towers in the image area are output as shown in Table 1.

[0114] Table 1 shows the number of transmission line towers and their coordinates within the image area.

[0115]

[0116] In the practical application of this embodiment, based on the second judgment result, the transmission line towers that are determined as the final detection targets of the YOLOv5 target detection model are specifically included:

[0117] If the second judgment result is that there is no overlap in the prediction box of the YOLOv5 target detection model, then the transmission line tower in the prediction box of the YOLOv5 target detection model is taken as the transmission line tower of the final detection target of the YOLOv5 target detection model.

[0118] If the second judgment result is that there is overlap in the predicted bounding boxes in the YOLOv5 object detection model, then it is determined whether the predicted bounding boxes with overlapping areas are located in the same image.

[0119] If the prediction boxes containing overlapping areas are in the same image, then all transmission line towers in the prediction boxes will be considered as transmission line towers as the final detection targets of the YOLOv5 target detection model.

[0120] If the predicted bounding boxes containing overlapping areas are not in the same image, then the transmission line towers with the highest confidence in the predicted bounding boxes will be used as the final transmission line towers of the YOLOv5 target detection model.

[0121] In the practical application of this embodiment, the first coordinate transformation is performed using formula (1).

[0122] The formula (1) is:

[0123]

[0124]

[0125] Among them, X min The x-coordinate of the top-left corner of the image in the image data after cropping according to a pre-defined overlap.

[0126] Y min The vertical coordinate of the top-left corner of the image in the image data after cropping according to a pre-defined overlap.

[0127] Y max The x-coordinate of the bottom right corner of the image in the image data after cropping according to a pre-defined overlap.

[0128] X max The vertical coordinate of the bottom right corner of the image in the image data after cropping according to a pre-defined overlap.

[0129] W represents the width of each image in the image data after cropping according to a pre-defined overlap.

[0130] H represents the height of each image in the image data after cropping according to a pre-defined overlap.

[0131] x is the normalized x-coordinate of the predicted bounding box center.

[0132] y is the normalized center ordinate of the prediction box.

[0133] w is the normalized width of the prediction box.

[0134] h is the normalized height of the prediction box.

[0135] The coordinates of the study area in the satellite remote sensing image to be predicted are obtained by performing the second coordinate transformation process using formula (2).

[0136] Formula (2) is:

[0137]

[0138]

[0139] Where x1 is the x-coordinate of the upper left corner point in the satellite remote sensing image of the study area to be predicted.

[0140] y1 is the ordinate of the top left corner point in the satellite remote sensing image of the study area to be predicted.

[0141] x2 is the x-coordinate of the lower right corner point in the satellite remote sensing image of the study area to be predicted.

[0142] y2 is the ordinate of the lower right corner point in the satellite remote sensing image of the study area to be predicted.

[0143] i represents the row number of the image's location within the satellite remote sensing imagery of the study area to be predicted.

[0144] j is the column number of the image's location in the satellite remote sensing imagery of the study area to be predicted.

[0145] q is the pre-defined degree of overlap.

[0146] Referring to Figure 2, in the practical application of this embodiment, before acquiring the image data cropped according to a preset overlap, the method further includes:

[0147] Preprocessing is performed on the satellite remote sensing image data of the study area to obtain the preprocessed satellite remote sensing image data of the study area.

[0148] The preprocessing involves initially cropping the areas containing power transmission line towers within the entire study area from the satellite remote sensing image data of the study area, discarding the areas that do not contain power transmission line towers.

[0149] In this embodiment, remote sensing images of the study area are downloaded. To reduce unnecessary work when creating the dataset, typical areas containing power transmission line towers in the entire study area are cropped, and some areas such as the ocean that do not contain power transmission line towers are discarded.

[0150] The preprocessed satellite remote sensing image data of the study area is cropped according to a pre-set format to obtain the first image set.

[0151] The first image set includes multiple satellite remote sensing images of the study area that meet a pre-defined format.

