CN114022761A - A kind of satellite remote sensing image transmission line tower detection and positioning method and device - Google Patents

Abstract

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

Description

A kind of satellite remote sensing image transmission line tower detection and positioning method and device

technical field

The invention relates to the technical field of operation and maintenance of power transmission lines, in particular to a method and a device for detecting and positioning towers of power transmission lines with satellite remote sensing images.

Background technique

Electric power is the main component of my country's energy structure and an important support for economic development. Transmission line towers are facilities for carrying and guiding high-voltage overhead lines in the power grid. With the popularization of electricity and the increasing complexity of the power grid, the number of transmission line towers is increasing year by year, and has the characteristics of wide distribution, large span and complex surrounding terrain. If relying on the traditional manual field measurement to obtain the position of the transmission line tower, although it has high accuracy, it needs to invest a lot of time and manpower, and the cost is high and the efficiency is low. Remote sensing images contain rich spatial information and are of great significance in resource surveys, environmental monitoring, geological disaster surveys, regional analysis and construction planning. If the remote sensing image visual interpretation method is used to obtain the position of the transmission line tower, it still requires high labor and time costs, and the accuracy depends on the professional technical level and meticulous degree of the interpreter. Due to the influence of visual fatigue, continuous Human visual interpretation work can significantly reduce the efficiency and accuracy of manual detection.

SUMMARY OF THE INVENTION

(1) Technical problems to be solved

In view of the above shortcomings and deficiencies of the prior art, the present invention provides a satellite remote sensing image transmission line tower detection and positioning method and device, which solves the problem of obtaining the position of the transmission line tower by manual field measurement. Although it has high accuracy, it needs to Invest a lot of time and manpower, high cost, low efficiency and technical problems.

(2) Technical solutions

In order to achieve the above-mentioned purpose, the main technical scheme adopted in the present invention includes:

In a first aspect, an embodiment of the present invention provides a satellite remote sensing image transmission line tower detection and positioning method, including:

The satellite remote sensing images of the study area to be predicted are cropped according to a preset degree of overlap, and the position information of each image in the satellite remote sensing image of the study area to be predicted after the cropping is numbered and recorded, and the obtained images are cropped according to the preset degree of overlap. post image data;

Inputting the image data trimmed according to the preset overlap degree into a pre-trained target detection model, and acquiring the prediction information of each picture in the image data trimmed according to the preset overlap degree;

The prediction information of each picture includes the normalized coordinates and confidence of the center point of the transmission line tower;

The trained target detection model is a model that uses the training set and test set of satellite remote sensing images to train and test the yolov5 target detection model in advance.

Preferably, the method further includes:

Analyze and process the prediction information of each picture in the image data cropped according to the preset overlap degree, and determine the number of independent and complete transmission line towers and transmission line towers in the image data cropped according to a certain degree of overlap , the coordinates of the transmission line tower.

Preferably, the prediction information of each picture in the image data trimmed according to the preset overlap degree is analyzed and processed, and the independent and complete transmission line towers and transmission lines in the image data trimmed according to the certain overlap degree are determined. The number of towers and the coordinates of the towers of the transmission line, including:

The prediction information of each picture includes the normalized coordinates and confidence of the center point of the transmission line tower;

Perform the first coordinate transformation process on the normalized coordinates of the center point of the transmission line tower in each picture to obtain the coordinates in the corresponding picture;

Based on the position information of each image in the satellite remote sensing image of the study area to be predicted, and the coordinates in each picture, a second coordinate conversion process is performed to obtain the coordinates in the satellite remote sensing image of the study area to be predicted;

Respectively judge whether the confidence of the center point of the transmission line tower in the prediction information of each picture is lower than the preset threshold, obtain the first judgment result, and use the first judgment result as lower than the set threshold as the yolov5 target detection model The transmission line towers of the detection target are deleted, and the transmission line towers whose first judgment result is greater than or equal to the set threshold are reserved as the detection target of the yolov5 target detection model;

Determine whether there is overlap in the prediction frame 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 according to the second judgment result;

The six-parameter transformation model of gdal is used to convert the coordinates in the satellite remote sensing image of the study area corresponding to the transmission line tower of the final detection target of the yolov5 target detection model to the corresponding geographic coordinates, and output the transmission line tower in the image area. The number and geographic coordinates of the transmission line towers.

