CN116843909B - Power line extraction method and device, storage medium and computer equipment - Google Patents
Power line extraction method and device, storage medium and computer equipment Download PDFInfo
- Publication number
- CN116843909B CN116843909B CN202310540202.4A CN202310540202A CN116843909B CN 116843909 B CN116843909 B CN 116843909B CN 202310540202 A CN202310540202 A CN 202310540202A CN 116843909 B CN116843909 B CN 116843909B
- Authority
- CN
- China
- Prior art keywords
- image
- tower
- transmission line
- power
- electric
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 109
- 238000003860 storage Methods 0.000 title claims abstract description 18
- 230000005540 biological transmission Effects 0.000 claims abstract description 206
- 238000003708 edge detection Methods 0.000 claims abstract description 85
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 82
- 238000001514 detection method Methods 0.000 claims abstract description 72
- 238000000034 method Methods 0.000 claims abstract description 60
- 239000012212 insulator Substances 0.000 claims description 40
- 239000011159 matrix material Substances 0.000 claims description 35
- 230000009466 transformation Effects 0.000 claims description 32
- 238000012549 training Methods 0.000 claims description 27
- 238000004364 calculation method Methods 0.000 claims description 26
- 230000004913 activation Effects 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 16
- 238000007781 pre-processing Methods 0.000 claims description 13
- 238000012795 verification Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 8
- 230000003321 amplification Effects 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 6
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 4
- 230000000007 visual effect Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 description 39
- 238000007689 inspection Methods 0.000 description 11
- 238000013527 convolutional neural network Methods 0.000 description 10
- 230000008901 benefit Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000000844 transformation Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000003064 k means clustering Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000003121 nonmonotonic effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 235000019504 cigarettes Nutrition 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012850 discrimination method Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 229910052704 radon Inorganic materials 0.000 description 1
- SYUHGPGVQRZVTB-UHFFFAOYSA-N radon atom Chemical compound [Rn] SYUHGPGVQRZVTB-UHFFFAOYSA-N 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000000779 smoke Substances 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 238000003892 spreading Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/17—Terrestrial scenes taken from planes or by drones
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Remote Sensing (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a power line extraction method and device, a storage medium and computer equipment, wherein the method comprises the following steps: performing electric tower coordinate prediction on the standardized image of the power transmission line corresponding to the power transmission line to be detected according to the coordinate prediction model to obtain electric tower prediction coordinates; performing edge detection on the standardized image of the power transmission line to obtain an edge detection result image and calculating an offset angle of the electric tower; determining a tower positioning frame in a standardized image of the power transmission line according to the tower prediction coordinates, and determining an effective power line extraction area according to the tower positioning frame and the tower offset angle; and carrying out linear detection on the effective power line extraction area, determining a straight line perpendicular to the electric tower offset angle in the detected straight line as a candidate power line, voting the candidate power line according to a preset power line template, and determining the power line of the power transmission line to be detected according to the voting result. By combining an unsupervised algorithm and a supervised algorithm, the accuracy of power line extraction is improved.
Description
Technical Field
The present disclosure relates to the field of power line inspection technologies, and in particular, to a power line extraction method and apparatus, a storage medium, and a computer device.
Background
With the continuous expansion of the scale of the national high-voltage transmission line, the environmental scope covered by the transmission network is more and more complex, and the safe and stable operation of the transmission line is more and more difficult to ensure. Traditional transmission line inspection mainly relies on artifical ground inspection and manned helicopter inspection, however high tension transmission line distributes in the rare mountain area of people's cigarette more, and its environment is different and traffic is inconvenient, and through the mode of manual inspection, timeliness and accuracy are not high, and there is the security risk, and the mode of inspection through the helicopter is then economic nature relatively poor. Compared with the traditional transmission line inspection method, the unmanned machine such as an unmanned plane is adopted for inspection, and the method has the advantages of low cost, convenience in operation and the like. Unmanned machines such as unmanned aerial vehicle can be used for carrying high-resolution aerial camera or other shooting machines for shoot aerial image, carry out the mode of patrolling and examining high-voltage transmission line through aerial image, with low costs and factor of safety is little.
Based on the relevant characteristics of the aerial image, the methods of power line detection of researchers at home and abroad aiming at the aerial image are mainly divided into two types, namely an unsupervised method and a supervised method. The non-supervision method is based on the linear characteristics of the power line, the power line is considered as a continuous straight line, the power line detection is realized through a classical straight line detection method, such as Radon transformation, hough transformation, LSD straight line detection and FLD linear discrimination methods, under a complex background, the non-supervision method is easy to be interfered by objects with linear characteristics such as roads and trees, noise interference exists in aerial images of the unmanned aerial vehicle, and the direct use of the non-supervision method can lead to false detection and missing detection of the power line.
The supervised method uses a deep learning model, usually a convolutional neural network, and trains the model through a large number of manually marked data sets, and the detection of the power line, the insulator and the power transmission tower can be realized through supervised learning, for example, the Mask R-CNN and Fast R-CNN detection methods based on targets have the characteristics of high detection precision, high detection time cost and the like, and the other method is a target detection method based on regression, such as Yolo and SSD algorithm. The above-described supervision algorithm requires a certain number of marked data sets and the algorithm performance is largely dependent on the data set quality.
Disclosure of Invention
In view of this, the present application provides a power line extraction method and apparatus, a storage medium, and a computer device, which improve accuracy of power line extraction by combining an unsupervised algorithm (line detection) and a supervised algorithm (coordinate prediction model).
According to one aspect of the present application, there is provided a power line extraction method, the method comprising:
acquiring a standardized image of a power transmission line corresponding to a power transmission line to be detected, and carrying out electric tower coordinate prediction on the standardized image of the power transmission line according to a coordinate prediction model to obtain electric tower prediction coordinates;
Performing edge detection on the standardized image of the power transmission line to obtain an edge detection result image, and calculating an electric tower offset angle based on the edge detection result image;
determining a tower positioning frame in the standardized power transmission line image according to the tower prediction coordinates, and determining an effective power line extraction area according to the tower positioning frame and the tower offset angle;
and detecting straight lines in the effective power line extraction area, determining straight lines perpendicular to the electric tower offset angle in the detected straight lines as candidate power lines, voting the candidate power lines according to a preset power line template, and determining the power lines of the power transmission line to be detected according to voting results.
According to another aspect of the present application, there is provided a power line extraction apparatus, the apparatus comprising:
the electric tower coordinate prediction module is used for acquiring a standardized image of the power transmission line corresponding to the power transmission line to be detected, and carrying out electric tower coordinate prediction on the standardized image of the power transmission line according to the coordinate prediction model to obtain electric tower prediction coordinates;
the electric tower angle calculation module is used for carrying out edge detection on the standardized image of the power transmission line to obtain an edge detection result image, and calculating an electric tower offset angle based on the edge detection result image;
The effective area determining module is used for determining an electric tower positioning frame in the standardized power transmission line image according to the electric tower prediction coordinates, and determining an effective power line extraction area according to the electric tower positioning frame and the electric tower offset angle;
the power line extraction module is used for detecting the straight line of the effective power line extraction area, determining the straight line perpendicular to the electric tower offset angle in the detected straight line as a candidate power line, voting the candidate power line according to a preset power line template, and determining the power line of the power transmission line to be detected according to the voting result.
According to still another aspect of the present application, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, implements the above-described power line extraction method.
According to still another aspect of the present application, there is provided a computer device including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, the processor implementing the above power line extraction method when executing the program.
By means of the technical scheme, the power line extraction method, the power line extraction device, the storage medium and the computer equipment provided by the application are used for carrying out electric tower coordinate prediction on the image of the standardized part of the power transmission line to be detected according to the coordinate prediction model to obtain electric tower prediction coordinates; performing edge detection on the standardized image of the power transmission line and calculating an offset angle of the electric tower; determining a tower positioning frame according to the tower prediction coordinates, and determining an effective power line extraction area according to the tower positioning frame and the tower offset angle; and carrying out linear detection on the effective power line extraction area, determining a straight line perpendicular to the electric tower offset angle in the detected straight line as a candidate power line, voting the candidate power line according to a preset power line template, and determining the power line of the power transmission line to be detected according to the voting result. Based on the characteristics of aerial images, electric tower prediction coordinates are obtained according to a supervised algorithm (a coordinate prediction model), then edge detection is carried out on a standardized image of a power transmission line so as to detect accurate and clear edges, an electric tower offset angle is calculated according to an edge detection result image, a priori knowledge is combined, a power line is required to be in a range of the electric power tower, the power line is perpendicular to the electric power tower, a straight line is detected in an effective power line extraction area through a straight line detection algorithm (an unsupervised algorithm), finally a straight line approximately perpendicular to the electric tower offset angle is selected as a candidate power line, and the power line is determined according to a preset power line template, so that the accuracy of power line detection is improved.
The foregoing description is only an overview of the technical solutions of the present application, and may be implemented according to the content of the specification in order to make the technical means of the present application more clearly understood, and in order to make the above-mentioned and other objects, features and advantages of the present application more clearly understood, the following detailed description of the present application will be given.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute an undue limitation to the application. In the drawings:
fig. 1 shows a schematic flow chart of a power line extraction method according to an embodiment of the present application;
fig. 2 is a schematic flow chart of another power line extraction method according to an embodiment of the present application;
fig. 3 is a schematic flow chart of another power line extraction method according to an embodiment of the present application;
fig. 4 shows a schematic structural diagram of a power line extraction device according to an embodiment of the present application;
fig. 5 shows a schematic structural diagram of another power line extraction device according to an embodiment of the present application.