[0152] For each satellite remote sensing image of the study area in the first image set that meets the pre-defined format, image processing is performed to obtain the second image set.

[0153] The image processing involves processing each satellite remote sensing image of the study area in the first image set that meets a pre-defined format using either grayscale world mode, automatic white balance mode, or histogram equalization mode.

[0154] In the specific application of this embodiment, each satellite remote sensing image of the study area in the first image set that meets the preset format is randomly subjected to grayscale world algorithm, automatic white balance algorithm and histogram equalization processing.

[0155] Use 60% of the images in the second image set as the training set, 20% as the validation set, and 20% as the test set.

[0156] The YOLOv5 object detection model is trained, tested, and validated using the training set, test set, and validation set to obtain the trained YOLOv5 object detection model.

[0157] The trained YOLOv5 object detection model is the model when the loss function of the YOLOv5 object detection model converges after training with the help of the training set.

[0158] In practical applications of this embodiment, the grayscale world implementation method specifically includes:

[0159] Adjust remote sensing image channel B using formula (A) n Band component C(B) n ′).

[0160] Formula (A) is:

[0161]

[0162] in,

[0163] in,

[0164] It is the average gray value of all pixels in band n.

[0165] C(B n ) represents the gray value of each pixel in band n.

[0166] In practical applications, the implementation process of the grayscale world algorithm is as follows: calculate the average pixel value of each channel of the remote sensing image. Taking the three channels B1, B2, and B3 as an example... Next, calculate the gain coefficients of channels B1, B2, and B3. Based on the Von Kries diagonal model, adjust the B1, B2, and B3 band components according to the pixel grayscale value C in the remote sensing image:

[0167] In the practical application of this embodiment, the implementation method of the automatic white balance algorithm specifically includes:

[0168] The remote sensing image is converted to the Lab color space, color cast is detected, the color cast value is calculated, and after color correction, it is converted to the RGB color space.

[0169] In the practical application of this embodiment, the bounding box loss function in the loss function of the YOLOv5 object detection model is CIoU.

[0170] The CIoU is:

[0171]

[0172]

[0173] Where IoU is the intersection-union ratio of the ground truth bounding box and the predicted bounding box.

[0174] C represents the diagonal length of the prediction box.

[0175] p 2 (b,b gt ) is used to measure the Euclidean distance between the center point of the predicted bounding box and the center point of the ground truth bounding box.

[0176] α is the weighting coefficient.

[0177] v is used to measure the similarity of aspect ratio.

[0178] In the specific application of this embodiment, CIoU integrates factors such as the aspect ratio of the predicted bounding box, the distance between the target and the anchor, the overlap, and the scale, making the regression of the target bounding box more stable and helping to accelerate the convergence speed.

[0179] Deep learning essentially involves building machine learning architecture models with multiple hidden layers, training them on large-scale data to obtain a wealth of more representative feature information. This allows for the classification and prediction of samples, improving the accuracy of classification and prediction. This process achieves feature learning through deep learning models. With the continuous improvement of deep learning theory and the iterative updates of detection algorithms, neural networks possess powerful feature extraction capabilities, and deep learning-based object detection methods have significant advantages over traditional object detection methods.

[0180] This embodiment uses deep learning to detect and locate transmission line towers. Compared with traditional methods such as manual on-site measurement and visual interpretation, it makes large-scale simultaneous detection possible, improves efficiency, and reduces costs.

[0181] In this embodiment, the gray-world algorithm, automatic white balance algorithm, and histogram equalization algorithm are used to process satellite remote sensing images of the study area. Compared with traditional deep learning object detection algorithms, this significantly improves precision and recall, enhances the model's generalization ability, and exhibits higher robustness.

[0182] Remote sensing images have larger dimensions and contain coordinate information. Directly inputting them into model predictions results in the loss of significant spatial detail and coordinate information. Testing has shown that the method proposed in this invention effectively preserves the spatial detail and coordinate information of remote sensing images.