Preferably, according to the second judgment result, determine the transmission line tower as the final detection target of the yolov5 target detection model, which specifically includes:

If the second judgment result is that there is no overlap in the prediction frame in the yolov5 target detection model, the transmission line tower in the prediction frame in the yolov5 target detection model is used as the transmission line tower of the final detection target of the yolov5 target detection model;

If the second judgment result is that there is overlap in the prediction frame in the yolov5 target detection model, then judge whether the prediction frame with the overlapping area is located in the same picture;

If the prediction frame containing the overlapping area is in the same picture, all the transmission line towers in the prediction frame are used as the transmission line towers of the final detection target of the yolov5 target detection model;

If the prediction frame containing the overlapping area is not in the same picture, the transmission line tower with the highest confidence in the prediction frame is used as the transmission line tower of the final detection target of the yolov5 target detection model.

preferably,

Wherein, formula (1) is used to perform the first coordinate conversion processing;

The formula (1) is:

Wherein, X _min is the abscissa of the upper left corner of the picture in the image data cropped according to the preset overlap degree;

Y _min is the ordinate of the upper left corner of the picture in the image data cropped according to the preset overlap degree;

X _max is the abscissa of the lower right corner of the picture in the image data cropped according to the preset overlap degree;

Y _max is the ordinate of the lower right corner of the picture in the image data cropped according to the preset overlap degree;

W is the width of each picture in the image data cropped according to the preset overlapping degree;

H is the height of each picture in the cropped image data according to the preset overlap degree;

x is the abscissa of the normalized center of the prediction frame;

y is the normalized center ordinate of the prediction frame;

w is the normalized width of the prediction frame;

h is the normalized height of the prediction frame;

Adopt formula (2) to carry out the second coordinate conversion processing to obtain the coordinates in the satellite remote sensing image of the study area to be predicted;

Among them, formula (2) is:

Among them, x ₁ is the abscissa of the upper left corner in the satellite remote sensing image of the study area to be predicted;

y ₁ is the ordinate of the upper left corner point in the satellite remote sensing image of the study area to be predicted;

x ₂ is the abscissa of the lower right corner point in the satellite remote sensing image of the study area to be predicted;

y ₂ is the ordinate of the lower right corner point in the satellite remote sensing image of the study area to be predicted;

i is the line number of the position of the picture in the satellite remote sensing image of the study area to be predicted;

j is the column number of the position of the picture in the satellite remote sensing image of the study area to be predicted;

q is a preset overlap degree.

Preferably, before acquiring the image data cropped according to the preset overlap degree, the method further includes:

Preprocess the satellite remote sensing image data of the study area, and obtain the preprocessed satellite remote sensing image data of the study area;

The preprocessing is to perform initial trimming of the area containing the transmission line towers in the entire study area in the satellite remote sensing image data of the study area, and discard the area that does not contain the transmission line towers;

According to the pre-processed satellite remote sensing image data of the research area, crop it according to the preset format, and obtain the first set of pictures;

The first picture set includes a plurality of satellite remote sensing image pictures of the research area that meet a preset format;

Perform image processing on each of the first set of pictures that satisfies the preset format of the satellite remote sensing image picture of the study area to obtain a second set of pictures;

The image processing is to use a grayscale world method, an automatic white balance method, or a histogram equalization method to process each of the first image sets in the study area satellite remote sensing image picture that meets the preset format;

Taking 60% of the pictures in the second picture set as the training set, 20% of the pictures as the validation set, and 20% of the pictures as the test set;