Detailed Description
The present application will be described in detail hereinafter with reference to the accompanying drawings in conjunction with embodiments. It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other.
In this embodiment, there is provided a power line extraction method, as shown in fig. 1, including:
step 101, obtaining a standardized image of the power transmission line corresponding to the power transmission line to be detected, and carrying out electric tower coordinate prediction on the standardized image of the power transmission line according to a coordinate prediction model to obtain electric tower prediction coordinates.
With the development of unmanned technology, unmanned technology is applied to various industries. To the technical field of power line inspection, unmanned machines such as unmanned aerial vehicle are adopted for inspection, and the power line inspection system has the advantages of being low in cost, convenient to operate and the like. Unmanned machines such as unmanned aerial vehicle can be used for carrying high-resolution aerial camera or other shooting machines for shoot aerial image, carry out the mode of patrolling and examining high-voltage transmission line through aerial image, with low costs and factor of safety is little. The aerial image has the following characteristics: the power lines are linear objects, approximately parallel; the power line background is natural wind and light; the power line is made of special materials, and the brightness is uniform; the power lines are shown as thick lines or thin lines in the aerial image, and the left side pixels and the right side pixels are opposite and partially parallel; the power line width and the appearance color are different.
In the above-described embodiments of the present application, the power line is extracted based on the features of the aerial image, so that the power transmission line is detected according to the extracted power line. Specifically, a transmission line basic image (aerial image) corresponding to the transmission line to be detected is obtained, the transmission line basic image is subjected to standardization processing, a transmission line standardization image is obtained, for example, an image with preset specification size (1280×1280 or 640×640) is intercepted by taking the center of the aerial image as a reference point, and finally unified normalization is carried out to obtain an image with 640×640 size. And carrying out electric tower coordinate prediction on the image of the standardized position of the power transmission line according to the coordinate prediction model to obtain electric tower prediction coordinates. The prior knowledge of the electric tower positioning power line can know that the power line is necessarily distributed in the electric tower range, and after the electric tower position is determined through the coordinate prediction model, the power line extraction can be performed according to the electric tower position, so that the power line extraction efficiency is improved.
And 102, performing edge detection on the standardized image of the power transmission line to obtain an edge detection result image, and calculating a tower offset angle based on the edge detection result image.
And step 103, determining a tower positioning frame in the standardized power transmission line image according to the tower prediction coordinates, and determining an effective power line extraction area according to the tower positioning frame and the tower offset angle.
And then, edge detection is carried out on the standardized image of the power transmission line so as to detect an accurate and clear edge, thereby being beneficial to further power line extraction. And obtaining an edge detection result image after edge detection, and calculating a tower offset angle based on the edge detection result image to prepare for subsequent power line extraction.
Then, according to the electric tower predicted coordinates, an electric tower positioning frame is determined in the standardized image of the power transmission line, in particular, according to the prior knowledge of electric tower positioning power lines, after the position of the electric tower is determined, the power lines are necessarily distributed in the electric tower range, so that according to the electric tower positioning frame and the electric tower offset angle, an effective power line extraction area can be further determined, and the interference of the straight line groups outside the electric tower range on the power line extraction process is eliminated.
And 104, detecting straight lines of the effective power line extraction area, determining straight lines perpendicular to the electric tower offset angle in the detected straight lines as candidate power lines, voting the candidate power lines according to a preset power line template, and determining the power lines of the power transmission line to be detected according to voting results.
And then, carrying out straight line detection on the effective power line extraction area, determining a straight line perpendicular to the electric tower offset angle in the detected straight lines as a candidate power line, voting the candidate power line according to a preset power line template, and determining the power line of the power transmission line to be detected according to the voting result. The power line is necessarily in the range of the power tower by combining priori knowledge, the power line is vertical to the power tower, the power tower offset angle is obtained based on an image (edge detection result image) after edge detection, the power line range (effective power line extraction area) is obtained through the power tower positioning frame and the power tower offset angle, then the environment interference outside the power line range is filtered, then a straight line is detected by using a straight line detection algorithm, a straight line approximately vertical to the power tower offset angle is selected as a candidate power line, finally the power line is extracted by combining a preset power line template, and the accuracy of power line detection is improved.
By applying the technical scheme of the embodiment, the electric tower coordinate prediction is carried out on the image of the standardized position of the transmission line to be detected according to the coordinate prediction model, so as to obtain electric tower prediction coordinates; performing edge detection on the standardized image of the power transmission line and calculating an offset angle of the electric tower; determining a tower positioning frame according to the tower prediction coordinates, and determining an effective power line extraction area according to the tower positioning frame and the tower offset angle; and carrying out linear detection on the effective power line extraction area, determining a straight line perpendicular to the electric tower offset angle in the detected straight line as a candidate power line, voting the candidate power line according to a preset power line template, and determining the power line of the power transmission line to be detected according to the voting result. Based on the characteristics of aerial images, electric tower prediction coordinates are obtained according to a supervised algorithm (a coordinate prediction model), then edge detection is carried out on a standardized image of a power transmission line so as to detect accurate and clear edges, an electric tower offset angle is calculated according to an edge detection result image, a priori knowledge is combined, a power line is required to be in a range of the electric power tower, the power line is perpendicular to the electric power tower, a straight line is detected in an effective power line extraction area through a straight line detection algorithm (an unsupervised algorithm), finally a straight line approximately perpendicular to the electric tower offset angle is selected as a candidate power line, and the power line is determined according to a preset power line template, so that the accuracy of power line detection is improved.
Further, as a refinement and extension of the foregoing embodiment, in order to fully describe the implementation procedure of this embodiment, another power line extraction method is provided, as shown in fig. 2, and the method includes:
step 201, a standardized image of a power transmission line corresponding to a power transmission line to be detected is obtained, and electric tower coordinate prediction is carried out on the standardized image of the power transmission line according to a coordinate prediction model, so that electric tower prediction coordinates are obtained.
In the embodiment of the application, the unmanned aerial vehicle can be adopted to shoot the power transmission line to be detected, the aerial image (power transmission line base image) is obtained, the power transmission line base image is subjected to standardized operation, such as cutting and normalization processing, so that the power transmission line standardized image is obtained, the aerial power line image quality can be effectively improved, and the power line extraction accuracy is improved. And carrying out electric tower coordinate prediction on the image of the standardized position of the power transmission line according to the coordinate prediction model to obtain electric tower prediction coordinates.
Step 202, gray scale adjustment is carried out on pixels in the standardized image of the power transmission line to obtain a gray scale image of the power transmission line, histogram equalization is carried out on the gray scale image of the power transmission line to obtain an equalized image of the power transmission line, and denoising is carried out on the equalized image of the power transmission line according to a Gaussian filtering method to obtain a preprocessed image of the power transmission line.
Then, carrying out image preprocessing on the standardized image of the power transmission line, specifically, carrying out gray scale adjustment on pixels in the standardized image of the power transmission line to obtain a gray scale image of the power transmission line, carrying out histogram equalization processing on the gray scale image of the power transmission line to obtain an equalized image of the power transmission line, and carrying out denoising processing on the equalized image of the power transmission line according to a Gaussian filtering method to obtain a preprocessed image of the power transmission line. The image preprocessing technology such as graying and histogram equalization is used for adjusting the image contrast, the power transmission line preprocessing image is obtained after normalization operation, the operand of the later-stage power line extraction is reduced, and meanwhile, convenience is provided for further edge detection.
Optionally, step 202 includes:
step 202-1, adjusting a gray value of each pixel in the standardized image of the power transmission line according to a pixel gray adjustment formula to obtain the gray image of the power transmission line, wherein the pixel gray adjustment formula is as follows:
Gray(i,j)＝0.29*R+0.578*G+0.114*B
gray (i, j) represents the Gray value of the j-th column of the i-th row in the pixel point matrix corresponding to the standardized image of the power transmission line, and R, G and B represent the red channel matrix, the green channel matrix and the blue channel matrix respectively.
Step 202-2, according to a histogram equalization formula, equalizing gray scale distribution in the transmission line gray scale image to obtain a transmission line equalized image, wherein the histogram equalization formula is as follows:
n represents the sum of pixels in the transmission line graying image, k represents the k-level gray level, n k Representing a gray level r k L represents the total number of gray levels in the transmission line gray-scale image, p r (r) represents the gray level probability density of the transmission line graying image.
In the above-described embodiments of the present application, the power tower and the power line are generally silvery white, and the gray scale image using the blue channel is more easily distinguished from the environment, so the blue channel is adopted as the gray scale image. And (3) carrying out gray scale processing on the standardized image of the power transmission line by adopting a weighted average method, namely adjusting the gray scale value of each pixel in the standardized image of the power transmission line according to a pixel gray scale adjustment formula to obtain the standardized image of the power transmission line, wherein the pixel gray scale adjustment formula is as follows:
Gray(i,j)＝0.29*R+0.578*G+0.114
in the formula, gray (i, j) represents Gray values of an ith row and a jth column in a pixel point matrix corresponding to the standardized image of the power transmission line, and R, G and B represent a red channel matrix, a green channel matrix and a blue channel matrix respectively.