[0183] Since the system described in the above embodiments of the present invention is a system used to implement the methods of the above embodiments of the present invention, those skilled in the art can understand the specific structure and modifications of the system / device based on the methods described in the above embodiments of the present invention, and therefore will not be repeated here. All systems used in the methods of the above embodiments of the present invention fall within the scope of protection of the present invention.

[0184] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0185] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0186] It should be noted that any reference numerals placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In claims that enumerate several means, several of these means may be embodied by the same hardware. The use of the terms first, second, third, etc., is merely for convenience of expression and does not indicate any order. These terms can be understood as part of the component names.

[0187] Furthermore, it should be noted that in the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0188] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the claims should be interpreted to include both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0189] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, then this invention should also include these modifications and variations.

Claims

1. A method for detecting and locating transmission line towers using satellite remote sensing imagery, characterized in that, include: The satellite remote sensing images of the study area to be predicted are cropped according to a pre-set overlap ratio, and the position information of each cropped image in the satellite remote sensing images of the study area to be predicted is recorded. Image data cropped according to the pre-set overlap ratio is obtained. This cropped image data is then input into a pre-trained target detection model, and prediction information for each image in the cropped image data is obtained. The prediction information for each image includes the normalized coordinates of the center point of the transmission line tower and the confidence score. The trained target detection model is used to detect YOLOv5 targets using a pre-set training and test set of satellite remote sensing images. The detection model is trained and tested; the method further includes: analyzing and processing the prediction information of each image in the image data cropped according to a pre-set overlap degree to determine the independent and complete transmission line towers, the number of transmission line towers, and the coordinates of the transmission line towers in the image data cropped according to a certain overlap degree; specifically, the prediction information of each image includes the transmission line towers... The system calculates the normalized coordinates and confidence score of the transmission line tower center point; it then performs a first coordinate transformation on the normalized coordinates of the transmission line tower center point in each image to obtain the corresponding coordinates in the image; based on the location information of each image in the satellite remote sensing image of the study area to be predicted and the coordinates in each image, it performs a second coordinate transformation to obtain the coordinates in the satellite remote sensing image of the study area to be predicted; it then determines whether the confidence score of the transmission line tower center point in the prediction information of each image is lower than a preset threshold, obtains a first judgment result, and deletes the transmission line towers whose first judgment result is lower than the preset threshold as targets detected by the YOLOv5 target detection model. The transmission line towers whose first judgment result is greater than or equal to the set threshold are retained as targets detected by the YOLOv5 target detection model; it is determined whether there is overlap in the prediction box in the YOLOv5 target detection model, and a second judgment result is obtained. Based on the second judgment result, the transmission line towers that are the final targets detected by the YOLOv5 target detection model are determined; the coordinates in the satellite remote sensing image of the study area corresponding to the transmission line towers of the final targets detected by the YOLOv5 target detection model are converted into corresponding geographic coordinates using the GDAL six-parameter transformation model, and the number of transmission line towers in the image area and the geographic coordinates of the transmission line towers are output.Based on the second judgment result, the transmission line towers are determined as the final detection targets of the YOLOv5 target detection model. Specifically, this includes: if the second judgment result indicates that there is no overlap in the predicted bounding boxes of the YOLOv5 target detection model, then the transmission line towers in the predicted bounding boxes of the YOLOv5 target detection model are taken as the final detection targets of the YOLOv5 target detection model; if the second judgment result indicates that there is overlap in the predicted bounding boxes of the YOLOv5 target detection model, then it is determined whether the predicted bounding boxes with overlapping areas are located in the same image; if the predicted bounding boxes with overlapping areas are in the same image, then all transmission line towers in the predicted bounding boxes are taken as the final detection targets of the YOLOv5 target detection model; if the predicted bounding boxes with overlapping areas are not in the same image, then the transmission line towers with the highest confidence in the predicted bounding boxes are taken as the final detection targets of the YOLOv5 target detection model.