Use the training set, the test set and the verification set to train and test and verify the yolov5 target detection model to obtain the trained yolov5 target detection model;

The trained yolov5 target detection model is the model of the yolov5 target detection model when the loss function of the yolov5 target detection model converges after training with the aid of the training set.

preferably,

The implementation method of the gray-scale world mode specifically includes:

Use formula (A) to adjust the B _n -band component C(B _n ′) of the remote sensing image channel;

Formula (A) is:

in,

in,

为波段n所有像元灰度值的平均值；

is the average value of the gray value of all pixels in band n;

C(B _n ) is the gray value of each pixel of band n.

preferably,

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

The remote sensing image is transferred to the Lab color space, the color cast is detected on the image, the color cast value is calculated, and the color is corrected and then transferred to the RGB color space.

preferably,

The frame loss function in the loss function of the yolov5 target detection model is CIoU;

The CIoU is:

where IoU is the intersection ratio of the real frame and the predicted frame;

c is the diagonal length of the prediction frame;

ρ ² (b, b ^gt ) is used to measure the Euclidean distance between the predicted frame and the center point of the real frame;

α is the weight coefficient;

v is the similarity used to measure the aspect ratio.

On the other hand, this embodiment also provides a satellite remote sensing image transmission line tower detection and positioning method device, the device includes:

at least one processor; and

At least one memory connected in communication with 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 as described above Video transmission line tower detection and positioning method.

(3) Beneficial effects

The beneficial effects of the present invention are as follows: in the present invention, a method and device for detecting and locating a power transmission line tower in a satellite remote sensing image, because a pre-trained target detection model is used, and each of the image data clipped according to the pre-set overlap degree is obtained. The prediction information of each picture, and further, based on the prediction information of each picture in the image data that has been cropped according to the preset overlap degree, analyze and process to determine the independent and complete transmission line tower in the image data that has been cropped according to a certain degree of overlap. , The number of transmission line towers, and the coordinates of transmission line towers. Compared with the existing technology, the detection and positioning of transmission line towers can be realized. Compared with the traditional methods of manual field measurement and visual interpretation, large-scale simultaneous detection is possible. , improve efficiency and reduce costs.

On the other hand, a method and device for detecting and positioning transmission line towers in satellite remote sensing images of the present invention use grayscale world algorithm, automatic white balance algorithm or histogram equalization algorithm for processing satellite remote sensing images in the study area. Compared with the traditional deep learning target detection algorithm, it significantly improves the precision and recall rate, enhances the generalization ability of the model, and has higher robustness.

Description of drawings

Fig. 1 is a flow chart of a method for detecting and positioning a transmission line tower in a satellite remote sensing image according to the present invention;

FIG. 2 is a schematic diagram of a method for detecting and positioning a transmission line tower in a satellite remote sensing image according to an embodiment of the present invention;

3 is a schematic flowchart of analysis processing in an embodiment of the present invention;

FIG. 4 is a schematic diagram of the detection result of the detection and positioning method of a kind of satellite remote sensing image transmission line tower detection and positioning method of the present invention;

FIG. 5 is a schematic diagram of detection details in the detection result of the detection and positioning method of a transmission line tower in a satellite remote sensing image of the present invention.

Detailed ways

In order to better explain the present invention and facilitate understanding, the present invention will be described in detail below with reference to the accompanying drawings and through specific embodiments.

For better understanding of the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present invention will be more clearly and thoroughly understood, and will fully convey the scope of the present invention to those skilled in the art.

The present invention is a deep learning-based satellite remote sensing image transmission line tower detection and positioning method.