Because the power line is more fuzzy after the power transmission line standardized image is gray, the contrast of the power transmission line gray image is regulated by using the histogram equalization, so that the detail of the power line is more obvious, and specifically, the gray distribution in the power transmission line gray image is equalized according to a histogram equalization formula to obtain the power transmission line equalized image, wherein the histogram equalization formula is as follows:
where n represents the sum of pixels in the image, k represents a gray level, n k Is of gray level r k L is the total number of gray levels in the image, p r (r) represents the gray level probability density of the image.
Step 203, adding a first diagonal direction gradient template and a second diagonal direction gradient template on the basis of a horizontal direction gradient template and a vertical direction gradient template included in the Canny edge detection algorithm, and calculating gradient values of the preprocessed image of the power transmission line according to the four direction gradient templates.
And 204, generating a highest threshold and a lowest threshold in gray values of pixels of the power transmission line preprocessing image according to an iterative algorithm, and reserving edge pixels corresponding to gradient values between the highest threshold and the lowest threshold to obtain an edge detection result image.
Then, edge detection is carried out on the filtered image (the preprocessed image of the power transmission line), a 2X 2 template is used for calculating gradient amplitude and size by a traditional Canny edge detection algorithm, and edge details of the image cannot be detected well due to large noise interference, so the Canny edge detection algorithm is improved, namely, after a first diagonal gradient template and a second diagonal gradient template are added on the basis of an original horizontal gradient template and a vertical gradient template, gradient values of the preprocessed image of the power transmission line are calculated according to the four direction gradient templates. And then generating a highest threshold value and a lowest threshold value in the gray values of all pixels of the preprocessing image of the power transmission line according to an iterative algorithm, and reserving edge pixels corresponding to gradient values between the highest threshold value and the lowest threshold value to obtain an edge detection result image.
Specifically, after two gradient direction templates (a first diagonal direction gradient template and a second diagonal direction gradient template) are added by improving the traditional Canny edge detection algorithm, four gradient templates are calculated horizontally, vertically and diagonally as follows:
the filtered image (transmission line pre-processed image) is convolved using the four templates described above, wherein,
The 45-degree directional gradient calculation formula is as follows:
the 135 DEG directional gradient calculation formula is:
the horizontal gradient was calculated as:
the vertical gradient calculation formula is:
the calculated gradients of 45 ° and 135 ° are projected into the horizontal and vertical directions and then summed to obtain new gradient values for the x-axis and y-axis, given by:
then calculating the gradient magnitude M and the direction D of the gray value of the current pixel, wherein the calculation formula is as follows:
the gradient calculation method in two directions is added, so that gradient values in four directions of 8 pixels around the pixel are fully considered, more image edge information can be obtained, edge positioning becomes more accurate, and the false detection rate and the omission rate are greatly reduced.
Further, because the Canny edge detection algorithm is limited by selecting the high threshold and the low threshold through manual experience, the optimal threshold is difficult to obtain, so that the power line detection result is affected.
Specifically, an initial threshold is set first, and the formula is:
T{T k |K＝0}
wherein T is the initial threshold of the image, K is the iterative times of the algorithm, Z max Is the maximum gray value, Z min Is the minimum gray value, and after the initial threshold value is obtained by calculation, the power transmission line pretreatment image is divided into H which is higher than the initial threshold value 0 And below the initial threshold H 1 Is H 0 H and H 1 The formulas are respectively as follows:
H 0 ＝{f(x,y)|f(x,y)>T}
H 1 ＝{f(x,y)|f(x,y)<T}
respectively calculating the gray average value T of the two parts H 、T L Wherein T is H T and T L The formulas are respectively as follows:
where f (x, y) represents the gray value of the (i, j) point in the image, where N 0 (i,j)、N 1 The value formulas of (i, j) are respectively as follows:
and calculating a new threshold TN, wherein the formula is as follows:
when the final iteration threshold is equal to the initial threshold or meets the set reasonable error range, the iteration is stopped, otherwise, the iteration continues to run, and finally the optimal T is obtained H 、T L For this reason, the noise interference to the threshold selection is greatly reduced.
Through carrying out edge detection on the Gaussian filtered image (the power transmission line pretreatment image) to obtain an image with less interference, improving a Canny edge detection algorithm, introducing a 45 DEG and 135 DEG gradient template, and fully considering gradient values of 8 pixels around the pixels, more image edge information can be obtained, the edge positioning becomes more accurate, and the false detection rate and the omission rate are greatly reduced. Aiming at the numerical relation of the high threshold value and the low threshold value in the traditional Canny edge detection algorithm, the optimal double threshold value is obtained through an iteration method, compared with the traditional Canny edge detection algorithm, the interference of noise on threshold value selection is reduced, and the detection precision and the robustness are greatly improved.
Step 205, calculating potential power lines and potential power line slopes corresponding to the edge detection result image according to a Hough transformation algorithm, and determining potential power line groups parallel to each other according to the potential power line slopes.
Step 206, clustering the slopes of the potential power line groups according to a K-means++ clustering algorithm, determining the target class with the largest number of slopes contained in a plurality of classes, and obtaining the clustering center slope of the target class as an angle theta 1 According to the angle theta 1 Calculating the electric tower offset angle theta by using an electric tower offset angle calculation formula 2 Wherein, the electric tower offset angle formula of calculation is:
then, calculating potential power lines and potential power line slopes corresponding to the edge detection result image according to a Hough transformation algorithm, determining potential power line groups parallel to each other according to the potential power line slopes, clustering the slopes of the potential power line groups according to a K-means++ clustering algorithm, determining target classes with the largest number of slopes contained in the classes, and acquiring the clustering center slope of the target class as an angle theta 1 According to angle theta 1 Calculating the electric tower offset angle theta by using an electric tower offset angle calculation formula 2 。
Specifically, the Hough transform algorithm is one of image feature extraction technologies, and converts a shape extraction problem into a parameter space peak value calculation problem according to the duality of points and lines, wherein an equation of converting a straight line into a parameter space can be expressed as:
r θ ＝x θ cosθ+y θ sinθ
Wherein r is θ Represents the distance from the origin to the straight line in the rectangular coordinate system, θ represents the angle between the straight line and the x-axis, and for each pair (r θ θ) represents (x θ ,y θ ) Is a straight line of (a).
And then, the local maximum value calculated by the parameter space accumulator corresponds to a specific shape to obtain the specific shape in the original image, specifically, a parameter space accumulator two-dimensional array (r, theta) is established, all pixels to be detected in the image are obtained into an accumulator array according to the linear polar coordinates, the accumulator array is added with 1, the (r ', theta') corresponding to the local maximum value in the accumulator is obtained, and the corresponding linear segment is extracted.
Because the power line penetrates through the whole image in the aerial image, the Hough transformation algorithm has an accurate detection result on the long line, and a detection threshold can be increased to filter some short lines, so that interference is reduced. In order to obtain a potential power line group, in the embodiment of the application, a Hough transformation algorithm is applied to an edge detection result image detected by an improved Canny edge detection algorithm, so that a straight line group parameter (r, theta) is obtained.
For straight lines detected by the Hough transform algorithm, not all straight lines are necessarily required power lines, so that filtering is required for the obtained straight line result. From a priori knowledge, in aerial images, the power lines are parallel to each other, thus filtering the slope of the straight line and preserving the parallel pairs. Meanwhile, as the power line penetrates through the whole image, the length of the lowest line segment detected by the Hough transformation algorithm is increased, and classification in the clustering algorithm is facilitated.
The K-Means clustering algorithm is one of the data mining algorithms, and K-means++ is selected in the embodiment of the application in order to improve the convergence speed because the K random initial clustering centers have uncertainty on the convergence effect of the clustering result. Specifically, a random point is selected from the input data set X as a first clustering center, a distance D (X) between each point X in the data set X and the selected clustering center is determined, then a p value of the possibility that the sample point X will be selected as the clustering center is determined, and a calculation formula of the p value is as follows:
sample points with high probability p values are selected until the necessary k cluster centers are selected as new cluster centers. The K-means algorithm is then performed using the selected K initial cluster centers.
In the embodiment of the application, the K-means++ algorithm is used for clustering the slopes of the parallel line groups to obtain the theta with the largest counting result 1 Through theta 1 To calculate the electric tower angle theta 2 The formula is:
for this purpose, the tower inclination angle θ was obtained using K-Means clustering 2 。
Step 207, determining a tower positioning frame in the standardized power transmission line image according to the tower prediction coordinates, and determining an effective power line extraction area according to the tower positioning frame and the tower offset angle.
And step 208, detecting the straight line of the effective power line extraction area, determining the straight line perpendicular to the electric tower offset angle in the detected straight line as a candidate power line, voting the candidate power line according to a preset power line template, and determining the power line of the power transmission line to be detected according to the voting result.
Next, determining a tower positioning frame in the standardized power transmission line image according to the tower prediction coordinates, determining an effective power line extraction area according to the tower positioning frame and the tower offset angle, for example, determining a tower offset frame according to the tower positioning frame and the tower offset angle, and determining the upper left corner point and the lower right corner point of the tower offset frameThe point (or upper right corner and lower left foot) is parallel to the power line θ 1 And (3) extending to obtain an effective power line extraction area so as to eliminate interference of straight lines outside the range of the power tower on power line extraction. And carrying out linear detection on the effective power line extraction area, determining a straight line perpendicular to the electric tower offset angle in the detected straight line as a candidate power line, voting the candidate power line according to a preset power line template, and determining the power line of the power transmission line to be detected according to the voting result.