2. The method according to claim 1, characterized in that, in, The first coordinate transformation is performed using formula (1); formula (1) is: ； ;in, The x-coordinate of the top-left corner of the image in the image data after cropping according to a preset overlap. The ordinate of the top left corner of the image in the image data after cropping according to a preset overlap. The x-coordinate of the bottom right corner of the image in the image data after cropping according to a preset overlap. W is the ordinate of the lower right corner of the image in the image data after cropping according to a pre-set overlap; H is the width of each image in the image data after cropping according to a pre-set overlap; x is the normalized center x-coordinate of the prediction box; y is the normalized center y-coordinate of the prediction box; w is the normalized width of the prediction box; h is the normalized height of the prediction box; the second coordinate transformation is performed using formula (2) to obtain the coordinates in the satellite remote sensing image of the study area to be predicted; where formula (2) is: ； Where x1 is the x-coordinate of the upper left corner point in the satellite remote sensing image of the study area to be predicted; y1 is the y-coordinate of the upper left corner point in the satellite remote sensing image of the study area to be predicted; x2 is the x-coordinate of the lower right corner point in the satellite remote sensing image of the study area to be predicted; y2 is the y-coordinate of the lower right corner point in the satellite remote sensing image of the study area to be predicted; i is the row number of the image's position in the satellite remote sensing image of the study area to be predicted; j is the column number of the image's position in the satellite remote sensing image of the study area to be predicted; and q is the pre-set overlap degree.

3. The method according to claim 2, characterized in that, Before acquiring the image data cropped according to a pre-set overlap, the method further includes: preprocessing the satellite remote sensing image data of the study area to acquire preprocessed satellite remote sensing image data of the study area; the preprocessing involves initially cropping the areas containing transmission line towers within the entire study area of ​​the satellite remote sensing image data of the study area, discarding areas that do not contain transmission line towers; cropping the preprocessed satellite remote sensing image data of the study area according to a pre-set format to acquire a first image set; the first image set includes multiple satellite remote sensing image images of the study area that meet the pre-set format; and performing image processing on each satellite remote sensing image of the study area that meets the pre-set format in the first image set. Obtain a second image set; the image processing involves processing each satellite remote sensing image of the study area in the first image set that meets a pre-defined format using grayscale world mode, automatic white balance mode, or histogram equalization mode; use 60% of the images in the second image set as the training set, 20% as the validation set, and 20% as the test set; use the training set, test set, and validation set to train, test, and validate the YOLOv5 object detection model to obtain the trained YOLOv5 object detection model; the trained YOLOv5 object detection model is the model when the loss function of the YOLOv5 object detection model converges after training with the help of the training set.

4. The method according to claim 3, characterized in that, The grayscale world mode implementation method specifically includes: adjusting remote sensing image channel B using formula (A). n Band components Formula (A) is: ;in, ;in, ； This is the average gray value of all pixels in band n; Let be the gray value of each pixel in band n.

5. The method according to claim 4, characterized in that, The implementation method of the automatic white balance algorithm specifically includes: converting the remote sensing image to the Lab color space, performing color cast detection on the image, calculating the color cast value, performing color correction, and then converting it to the RGB color space.

6. The method according to claim 5, characterized in that, The bounding box loss function in the loss function of the YOLOv5 object detection model is CIoU; CIoU is: ； ;in This represents the intersection-union ratio (IoU) of the ground truth bounding boxes and the predicted bounding boxes. This is the length of the diagonal of the predicted bounding box; This is used to measure the Euclidean distance between the center points of the predicted bounding box and the ground truth bounding box. These are the weighting coefficients; This is used to measure the similarity of aspect ratios.

7. A method and device for detecting and locating transmission line towers using satellite remote sensing imagery, characterized in that, The device includes: at least one processor; and at least one memory communicatively connected to the processor, wherein the memory stores program instructions executable by the processor, and the processor invokes the program instructions to execute the satellite remote sensing image transmission line tower detection and positioning method as described in any one of claims 1 to 6.

Images