The invention innovatively uses the grayscale world algorithm, the automatic white balance algorithm and the histogram equalization algorithm for the remote sensing image, so as to eliminate the influence of ambient light in the remote sensing image, obtain the image of the original scene, reduce the influence of the light source on the satellite camera sensor, and simulate the Due to the constancy of the human visual system, in order to retain the spatial detail information and spatial coordinate information of the remote sensing image, a CPAS detection and positioning method is designed to complete the detection of the transmission line tower in the remote sensing image, and output the position information of the transmission line tower. Compared with the traditional methods of manual field measurement and visual interpretation, it enables large-scale simultaneous detection, improves efficiency and reduces costs. Compared with the traditional deep learning target detection algorithm, it improves the precision rate and recall rate, enhances the generalization ability of the model, and has higher robustness.

Referring to FIG. 1 , the present embodiment provides a method for detecting and positioning a transmission line tower in a satellite remote sensing image, including:

The satellite remote sensing images of the study area to be predicted are cropped according to a preset degree of overlap, and the position information of each image in the satellite remote sensing image of the study area to be predicted after the cropping is numbered and recorded, and the obtained images are cropped according to the preset degree of overlap. post image data.

The image data trimmed according to the preset overlap degree is input into a pre-trained target detection model, and the prediction information of each picture in the image data trimmed according to the preset overlap degree is acquired.

The prediction information of each picture includes the normalized coordinates and confidence level of the center point of the transmission line tower.

The trained target detection model is a model that uses the training set and test set of satellite remote sensing images to train and test the yolov5 target detection model in advance.

In the practical application of this embodiment, the method further includes:

Analyze and process the prediction information of each picture in the image data cropped according to the preset overlap degree, and determine the number of independent and complete transmission line towers and transmission line towers in the image data cropped according to a certain degree of overlap , the coordinates of the transmission line tower.

Referring to FIG. 3 , in the practical application of this embodiment, the prediction information of each picture in the image data trimmed according to the preset overlap degree is analyzed and processed, and the image data trimmed according to a certain overlap degree is determined. The independent complete transmission line towers, the number of transmission line towers, and the coordinates of the transmission line towers in the list include:

The prediction information of each picture includes the normalized coordinates and confidence level of the center point of the transmission line tower.

Perform a first coordinate transformation process on the normalized coordinates of the center point of the transmission line tower in each picture to obtain the coordinates in the corresponding picture.

Based on the position information of each image in the satellite remote sensing image of the study area to be predicted, and the coordinates in each picture, a second coordinate conversion process is performed to obtain the coordinates in the satellite remote sensing image of the study area to be predicted.

Respectively judge whether the confidence of the center point of the transmission line tower in the prediction information of each picture is lower than the preset threshold, obtain the first judgment result, and use the first judgment result as lower than the set threshold as the yolov5 target detection model The transmission line towers of the detection target are deleted, and the transmission line towers whose first judgment result is greater than or equal to the set threshold are reserved as the detection target of the yolov5 target detection model.

Determine whether there is overlap in the prediction frame of 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 according to the second judgment result.

The six-parameter transformation model of gdal is used to convert the coordinates in the satellite remote sensing image of the study area corresponding to the transmission line tower of the final detection target of the yolov5 target detection model to the corresponding geographic coordinates, and output the transmission line tower in the image area. The number and geographic coordinates of the transmission line towers.

In the specific application of this embodiment, the transmission line tower of the final detection target of the yolov5 target detection model is displayed in the input satellite remote sensing image of the study area to be predicted, as shown in Figure 4 and Figure 5, and the transmission line tower in the image area is output. The number and coordinates of transmission line towers are shown in Table 1.

Table 1 The number of transmission line towers in the image area and the coordinate information of transmission line towers

In the practical application of this embodiment, according to the second judgment result, the transmission line tower as the final detection target of the yolov5 target detection model is determined, which specifically includes:

If the second judgment result is that there is no overlap in the prediction frame in the yolov5 target detection model, the transmission line tower in the prediction frame in the yolov5 target detection model is used as the transmission line tower of the final detection target of the yolov5 target detection model.