At present, an unsupervised power line detection algorithm is realized by regarding a power line as a straight line, removing noise through a Gaussian filtering method, performing edge detection by using a Canny operator, extracting an object edge, extracting the straight line in an image by combining a Hough transformation algorithm, and finally connecting the power lines through a parallel constraint and straight line grouping method. The Canny edge detection algorithm has the characteristics of few calculation steps and easiness in implementation, but has the problems of false edges and edge deletion due to noise, environmental interference of roads, trees and the like, double edges are easy to appear in the power line detection, and the accuracy of edge positioning is affected. The Hough transformation algorithm has good anti-interference characteristics, but has high time and space complexity, only detects the straight line direction, loses straight line length information, is interfered by trees and ridges under a complex background, has obvious influence on performance, cannot be completely detected by a power line, and can detect irrelevant line segments. According to the embodiment of the application, based on the image characteristics of the aerial image, a Canny edge detection algorithm is improved, 45 DEG and 135 DEG gradient templates are introduced, meanwhile, the optimal double threshold value is obtained through an iteration method aiming at the numerical relation of the high threshold value and the low threshold value in the traditional Canny edge detection algorithm, the accuracy and the robustness of edge detection are improved, a Hough transformation algorithm is adopted to extract a power line, a Kmeans++ clustering method is adopted to calculate the offset angle of the power tower, and the accuracy of power line extraction is improved.
By applying the technical scheme of the embodiment, the electric tower prediction coordinates are obtained according to the coordinate prediction model. And carrying out image preprocessing on the standardized image of the power transmission line to obtain a preprocessed image of the power transmission line. The Canny edge detection algorithm is improved, a first diagonal gradient template and a second diagonal gradient template (for example, 45 DEG and 135 DEG gradient templates) are introduced, gradient values of the preprocessed image of the power transmission line are calculated, the highest threshold and the lowest threshold are determined according to an iterative algorithm, edge pixels corresponding to the gradient values between the highest threshold and the lowest threshold are reserved, and an edge detection result image is obtained. And determining a potential power line group according to the Hough transformation algorithm, calculating a tower offset angle according to the K-means++ clustering algorithm and the potential power line group, and determining an effective power line extraction area according to the tower prediction coordinates. And carrying out linear detection on the effective power line extraction area, determining a straight line perpendicular to the tower offset angle in the detected straight lines as a candidate power line, and finally voting to determine the power line of the power transmission line to be detected. Based on the image characteristics of the aerial image, a Canny edge detection algorithm is improved, 45-degree and 135-degree gradient templates are introduced, so that four-direction gradient values of 8 pixels around the pixels are fully considered, and an optimal double threshold value is obtained through an iteration method according to the numerical relation of high and low threshold values in the traditional Canny edge detection algorithm, so that the accuracy and the robustness of edge detection are improved, and the accuracy of power line extraction is improved.
Further, as a refinement and extension of the foregoing embodiment, in order to fully describe the implementation procedure of this embodiment, another power line extraction method is provided, as shown in fig. 3, and the method includes:
step 301, obtaining a standardized image of a power transmission line corresponding to the power transmission line to be detected, and predicting coordinates of an electric tower and an insulator according to the standardized image of the power transmission line according to a coordinate prediction model to obtain predicted coordinates of the electric tower, predicted coordinates of the insulator and predicted center points of the insulator.
And step 302, performing edge detection on the standardized image of the power transmission line to obtain an edge detection result image, and calculating a tower offset angle based on the edge detection result image.
In the above embodiment of the present application, a standardized image of a power transmission line corresponding to a power transmission line to be detected is obtained, and electric tower and insulator coordinate prediction is performed on the image of the standardized position of the power transmission line according to a coordinate prediction model, so as to obtain electric tower prediction coordinates, insulator prediction coordinates and an insulator prediction center point. And carrying out edge detection on the standardized image of the power transmission line to obtain an edge detection result image, and calculating a tower offset angle based on the edge detection result image so as to prepare for the subsequent power line extraction.
Step 303, determining a tower positioning frame in the standardized power transmission line image according to the tower prediction coordinates, and according to the tower direction and the tower offset angle theta 2 And rotating the electric tower positioning frame, and amplifying the rotated electric tower positioning frame by taking the central point of the rotated electric tower positioning frame as the center according to a preset amplification factor to obtain an electric tower offset frame, wherein the electric tower positioning frame is rectangular.
Step 304, determining a first extension point and a second extension point according to the opposite angles of the electric tower offset frame, and determining an effective power line extraction area according to the first extension point and the second extension point.
Specifically, determining a tower positioning frame in a standardized power transmission line image according to a tower prediction coordinate, and obtaining a power line range (effective power line extraction area) by rotating the tower positioning frame and adding a certain offset, specifically, using the tower positioning frame according to a tower offset angle theta 2 The rotation is performed, in particular, the orientation of the electric towers in the image needs to be distinguished when the rotation is performed, so that the whole electric towers can be displayed in the rotated electric tower positioning frame, and the electric towers are prevented from being missed. And adding preset offset to four points of the rectangle of the electric tower positioning frame, or amplifying the rotating electric tower positioning frame by taking the central point of the rotating electric tower positioning frame as the center according to a preset amplification factor to obtain the electric tower offset frame, wherein the preset offset is set according to different details of detection lines, and a first extension point and a second extension point (an upper left corner point and a lower right corner point or an upper right corner and a lower left foot) are determined according to opposite angles of the electric tower offset frame so as to determine an effective power line extraction area according to the first extension point and the second extension point.
After determining the effective power line extraction area, verification may be performed, in particular, for a point in space, whether it is to the left or right of a straight line. Assuming straight lines with itTwo points a (x 1 ,y 1 )，B(x 2 ,y 2 ) And the straight line direction points from a to B, then the straight line can be expressed as:
αx+βy+γ＝0
wherein alpha is y 2 -y 1 Beta is x 2 -x 1 Gamma is x 2 *y 1 -x 1 *y 2 。
C＝αx d +βy p +γ
The inspector can determine a certain point D (x) in the space by calculating C d ,y d ) On which side of the line.
And 305, performing linear detection on the effective power line extraction area according to an LSD linear detection algorithm to obtain candidate power line segments to be fitted, and placing the candidate power line segments to be fitted into a line segment pool.
And step 306, placing the candidate power line segments to be fitted smaller than a preset distance threshold into the candidate power line segment group to be fitted according to the distance between the candidate power line segments to be fitted in the line segment pool.
Step 307, fitting all the candidate power line segments to be fitted in the candidate power line segment group to be fitted according to a least square method, obtaining a fitting straight line, and determining the fitting straight line perpendicular to the tower offset angle in the fitting straight line as a candidate power line.
In the effective power line extraction area, line detection is performed using an LSD straight line detection algorithm, which is a local extraction algorithm for straight lines, has an advantage of being faster than the Hough transform algorithm, and is designed to be parameter-free on digital images. However, a disadvantage is that long line segments are often cut into multiple straight lines due to the self-growing nature of the local detection algorithm, occlusion, and local blurring, which is not present in the Hough transform algorithm. The LSD straight line detection algorithm is chosen because it can accommodate various complex environments without parameter adjustment and straight line detection is preferred over Hough transform algorithms.
And carrying out linear detection on the effective power line extraction area according to an LSD linear detection algorithm to obtain a candidate power line segment to be fitted, filtering and merging the candidate power line segment to be fitted, and specifically, putting the candidate power line segment to be fitted into a line segment pool. In the line segment pool, a preset distance threshold is set to 7 pixel values, and the distance between the candidate power line segments to be fitted can be calculated by the following formula:
wherein k is the slope of the current line segment, I is the intercept, a line segment group of each line is obtained, then a straight line is fitted through a least square method, and a straight line which is approximately perpendicular to the tower angle obtained by Kmeans++ clustering is selected as a candidate power line.
Step 308, according to the predicted coordinates of the insulator, acquiring a tower standard image in a tower standard template library corresponding to the power transmission line to be detected, mapping the tower standard image to a power transmission line standardized image according to a preset power line template, voting the candidate power line according to the distance between the predicted center point of the insulator and the candidate power line, and determining the power line of the power transmission line to be detected according to the voting result, wherein the preset power line template is obtained by calculating a homography matrix of the power transmission line standardized image mapped to the tower standard image.
In computer vision, homography is essentially a property that exists between image transformations, and can be classified into rigid transformations, affine transformations, and projective transformations according to the transformation property. Wherein, the rigid transformation refers to a transformation form only comprising translation and rotation, only changing the position and the direction of the original image, and not changing the shape, and the affine transformation further changes the shape of the image compared with the rigid transformation, but maintains the parallel relation of the original image in the transformation process.
The transformation property between planes is essentially the position change of the point, which can be represented by a matrix, and the corresponding relation between the two planes of the point can be represented by a homography constraint on the assumption that the point Q in the original plane becomes Q' in another plane after projection transformation:
Q′＝AQ
where a is referred to as a homography matrix, typically of size 3*3, and the transformation between two points after expansion can be expressed as:
according to different specific transformation processes, the values of the parameters in the homography matrix are different, and in general cases, attention is required to how to calculate the matrix A, so that for any point in the original plane, the position of the point on another plane can be calculated through the homography matrix.
Spreading Q' =aq gives three equations as follows:
x′＝a 11 x+a 12 y+a 13
y′＝a 21 x+a 22 y+a 23
1＝a 31 x+a 32 y+1
after multiplication, we obtain:
xa 11 +ya 12 +a 13 -xx′a 31 -yx′a 32 ＝x′
xa 21 +ya 22 +a 23 -xy′a 31 -yy′a 32 ＝y′
the final unified representation is the matrix form ba=b, namely:
the equation has 8 unknowns, two equations can be obtained by a pair of matching point pairs, according to the polynomial theory, at least four pairs of matching point pairs are needed to solve the homography matrix, and the two sides are multiplied by B at the same time T And then performing term transfer by utilizing the property of the inverse matrix to obtain a solving formula of a, wherein the solving formula is as follows:
B T Ba＝B T b
a＝(B T B) -1 B T b
for multiple-view aerial images at the same place, the precondition of homography is that two plane spaces under different view angles belong to the same plane in the actual space, and a public plane is found under different view angles for solving the homography matrix. For aerial power line images (transmission line standardized images), the planes of the electric towers and the parts where the insulators are located meet the condition, so that in the embodiment of the application, homography matrixes under different visual angles are calculated according to the positions of the centers of calibrated insulator detection frames (electric tower standard images obtained from an electric tower standard template library corresponding to the transmission line to be detected) and the positions of insulator prediction coordinate center points predicted by a coordinate prediction model, matching point pair selection is respectively carried out in two power line images (the electric tower standard images corresponding to the same electric tower and the transmission line standardized images), and six pairs of matching points are selected for homography matrix calculation so that detection information is more accurately and fully used for homography matrix calculation.
In order to verify the accuracy of the homography matrix, in the embodiment of the application, between two visual angles, the homography matrix is used for mapping the center point of an actual frame (an electric tower standard image) in the standard template into the standardized image of the power transmission line, the center point coordinates of the predicted coordinates of the comparative insulator are compared with the center point coordinates of the mapped standard insulator, and the error between the center point coordinates of the mapped standard insulator and the center point coordinates of the predicted coordinates of the insulator is smaller. Therefore, the coordinates of the insulator in the standard template are mapped into the power line image to be detected through homography transformation, and the power line candidate group is voted according to the distance between the center point of the insulator and the power line, so that the corresponding power line is obtained.
Through the technical scheme of the embodiment, firstly, image standardization operation is carried out, and the acquired unmanned aerial vehicle aerial image is cut and normalized. And secondly, preprocessing an image, obtaining a power tower position (power tower predicted coordinates) by using a coordinate prediction model, preprocessing an image with less interference by carrying out gray level and histogram equalization on a standardized power transmission line image, and removing background information outside the power tower range (power transmission line preprocessed image) by using Gaussian filtering in combination with power tower position information. And thirdly, carrying out edge detection on the preprocessed image of the power transmission line, and obtaining an accurate and clear image edge by using an improved Canny edge detection algorithm. Finally, extracting the power line by combining priori knowledge, acquiring the angle of the power tower (the electric tower offset angle) by using K-means clustering of the image after edge detection, obtaining the power line range (the effective power line extraction area) by rotating the electric tower positioning frame and adding a certain offset, filtering out the environment interference outside the power line range, detecting the power line by using an LSD straight line detection algorithm, finally, carrying out straight line combination, selecting a straight line approximately perpendicular to the electric tower offset angle, namely the candidate power line in the aerial image, combining the power line template and the homography matrix, voting to determine the straight line of the power line, and improving the accuracy of power line extraction.
In an embodiment of the present application, optionally, the coordinate prediction model is an improved model of the YOLOv7 model, and the coordinate prediction model is obtained by replacing a SiLU activation function in the YOLOv7 model with a mix activation function, and replacing a CIou loss function in the YOLOv7 model with a SIoU loss function.
In this embodiment of the present application, optionally, the training method of the coordinate prediction model includes obtaining a plurality of data set base images including an electric tower, performing standardization processing on the data set base images to obtain data set standardized images, and labeling the electric tower and an insulator included in the electric tower in the data set standardized images by a preset data labeling tool to generate an electric tower positioning data set; dividing the electric tower positioning data set into an electric tower positioning training data set with a preset first proportion, an electric tower positioning test data set with a preset second proportion and an electric tower positioning verification data set with a preset third proportion in a uniform random sampling mode; and inputting the electric tower positioning training data set, the electric tower positioning test data set and the electric tower positioning verification data set into a coordinate prediction model for training, and obtaining a trained coordinate prediction model.
The implementation approach of the supervised learning power line detection method is to establish a convolutional neural network model, construct an artificial labeling power line data set, extract and classify the characteristics of the preprocessed aerial power line data set images, and filter the characteristic images through structural information to obtain power line information.
In the supervised learning method, mask R-CNN and Fast R-CNN are divided into two steps, candidate frames are generated, CNN extraction features are generated, classification and frame regression are generated, the R-CNN series can continuously improve the precision by optimizing a network, but the detection speed is reduced by two detection steps, and the calculation load is high. The YOLO series divides an input picture into S grids, K prediction boundary boxes and C target classification scores of the grids are set, then grids with highest scores are used as final prediction results according to the confidence level of the K prediction frames and the final score level of the C target boundary boxes, and experimental results show that the YOLO network is 42 times Faster than Fast R-CNN network running time and twice ahead than Fast R-CNN, but because each grid only outputs one result, misjudgment and missed detection can occur when the grid division is overlarge or the power line targets in aerial images are too small. When the supervision and learning method is used for detecting the power line, the quality requirements on the data sets are higher, most of the power line data sets are non-public data sets, and images of broken power lines, hanging of foreign matters, smoke and fire conditions and the like in the public data sets are too few, so that the condition that the supervision and learning method is used for training is difficult to meet the detection requirements of safe and stable operation of the power system.
In the above embodiment of the present application, for an aerial high-voltage transmission line image, a dataset is made, an aerial image (including a plurality of dataset base images of an electric tower) is taken as a reference point from the center of 1280×1280, 640×640 size images, unified normalized to 640×640, and a labelme is adopted to label an electric tower in the image, so as to generate a dataset standardized image (dataset), and because the aerial image includes a plurality of backgrounds, all aerial images come from an unmanned plane along the line for aerial photography, and a specific dataset may include: grassland background (400), forest background (500), farmland background (350), water area background (50), city background (100).
The data set is enhanced by adopting a mode of rotating the images at 45 degrees, the data set is expanded to 16 times of the original size, in the subsequent training, the image weight coefficient is used, the images are selected according to the number of the label boxes in the training process, if the number of the label boxes is larger, the weight is larger, and the sampling probability is correspondingly increased. And for label distribution, consistent with the YOLOv7 model, a SimOTA strategy is adopted, k positive samples are distributed to each label box in a self-adaptive and dynamic mode in the training process, and compared with OTA, the training time is remarkably shortened.
Secondly, dividing the electric tower positioning data set into an electric tower positioning training data set with a preset first proportion, an electric tower positioning test data set with a preset second proportion and an electric tower positioning verification data set with a preset third proportion without intersection, for example, setting the proportion of a training set (electric tower positioning training data set), a test set (electric tower positioning test data set) and a verification set (electric tower positioning verification data set) to be 6:2:2, and finally inputting 640 multiplied by 3 images in the processed data set into a YOLOv7 network.
The YOLOv7 network is divided into a back box and a Head, wherein the back box is composed of a Conv convolution module, an ELAN module and an MP-1 module, the Conv in the YOLOv7 is not a common convolution layer and is composed of a common two-dimensional convolution, a BN layer and a SiLU activation function, and the Head is composed of a Conv convolution module, an SPPCSPC module, an MP-2 module and a detection module.
The SiLU activation function is a Sigmoid weighted linear combination function commonly used in machine learning
SiLU(x)＝x*Sigmoid(x)
The method is obtained by linearly combining the input x and the Sigmoid function, has excellent characteristics of no upper bound, no lower bound, smoothness and non-monotonic, the no upper bound avoids saturation caused by the upper bound, effectively reduces the disappearance of gradients, the lower bound can reduce overfitting, has a certain regularization effect, and the non-monotonic ensures the negative output generated by the negative input, thereby improving the network expression capability.
A large number of experiments show that the Mish activation function
Mish(x)＝x*tanh(ln(1+e x ))
The method has almost the same characteristics as the SiLU function and better accuracy, and based on the YOLOv7 model, the SiLU activation function in Conv is replaced by the Mish activation function, so that the data set in the facts of the application improves the mAP value by 1% by replacing the activation function.
The target detection loss function is mainly composed of two parts: classification loss and regression positioning loss. Regression localization loss as an integral part of the target detection loss function, YOLOv7 uses a loss function of CIoU that considers three geometric parameters: the overlapping area, the center point distance and the aspect ratio are increased by the loss of the length and the width on the basis of the DIoU function, and the CIoU loss function formula is as follows:
wherein alpha is a weight parameter, ρ is a Euclidean distance between two points, v is the consistency of the aspect ratio of the prediction frame and the target frame, and the calculation method formula of v is as follows:
wherein w is gt 、h gt W, h represent the width and height of the real and predicted frames, respectively.