If the second judgment result is that the prediction frames in the yolov5 target detection model overlap, it is determined whether the prediction frames with the overlapping area are located in the same picture.

If the prediction frame containing the overlapping area is in the same picture, all the transmission line towers in the prediction frame are used as the transmission line towers of the final detection target of the yolov5 target detection model.

If the prediction frame containing the overlapping area is not in the same picture, the transmission line tower with the highest confidence in the prediction frame is used as the transmission line tower of the final detection target of the yolov5 target detection model.

In the practical application of this embodiment, the formula (1) is used to perform the first coordinate conversion processing.

The formula (1) is:

Wherein, X _min is the abscissa of the upper left corner of the picture in the image data cropped according to the preset overlap degree.

Y _min is the ordinate of the upper left corner of the picture in the image data cropped according to the preset overlap degree.

Y _max is the abscissa of the lower right corner of the picture in the image data cropped according to the preset overlap degree.

X _max is the ordinate of the lower right corner of the picture in the image data cropped according to the preset overlap degree.

W is the width of each picture in the image data cropped according to the preset overlap degree.

H is the height of each picture in the cropped image data according to the preset overlap degree.

x is the normalized center abscissa of the prediction box.

y is the normalized center ordinate of the prediction frame.

w is the normalized width of the prediction box.

h is the normalized height of the prediction box.

Use formula (2) to perform the second coordinate transformation process to obtain the coordinates in the satellite remote sensing image of the study area to be predicted.

Among them, formula (2) is:

Among them, x ₁ is the abscissa of the upper left corner in the satellite remote sensing image of the study area to be predicted.

y ₁ is the ordinate of the upper left corner in the satellite remote sensing image of the study area to be predicted.

x ₂ is the abscissa of the lower right corner point in the satellite remote sensing image of the study area to be predicted.

y ₂ is the ordinate of the lower right corner point in the satellite remote sensing image of the study area to be predicted.

i is the line number of the position of the picture in the satellite remote sensing image of the study area to be predicted.

j is the column number of the position of the picture in the satellite remote sensing image of the study area to be predicted.

q is a preset overlap degree.

Referring to FIG. 2, in the practical application of this embodiment, before acquiring the image data cropped according to the preset overlap degree, the method further includes:

The satellite remote sensing image data of the study area is preprocessed, and the preprocessed satellite remote sensing image data of the study area is obtained.

The preprocessing is to initially trim the area containing the transmission line tower in the satellite remote sensing image data of the study area, and discard the area that does not contain the transmission line tower.

In this example, the remote sensing images in the study area are downloaded. In order to reduce unnecessary work when creating data sets, the typical area containing transmission line towers is cut out in the entire study area, and some areas such as oceans that do not contain transmission line towers are discarded.

The preprocessed satellite remote sensing image data of the study area is cropped according to a preset format to obtain a first set of images.

The first picture set includes a plurality of satellite remote sensing image pictures of the research area that meet a preset format.

Perform image processing on each satellite remote sensing image picture of the study area that meets a preset format in the first picture set to obtain a second picture set.

The image processing is to process each satellite remote sensing image picture of the research area in the first set of images that meets a preset format by using a grayscale world method, an automatic white balance method, or a histogram equalization method.

In the specific application of this embodiment, the gray-scale world algorithm, the automatic white balance algorithm and the histogram equalization process are randomly performed for each satellite remote sensing image picture of the research area that meets the preset format in the first picture set. .

60% of the pictures in the second picture set are used as the training set, 20% of the pictures are used as the validation set, and 20% of the pictures are used as the test set.

The training set, the test set and the verification set are used to train, test and verify the yolov5 target detection model to obtain the trained yolov5 target detection model.

The trained yolov5 target detection model is the model of the yolov5 target detection model when the loss function of the yolov5 target detection model converges after training with the aid of the training set.

In the practical application of this embodiment, the method for realizing the grayscale world mode specifically includes:

Use formula (A) to adjust the B _n -band component C(B _n ') of the remote sensing image channel.