The CIou function does not consider the angle problem, so that the convergence speed is low and the efficiency is low, therefore, the SIoU function is introduced, the matching direction of the prediction frame and the real frame is introduced, the model convergence is quickened, the SIoU function is composed of four cost functions, and the angle cost function is defined as the following formula:
Wherein,representing the center coordinates of the real and predicted frames respectively,
since the angle loss is introduced, the distance cost function is modified as follows:
here ρ x And ρ y And γ is respectively:
γ＝2-Λ
wherein c w And c h The width and height of the minimum boundary rectangle of the real frame prediction frame are respectively defined as the following shape cost function definition formula:
w herein w 、w h The method comprises the following steps of:
finally defining the IOU cost function as a formula:
wherein B and B GT And respectively representing the candidate frame and the standard frame, and finally obtaining an SiOU loss function formula as follows:
therefore, the improved YOLOv7 model is used for training the data set to obtain the center point, width and height of the prediction frame of the power tower and the insulator. The acquired unmanned aerial vehicle aerial image is cut and normalized (image standardization operation is carried out), then image preprocessing is carried out, an improved YOLOv7 network is used for training a data set, a Mish activation function is used for replacing a SiLU activation function, the disappearance of gradients is effectively reduced, the over-fitting phenomenon is reduced, a certain regularization effect is achieved, negative output is allowed, the expression capacity of the network is improved, the convergence and the accuracy of a model are accelerated, a CIou loss function is replaced by a CIou loss function, the angle parameter is considered, the training speed and the reasoning accuracy are effectively improved, and the Map value detected by an electric tower is improved by 2% (YOLOv 7.864, and the embodiment of the application is 0.881).
By applying the technical scheme of the embodiment, the YOLOv7 network model, the Canny edge detection algorithm, the Hough transformation algorithm and the power line extraction method are improved, the YOLOv model is a classical single-stage detection network, and has the characteristics of high running speed and small memory occupation ratio, but the detection accuracy is not ideal, the YOLOv7 network is improved, the Mish activation function is used for replacing the original activation function, the SIoU is used as a loss function, angle terms are added into the previous loss function, the model convergence is accelerated, the detection performance of the model on the power tower is improved, meanwhile, LSD line detection is adopted for linear detection, linear screening and least square connection are carried out to obtain candidate power lines, the power line range (effective power line extraction area) is determined by adopting a rotary electric tower positioning frame and increasing an offset amount mode based on the structural information of the power line, finally, the power line range is mixed and disordered by adopting a prediction frame center point, a preset high and high-speed information of an insulator of the YOLOv7 detection, the candidate power line is combined with a projection template according to the projection distance of the power line, and the candidate power line is finally, the candidate power line is transformed according to the projection standard.
Further, as a specific implementation of the method of fig. 1, an embodiment of the present application provides a power line extraction device, as shown in fig. 4, including:
the electric tower coordinate prediction module 401 is configured to obtain a standardized image of a power transmission line corresponding to a power transmission line to be detected, and perform electric tower coordinate prediction on the standardized image of the power transmission line according to a coordinate prediction model to obtain electric tower prediction coordinates;
the tower angle calculating module 402 is configured to perform edge detection on the standardized image of the power transmission line, obtain an edge detection result image, and calculate a tower offset angle based on the edge detection result image;
an effective area determining module 403, configured to determine an electric tower positioning frame in the standardized power transmission line image according to the electric tower prediction coordinates, and determine an effective power line extraction area according to the electric tower positioning frame and the electric tower offset angle;
the power line extraction module 404 is configured to perform line detection on the effective power line extraction area, determine a line perpendicular to the tower offset angle from among the detected lines as a candidate power line, vote the candidate power line according to a preset power line template, and determine a power line of the power transmission line to be detected according to a voting result.
Optionally, the tower angle calculation module 402 is further configured to:
calculating potential power lines and potential power line slopes corresponding to the edge detection result image according to a Hough transformation algorithm, and determining potential power line groups parallel to each other according to the potential power line slopes;
clustering the slopes of the potential power line group according to a K-means++ clustering algorithm, determining the target class with the largest number of slopes contained in a plurality of classes, and obtaining the clustering center slope of the target class as an angle theta 1 According to the angle theta 1 Calculating the electric tower offset angle theta by using an electric tower offset angle calculation formula 2 Wherein, the electric tower offset angle formula of calculation is:
optionally, the active area determining module 403 is further configured to:
according to the direction of the electric tower and the offset angle theta of the electric tower 2 Rotating the electric tower positioning frame, and amplifying the rotated electric tower positioning frame by taking the central point of the rotated electric tower positioning frame as the center according to a preset amplification coefficient to obtain an electric tower offset frame;
and determining a first extending point and a second extending point according to the opposite angles of the electric tower offset frame, and determining an effective power line extraction area according to the first extending point and the second extending point.
Optionally, the power line extraction module 404 is further configured to:
performing linear detection on the effective power line extraction area according to an LSD linear detection algorithm to obtain candidate power line segments to be fitted, and placing the candidate power line segments to be fitted into a line segment pool;
according to the distance between the candidate power line segments to be fitted in the line segment pool, placing the candidate power line segments to be fitted smaller than a preset distance threshold into a candidate power line segment group to be fitted;
and fitting all the candidate power line segments to be fitted in the candidate power line segment group to be fitted according to a least square method to obtain a fitting straight line, and determining the fitting straight line perpendicular to the electric tower offset angle in the fitting straight line as a candidate power line.
Optionally, the tower coordinate prediction module 401 is further configured to:
and carrying out insulator coordinate prediction on the image of the standardized position of the power transmission line according to a coordinate prediction model to obtain an insulator prediction coordinate and an insulator prediction center point.
Optionally, the power line extraction module 404 is further configured to:
and after the electric tower standard image is mapped to the electric transmission line standardized image according to a preset power line template, voting is carried out on the candidate power lines according to the distance between the insulator prediction center point and the candidate power lines, and the power lines of the electric transmission line to be detected are determined according to voting results, wherein the preset power line template is obtained by calculating a homography matrix of the electric transmission line standardized image mapped to the electric tower standard image.
Further, another power line extraction device is provided in an embodiment of the present application, as shown in fig. 5, and the device includes:
the electric tower coordinate prediction module 501 is used for acquiring a standardized image of the electric transmission line corresponding to the electric transmission line to be detected, and performing electric tower coordinate prediction on the standardized image of the electric transmission line according to a coordinate prediction model to obtain electric tower prediction coordinates;
the tower angle calculating module 502 is configured to perform edge detection on the standardized image of the power transmission line, obtain an edge detection result image, and calculate a tower offset angle based on the edge detection result image;
an effective area determining module 503, configured to determine an electric tower positioning frame in the standardized power transmission line image according to the electric tower prediction coordinates, and determine an effective power line extraction area according to the electric tower positioning frame and the electric tower offset angle;
the power line extraction module 504 is configured to perform line detection on the effective power line extraction area, determine a line perpendicular to the tower offset angle from among the detected lines as a candidate power line, vote the candidate power line according to a preset power line template, and determine a power line of the power transmission line to be detected according to a voting result.
The line image processing module 505 is configured to perform gray scale adjustment on pixels in the transmission line standardized image to obtain a transmission line gray scale image, perform histogram equalization processing on the transmission line gray scale image to obtain a transmission line equalized image, and perform denoising processing on the transmission line equalized image according to a gaussian filtering method to obtain a transmission line preprocessed image; on the basis of a horizontal direction gradient template and a vertical direction gradient template contained in a Canny edge detection algorithm, adding a first diagonal direction gradient template and a second diagonal direction gradient template, and calculating gradient values of the preprocessing image of the power transmission line according to the four direction gradient templates; and generating a highest threshold value and a lowest threshold value in the gray values of all pixels of the preprocessing image of the power transmission line according to an iterative algorithm, and reserving edge pixels corresponding to gradient values between the highest threshold value and the lowest threshold value to obtain an edge detection result image.
A prediction model improvement module 506, configured to replace a SiLU activation function in the YOLOv7 model with a mix activation function and replace a CIou loss function in the YOLOv7 model with a SIoU loss function, where the coordinate prediction model is an improved model of the YOLOv7 model.
The prediction model training module 507 is configured to obtain a plurality of data set base images including an electric tower, normalize the data set base images to obtain data set normalized images, and label the electric tower and an insulator included in the electric tower in the data set normalized images by using a preset data labeling tool to generate an electric tower positioning data set; dividing the electric tower positioning data set into an electric tower positioning training data set with a preset first proportion, an electric tower positioning test data set with a preset second proportion and an electric tower positioning verification data set with a preset third proportion in a uniform random sampling mode; and inputting the electric tower positioning training data set, the electric tower positioning test data set and the electric tower positioning verification data set into a coordinate prediction model for training, and obtaining a trained coordinate prediction model.
Optionally, the line image processing module 505 is further configured to:
the gray value of each pixel in the standardized image of the power transmission line is adjusted according to a pixel gray adjustment formula to obtain the gray image of the power transmission line, wherein the pixel gray adjustment formula is as follows:
Gray(i,j)＝0.29*R+0.578*G+0.114*B
gray (i, j) represents the Gray value of the j-th column of the i-th row in the pixel point matrix corresponding to the standardized image of the power transmission line, and R, G and B represent the red channel matrix, the green channel matrix and the blue channel matrix respectively.
Optionally, the line image processing module 505 is further configured to:
equalizing gray scale distribution in the transmission line gray scale image according to a histogram equalization formula to obtain the transmission line equalization image, wherein the histogram equalization formula is as follows:
n represents the sum of pixels in the transmission line graying image, k represents the k-level gray level, n k Representing a gray level r k L represents the total number of gray levels in the transmission line gray-scale image, p r (r) represents the gray level probability density of the transmission line graying image.
Optionally, the tower angle calculation module 502 is further configured to:
calculating potential power lines and potential power line slopes corresponding to the edge detection result image according to a Hough transformation algorithm, and determining potential power line groups parallel to each other according to the potential power line slopes;
clustering the slopes of the potential power line group according to a K-means++ clustering algorithm, determining the target class with the largest number of slopes contained in a plurality of classes, and obtaining the clustering center slope of the target class as an angle theta 1 According to the angle theta 1 Calculating the electric tower offset angle theta by using an electric tower offset angle calculation formula 2 Wherein the electric tower offset angle The degree calculation formula is:
optionally, the effective area determining module 503 is further configured to:
according to the direction of the electric tower and the offset angle theta of the electric tower 2 Rotating the electric tower positioning frame, and amplifying the rotated electric tower positioning frame by taking the central point of the rotated electric tower positioning frame as the center according to a preset amplification coefficient to obtain an electric tower offset frame;
and determining a first extending point and a second extending point according to the opposite angles of the electric tower offset frame, and determining an effective power line extraction area according to the first extending point and the second extending point.
Optionally, the power line extraction module 504 is further configured to:
performing linear detection on the effective power line extraction area according to an LSD linear detection algorithm to obtain candidate power line segments to be fitted, and placing the candidate power line segments to be fitted into a line segment pool;
according to the distance between the candidate power line segments to be fitted in the line segment pool, placing the candidate power line segments to be fitted smaller than a preset distance threshold into a candidate power line segment group to be fitted;
and fitting all the candidate power line segments to be fitted in the candidate power line segment group to be fitted according to a least square method to obtain a fitting straight line, and determining the fitting straight line perpendicular to the electric tower offset angle in the fitting straight line as a candidate power line.
Optionally, the tower coordinate prediction module 501 is further configured to:
and carrying out insulator coordinate prediction on the image of the standardized position of the power transmission line according to a coordinate prediction model to obtain an insulator prediction coordinate and an insulator prediction center point.
Optionally, the power line extraction module 504 is further configured to:
and after the electric tower standard image is mapped to the electric transmission line standardized image according to a preset power line template, voting is carried out on the candidate power lines according to the distance between the insulator prediction center point and the candidate power lines, and the power lines of the electric transmission line to be detected are determined according to voting results, wherein the preset power line template is obtained by calculating a homography matrix of the electric transmission line standardized image mapped to the electric tower standard image.
It should be noted that, for other corresponding descriptions of each functional unit related to the power line extraction device provided in the embodiment of the present application, reference may be made to corresponding descriptions in the methods of fig. 1 to 3, which are not repeated herein.
Based on the above-described method shown in fig. 1 to 3, correspondingly, the embodiment of the present application further provides a storage medium, on which a computer program is stored, which when executed by a processor, implements the above-described power line extraction method shown in fig. 1 to 2.
Based on such understanding, the technical solution of the present application may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.), and includes several instructions for causing a computer device (may be a personal computer, a server, or a network device, etc.) to perform the methods described in various implementation scenarios of the present application.
Based on the method shown in fig. 1 to 3 and the virtual device embodiments shown in fig. 4 and 5, in order to achieve the above objective, the embodiments of the present application further provide a computer device, which may specifically be a personal computer, a server, a network device, etc., where the computer device includes a storage medium and a processor; a storage medium storing a computer program; a processor for executing a computer program to implement the power line extraction method as described above and shown in fig. 1 to 3.
Optionally, the computer device may also include a user interface, a network interface, a camera, radio Frequency (RF) circuitry, sensors, audio circuitry, WI-FI modules, and the like. The user interface may include a Display screen (Display), an input unit such as a Keyboard (Keyboard), etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (e.g., bluetooth interface, WI-FI interface), etc.
It will be appreciated by those skilled in the art that the architecture of a computer device provided in the present embodiment is not limited to the computer device, and may include more or fewer components, or may combine certain components, or may be arranged in different components.
The storage medium may also include an operating system, a network communication module. An operating system is a program that manages and saves computer device hardware and software resources, supporting the execution of information handling programs and other software and/or programs. The network communication module is used for realizing communication among all components in the storage medium and communication with other hardware and software in the entity equipment.
Through the description of the above embodiments, it can be clearly understood by those skilled in the art that the present application may be implemented by means of software plus a necessary general hardware platform, or may be implemented by hardware, and electric tower coordinate prediction is performed on an image of a standardized position of an electric transmission line to be detected according to a coordinate prediction model, so as to obtain electric tower prediction coordinates; performing edge detection on the standardized image of the power transmission line and calculating an offset angle of the electric tower; determining a tower positioning frame according to the tower prediction coordinates, and determining an effective power line extraction area according to the tower positioning frame and the tower offset angle; and carrying out linear detection on the effective power line extraction area, determining a straight line perpendicular to the electric tower offset angle in the detected straight line as a candidate power line, voting the candidate power line according to a preset power line template, and determining the power line of the power transmission line to be detected according to the voting result. Based on the characteristics of aerial images, electric tower prediction coordinates are obtained according to a supervised algorithm (a coordinate prediction model), then edge detection is carried out on a standardized image of a power transmission line so as to detect accurate and clear edges, an electric tower offset angle is calculated according to an edge detection result image, a priori knowledge is combined, a power line is required to be in a range of the electric power tower, the power line is perpendicular to the electric power tower, a straight line is detected in an effective power line extraction area through a straight line detection algorithm (an unsupervised algorithm), finally a straight line approximately perpendicular to the electric tower offset angle is selected as a candidate power line, and the power line is determined according to a preset power line template, so that the accuracy of power line detection is improved.
Those skilled in the art will appreciate that the drawings are merely schematic illustrations of one preferred implementation scenario, and that the modules or flows in the drawings are not necessarily required to practice the present application. Those skilled in the art will appreciate that modules in an apparatus in an implementation scenario may be distributed in an apparatus in an implementation scenario according to an implementation scenario description, or that corresponding changes may be located in one or more apparatuses different from the implementation scenario. The modules of the implementation scenario may be combined into one module, or may be further split into a plurality of sub-modules.
The foregoing application serial numbers are merely for description, and do not represent advantages or disadvantages of the implementation scenario. The foregoing disclosure is merely a few specific implementations of the present application, but the present application is not limited thereto and any variations that can be considered by a person skilled in the art shall fall within the protection scope of the present application.
Claims (11)
1. A power line extraction method, the method comprising:
acquiring a standardized image of a power transmission line corresponding to a power transmission line to be detected, and carrying out electric tower coordinate prediction on the standardized image of the power transmission line according to a coordinate prediction model to obtain electric tower prediction coordinates;
the training method of the coordinate prediction model comprises the following steps:
Acquiring a plurality of data set basic images containing electric towers, carrying out standardization processing on the data set basic images to obtain data set standardization images, and marking the electric towers and insulators contained in the electric towers in the data set standardization images through a preset data marking tool to generate electric tower positioning data sets;
dividing the electric tower positioning data set into an electric tower positioning training data set with a preset first proportion, an electric tower positioning test data set with a preset second proportion and an electric tower positioning verification data set with a preset third proportion in a uniform random sampling mode;
inputting the electric tower positioning training data set, the electric tower positioning test data set and the electric tower positioning verification data set into a coordinate prediction model for training, and obtaining a trained coordinate prediction model;
determining a tower positioning frame in the standardized power transmission line image according to the tower prediction coordinates;
performing edge detection on the standardized image of the power transmission line to obtain an edge detection result image, and calculating an electric tower offset angle based on the edge detection result image;
determining an effective power line extraction area according to the electric tower positioning frame and the electric tower offset angle;
Performing linear detection on the effective power line extraction area, and determining a straight line perpendicular to the tower offset angle in the detected straight lines as a candidate power line;
searching a tower standard image according to tower prediction coordinates in a tower standard template library corresponding to a power transmission line to be detected, wherein the tower standard image and the power transmission line standardized image are images of different visual angles of the same tower;
mapping the insulator in the electric tower standard image into the electric transmission line standardized image through homography, voting the candidate electric power lines according to the distance between the mapped insulator and the candidate electric power lines, and determining the straight line of the electric power line of the electric transmission line to be detected according to the voting result, wherein the homography is realized by using homography matrix, the homography matrix is calculated based on six pairs of insulator matching points, and the six pairs of insulator matching points are selected from the electric tower standard image and the electric transmission line standardized image which correspond to the same electric tower and are mutually different in view angles.
2. The method according to claim 1, wherein performing edge detection on the standardized image of the power transmission line to obtain an edge detection result image includes:
Gray scale adjustment is carried out on pixels in the standardized power transmission line image to obtain a gray scale image of the power transmission line, the gray scale image of the power transmission line is subjected to histogram equalization processing to obtain an equalized image of the power transmission line, and denoising processing is carried out on the equalized image of the power transmission line according to a Gaussian filtering method to obtain a preprocessed image of the power transmission line;
on the basis of a horizontal direction gradient template and a vertical direction gradient template contained in a Canny edge detection algorithm, adding a first diagonal direction gradient template and a second diagonal direction gradient template, and calculating gradient values of the preprocessing image of the power transmission line according to the four direction gradient templates;
and generating a highest threshold value and a lowest threshold value in the gray values of all pixels of the preprocessing image of the power transmission line according to an iterative algorithm, and reserving edge pixels corresponding to gradient values between the highest threshold value and the lowest threshold value to obtain an edge detection result image.