Formula (A) is:

in,

in,

为波段n所有像元灰度值的平均值。

is the average value of the grayscale values of all pixels in band n.

C(B _n ) is the gray value of each pixel of band n.

再计算B₁，B₂，B₃三通道的增益系数

根据Von Kries对角模型，根据遥感影像中像元灰度值C调整其B₁，B₂，B₃波段分量：

In the specific application, the realization process of the gray _- scale world algorithm is as follows: calculate the average value _of each channel pixel _of the remote sensing image.

Then calculate the gain coefficients of the three channels B ₁ , B ₂ , and B ₃

According to the Von Kries diagonal model, adjust its B ₁ , B ₂ , B ₃ band components according to the pixel gray value C in the remote sensing image:

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

The remote sensing image is transferred to the Lab color space, the color cast is detected on the image, the color cast value is calculated, and the color is corrected and then transferred to the RGB color space.

In the practical application of this embodiment, the frame loss function in the loss function of the yolov5 target detection model is CIoU.

The CIoU is:

where IoU is the intersection ratio of the ground-truth box and the predicted box.

C is the diagonal length of the prediction box.

p ² (b,b ^gt ) is used to measure the Euclidean distance between the predicted frame and the center point of the true frame.

α is the weight coefficient.

v is the similarity used to measure the aspect ratio.

In the specific application of this embodiment, CIoU integrates factors such as the aspect ratio of the prediction frame, the distance between the target and the anchor, the degree of overlap, and the scale, so that the regression of the target frame is more stable, which is conducive to speeding up the convergence. speed.

The essence of deep learning is to build a machine learning architecture model with multiple hidden layers, and to obtain a large amount of more representative feature information through training on large-scale data. Thereby, the samples are classified and predicted, and the accuracy of classification and prediction is improved. This process is to achieve the purpose of feature learning by means of deep learning models. With the continuous improvement of deep learning theory and the iterative update of detection algorithms, neural networks have powerful feature extraction capabilities, and target detection methods based on deep learning have obvious advantages over traditional target detection methods.

In this embodiment, the deep learning method is used to realize the detection and positioning of transmission line towers. Compared with traditional methods of manual field measurement and visual interpretation, simultaneous detection in a large area is possible, efficiency is improved, and cost is reduced.

In this embodiment, the grayscale world algorithm, the automatic white balance algorithm and the histogram equalization algorithm are used for the processing of satellite remote sensing images in the study area. Compared with the traditional deep learning target detection algorithm, it significantly improves the precision and recall rate, enhances the generalization ability of the model, and has higher robustness.

Remote sensing images have larger size and contain coordinate information. Bringing it directly into the model prediction will lose a lot of spatial detail information and coordinate information. After testing, the method proposed by the present invention can effectively retain the spatial detail information and coordinate information of the remote sensing image.

Since the system described in the above embodiments of the present invention is the system used to implement the methods in the above embodiments of the present invention, based on the methods described in the above embodiments of the present invention, those skilled in the art can understand the specific details of the system/device. The structure and deformation will not be repeated here. All systems used in the methods of the above embodiments of the present invention belong to the scope of protection of the present invention.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may 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.) having computer-usable program code embodied therein.

The present 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 flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions.

It should be noted that, in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not preclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In the claims enumerating several means, several of these means can be embodied by one and the same item of hardware. The words first, second, third, etc. are used for convenience only and do not imply any order. These words can be understood as part of the part name.

In addition, it should be noted that in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiments", "examples", "specific examples" or "some examples", etc., are Indicates that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, those skilled in the art may combine and combine the different embodiments or examples described in this specification, as well as the features of the different embodiments or examples, without conflicting each other.

Although the preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments will occur to those skilled in the art after learning the basic inventive concepts. Therefore, the claims should be construed to include the preferred embodiment and all changes and modifications that fall within the scope of the present invention.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit and scope of the invention. Thus, provided that these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention should also include these modifications and variations.

Claims (10)

1. A detection and positioning method for a power transmission line tower based on a satellite remote sensing image is characterized by comprising the following steps:

cutting the satellite remote sensing images of the research area to be predicted according to a preset overlapping degree, numbering and recording the position information of each cut image in the satellite remote sensing images of the research area to be predicted, and acquiring image data cut according to the preset overlapping degree;

inputting the image data cut according to the preset overlapping degree into a pre-trained target detection model, and acquiring the prediction information of each picture in the image data cut according to the preset overlapping degree;

the prediction information of each picture comprises normalized coordinates and confidence degrees of the center point of the power transmission line tower;

the trained target detection model is a model obtained by training and testing a yolov5 target detection model in advance by adopting a training set and a testing set of satellite remote sensing images.

2. The method of claim 1, further comprising:

analyzing and processing the prediction information of each picture in the image data cut according to the preset overlapping degree, and determining independent complete transmission line towers, the number of the transmission line towers and the coordinates of the transmission line towers in the image data cut according to the certain overlapping degree.

3. The method according to claim 2, wherein the analyzing the prediction information of each picture in the image data cut according to the preset overlapping degree to determine the number of independent complete transmission line towers, the number of transmission line towers and the coordinates of the transmission line towers in the image data cut according to the certain overlapping degree specifically comprises:

the prediction information of each picture comprises normalized coordinates and confidence degrees of the center point of the power transmission line tower;

performing first coordinate conversion processing on the normalized coordinates of the central point of the power transmission line tower in each picture to obtain the coordinates in the corresponding picture;

performing second coordinate conversion processing on the position information of each image in the satellite remote sensing image of the research area to be predicted and the coordinates in each picture to obtain the coordinates in the satellite remote sensing image of the research area to be predicted;

respectively judging whether the confidence of the center point of the power transmission line tower in the prediction information of each picture is lower than a preset threshold value, acquiring a first judgment result, deleting the power transmission line tower which is lower than the preset threshold value and is used as a detection target of the yolov5 target detection model, and keeping the power transmission line tower which is larger than or equal to the preset threshold value and is used as a detection target of the yolov5 target detection model;

judging whether the prediction frames in the yolov5 target detection model are overlapped or not, obtaining a second judgment result, and determining the power transmission line tower serving as the final detection target of the yolov5 target detection model according to the second judgment result;

and converting coordinates in the satellite remote sensing image of the research area to be predicted, which corresponds to the power transmission line pole tower of the final detection target of the yolov5 target detection model, into corresponding geographic coordinates by adopting a gdal six-parameter conversion model, and outputting the number of the power transmission line poles and the geographic coordinates of the power transmission line poles and towers in an image area.

4. The method according to claim 3, wherein determining the transmission line tower as the final detection target of the yolov5 target detection model according to the second determination result specifically comprises:

if the second judgment result is that no overlap exists in the prediction frames in the yolov5 target detection model, taking the power transmission line tower in the prediction frames in the yolov5 target detection model as the power transmission line tower of the final detection target of the yolov5 target detection model;

if the second judgment result is that the prediction frames in the yolov5 target detection model are overlapped, judging whether the prediction frames with the overlapped areas are positioned in the same picture;

if the prediction frames containing the overlapping areas are in the same picture, all the transmission line towers in the prediction frames are used as the transmission line towers of the final detection target of the yolov5 target detection model;

and if the prediction frames containing the overlapping areas are not in the same picture, all the power transmission line towers with the maximum confidence in the prediction frames are used as the power transmission line towers of the final detection target of the yolov5 target detection model.

5. The method of claim 4,

wherein the first coordinate conversion processing is performed by using a formula (1);

the formula (1) is:

wherein, X_minThe horizontal coordinate of the upper left corner point of the picture in the image data cut according to the preset overlapping degree;

Y_minthe vertical coordinate of the upper left corner point of the picture in the image data cut according to the preset overlapping degree;

X_maxthe horizontal coordinate of the lower right corner point of the picture in the image data cut according to the preset overlapping degree;

Y_maxthe vertical coordinate of the lower right corner point of the picture in the image data cut according to the preset overlapping degree;

w is the width of each picture in the image data cut according to the preset overlapping degree;

h is the height of each picture in the image data cut according to the preset overlapping degree;

x is the normalized central abscissa of the prediction box;

y is the normalized central ordinate of the prediction frame;

w is the normalized width of the prediction box;

h is the normalized height of the prediction box;

adopting a formula (2) to perform second coordinate conversion processing to obtain coordinates in the satellite remote sensing image of the research area to be predicted;

wherein, the formula (2) is:

wherein x is₁The horizontal coordinate of the upper left corner point in a satellite remote sensing image of a research area to be predicted;

y₁for satellite remote in the research area to be predictedSensing the vertical coordinate of the upper left corner point in the image;

x₂the horizontal coordinate of the lower right corner point in the satellite remote sensing image of the research area to be predicted;

y₂the vertical coordinate of the lower right corner in the satellite remote sensing image of the research area to be predicted;

i is a row number of the position of the picture in the satellite remote sensing image of the research area to be predicted;

j is the column number of the position of the picture in the satellite remote sensing image of the research area to be predicted;

q is a preset overlap.

6. The method of claim 5, wherein prior to acquiring the image data cropped to the predetermined degree of overlap, the method further comprises:

preprocessing the satellite remote sensing image data of the research area to obtain preprocessed satellite remote sensing image data of the research area;

the preprocessing comprises the steps of carrying out primary cutting on an area containing the transmission line tower in the whole research area in the satellite remote sensing image data of the research area, and abandoning the area containing no transmission line tower;

cutting the preprocessed satellite remote sensing image data of the research area according to a preset format to obtain a first picture set;

the first picture set comprises a plurality of research area satellite remote sensing image pictures meeting a preset format;

performing image processing on each research area satellite remote sensing image picture meeting a preset format in the first picture set to obtain a second picture set;

the image processing is to process each research area satellite remote sensing image picture meeting a preset format in a first picture set by adopting a gray world mode or an automatic white balance mode or a histogram equalization mode;

taking 60% of pictures in the second picture set as a training set, 20% of pictures as a verification set and 20% of pictures as a test set;

training, testing and verifying the yolov5 target detection model by adopting the training set, the testing set and the verifying set to obtain a trained yolov5 target detection model;

the trained yolov5 target inspection model is the model of the yolov5 target inspection model when the loss function of the yolov5 target inspection model after training with the training set converges.

7. The method of claim 6,

the gray world mode implementation method specifically comprises the following steps:

adjusting remote sensing image channel B by formula (A)_nBand component C (B)_n′)；

The formula (A) is:

wherein,

wherein,

the average value of all pixel gray values of the wave band n is obtained;

C(B_n) Is the gray value of each pixel of the band n.

8. The method of claim 7,

the implementation method of the automatic white balance algorithm specifically comprises the following steps:

and transferring the remote sensing image to a Lab color space, performing color cast detection on the image, calculating a color cast value, and transferring to an RGB color space after performing color correction.

9. The method of claim 8,

a frame loss function in the loss functions of the yolov5 target detection model is CIoU;

the CIoU is as follows:

wherein IoU is the intersection ratio of the real box and the predicted box;

c is the diagonal length of the prediction frame;

ρ²(b，b^gt) The Euclidean distance between the central points of the prediction frame and the real frame is measured;

alpha is a weight coefficient;

v is a measure of the similarity of aspect ratios.

10. A detection and positioning method device for a power transmission line tower based on a satellite remote sensing image is characterized by comprising the following steps:

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 calls the program instructions to perform the method according to any one of claims 1 to 9.

Images