3. The method according to claim 2, wherein the gray-scale adjustment of the pixels in the standardized image of the power transmission line to obtain the gray-scale image of the power transmission line includes:
the gray value of each pixel in the standardized image of the power transmission line is adjusted according to a pixel gray adjustment formula to obtain the gray image of the power transmission line, wherein the pixel gray adjustment formula is as follows:
Gray(i,j)＝0.29*R+0.578*G+0.114*B
Gray (i, j) represents the Gray value of the j-th column of the i-th row in the pixel point matrix corresponding to the standardized image of the power transmission line, and R, G and B represent the red channel matrix, the green channel matrix and the blue channel matrix respectively.
4. The method of claim 2, wherein the histogram equalization processing the transmission line gray scale image to obtain a transmission line equalized image comprises:
equalizing gray scale distribution in the transmission line gray scale image according to a histogram equalization formula to obtain the transmission line equalization image, wherein the histogram equalization formula is as follows:
n represents the sum of pixels in the transmission line graying image, k represents the k-level gray level, n k Representing a gray level r k L represents the total number of gray levels in the transmission line gray-scale image, p r (r) represents the gray level probability density of the transmission line graying image.
5. The method of claim 1, wherein the calculating a tower offset angle based on the edge detection result image comprises:
calculating potential power lines and potential power line slopes corresponding to the edge detection result image according to a Hough transformation algorithm, and determining potential power line groups parallel to each other according to the potential power line slopes;
Clustering the slopes of the potential power line group according to a K-means++ clustering algorithm, determining the target class with the largest number of slopes contained in a plurality of classes, and obtaining the clustering center slope of the target class as an angle theta 1 According to the angle theta 1 Calculating the electric tower offset angle theta by using an electric tower offset angle calculation formula 2 Wherein, the electric tower offset angle formula of calculation is:
6. the method of claim 5, wherein the tower locating frame is rectangular; the determining an effective power line extraction area according to the electric tower positioning frame and the electric tower offset angle comprises the following steps:
according to the direction of the electric tower and the offset angle theta of the electric tower 2 Rotating the electric tower positioning frame, centering around the center point of the rotated electric tower positioning frame, and feeding the rotated electric tower positioning frame according to a preset amplification factorRow amplification is carried out to obtain a tower offset frame;
and determining a first extending point and a second extending point according to the opposite angles of the electric tower offset frame, and determining an effective power line extraction area according to the first extending point and the second extending point.
7. The method of claim 6, wherein the performing the line detection on the effective power line extraction area and determining a line perpendicular to the tower offset angle among the detected lines as a candidate power line includes:
Performing linear detection on the effective power line extraction area according to an LSD linear detection algorithm to obtain candidate power line segments to be fitted, and placing the candidate power line segments to be fitted into a line segment pool;
according to the distance between the candidate power line segments to be fitted in the line segment pool, placing the candidate power line segments to be fitted smaller than a preset distance threshold into a candidate power line segment group to be fitted;
and fitting all the candidate power line segments to be fitted in the candidate power line segment group to be fitted according to a least square method to obtain a fitting straight line, and determining the fitting straight line perpendicular to the electric tower offset angle in the fitting straight line as a candidate power line.
8. The method according to any one of claims 1 to 7, wherein,
the coordinate prediction model is an improved model of the YOLOv7 model, and is obtained by replacing a SiLU activation function in the YOLOv7 model with a mix activation function and replacing a CIou loss function in the YOLOv7 model with a SIoU loss function.
9. A power line extraction device, the device comprising:
the electric tower coordinate prediction module is used for acquiring a standardized image of the electric transmission line corresponding to the electric transmission line to be detected, carrying out electric tower coordinate prediction on the image of the standardized position of the electric transmission line according to a coordinate prediction model to obtain electric tower prediction coordinates, wherein the training method of the coordinate prediction model comprises the steps of acquiring a plurality of data set basic images containing electric towers, carrying out standardization processing on the data set basic images to obtain a data set standardized image, labeling the electric towers and insulators contained in the electric towers in the data set standardized image through a preset data labeling tool, generating a tower positioning data set, dividing the tower positioning data set into a preset first proportion of tower positioning training data set, a preset second proportion of tower positioning test data set and a preset third proportion of tower positioning verification data set without intersection by uniformly and randomly sampling the tower positioning data set, and inputting the tower positioning training data set, the tower positioning test data set and the tower positioning verification data set into a coordinate prediction model for training to obtain a trained coordinate prediction model;
The effective area determining module is used for determining a tower positioning frame in the standardized power transmission line image according to the tower prediction coordinates;
the electric tower angle calculation module is used for carrying out edge detection on the standardized image of the power transmission line to obtain an edge detection result image, and calculating an electric tower offset angle based on the edge detection result image;
the effective area determining module is further used for determining an effective power line extraction area according to the electric tower positioning frame and the electric tower offset angle;
the power line extraction module is used for detecting straight lines of the effective power line extraction area and determining straight lines perpendicular to the electric tower offset angle in the detected straight lines as candidate power lines;
the power line extraction module is further used for searching a power tower standard image according to the power tower prediction coordinates in a power tower standard template library corresponding to the power transmission line to be detected, wherein the power tower standard image and the power transmission line standardized image are images of different visual angles of the same power tower;
the power line extraction module is further used for mapping the insulators in the electric tower standard image into the power transmission line standardized image through homography transformation, voting is conducted on the candidate power lines according to the distance between the mapped insulators and the candidate power lines, and the line where the power line of the power transmission line to be detected is located is determined according to the voting result, wherein homography transformation is achieved through homography matrixes, the homography matrixes are calculated based on six pairs of insulator matching points, and the six pairs of insulator matching points are selected from the electric tower standard image and the power transmission line standardized image which correspond to the same electric tower and are different in view angles.
10. A storage medium having stored thereon a computer program, which when executed by a processor, implements the method of power line extraction of any one of claims 1 to 8.
11. A computer device comprising a storage medium, a processor and a computer program stored on the storage medium and executable on the processor, characterized in that the processor implements the method of power line extraction according to any one of claims 1 to 8 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310540202.4A CN116843909B (en) | 2023-05-12 | 2023-05-12 | Power line extraction method and device, storage medium and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310540202.4A CN116843909B (en) | 2023-05-12 | 2023-05-12 | Power line extraction method and device, storage medium and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116843909A CN116843909A (en) | 2023-10-03 |
| CN116843909B true CN116843909B (en) | 2024-03-08 |
Family
ID=88169440
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310540202.4A Active CN116843909B (en) | 2023-05-12 | 2023-05-12 | Power line extraction method and device, storage medium and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116843909B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104978580A (en) * | 2015-06-15 | 2015-10-14 | 国网山东省电力公司电力科学研究院 | Insulator identification method for unmanned aerial vehicle polling electric transmission line |
| CN108109155A (en) * | 2017-11-28 | 2018-06-01 | 东北林业大学 | A kind of automatic threshold edge detection method based on improvement Canny |
| CN111444809A (en) * | 2020-03-23 | 2020-07-24 | 华南理工大学 | Power transmission line abnormal target detection method based on improved YO L Ov3 |
| KR20210090489A (en) * | 2020-01-10 | 2021-07-20 | 한국전력공사 | CAMERA DRONE for INSPECTING TRANSMISSION TOWER and SHOOTING METHOD of POWER LINE or TRANSMISSION TOWER |
| CN113762276A (en) * | 2021-08-17 | 2021-12-07 | 湖南大学 | Method for detecting angle of ladder in electric power operation |
| CN114022761A (en) * | 2021-10-26 | 2022-02-08 | 国网辽宁省电力有限公司经济技术研究院 | Detection and positioning method and device for power transmission line tower based on satellite remote sensing image |
| CN114114314A (en) * | 2021-11-08 | 2022-03-01 | 国网福建省电力有限公司漳州供电公司 | Power transmission line inspection detection system and detection method based on laser point cloud |
| CN115239667A (en) * | 2022-07-21 | 2022-10-25 | 淮阴工学院 | Tire qualification detection method based on multi-view camera |
| CN115564725A (en) * | 2022-09-29 | 2023-01-03 | 四川省公路规划勘察设计研究院有限公司 | Slope monitoring method and device based on intelligent image identification |
| CN115731164A (en) * | 2022-09-14 | 2023-03-03 | 常州大学 | Insulator defect detection method based on improved YOLOv7 |
| CN115761404A (en) * | 2022-11-18 | 2023-03-07 | 云南电网有限责任公司电力科学研究院 | Training method, using method and training device of YOLOv5s model |
| CN115841669A (en) * | 2022-09-30 | 2023-03-24 | 国网山西省电力公司大同供电公司 | Pointer instrument detection and reading identification method based on deep learning technology |
| CN115841633A (en) * | 2022-12-05 | 2023-03-24 | 深圳市优智创芯科技有限公司 | Power tower and power line associated correction power tower and power line detection method |
| CN115861186A (en) * | 2022-11-15 | 2023-03-28 | 佳源科技股份有限公司 | Electric power tower detection method and device based on deep learning and unmanned aerial vehicle equipment |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017021751A1 (en) * | 2015-08-06 | 2017-02-09 | Accenture Global Services Limited | Vegetation management for power line corridor monitoring using computer vision |
| CN106356757B (en) * | 2016-08-11 | 2018-03-20 | 河海大学常州校区 | A kind of power circuit unmanned plane method for inspecting based on human-eye visual characteristic |
-
2023
- 2023-05-12 CN CN202310540202.4A patent/CN116843909B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104978580A (en) * | 2015-06-15 | 2015-10-14 | 国网山东省电力公司电力科学研究院 | Insulator identification method for unmanned aerial vehicle polling electric transmission line |
| CN108109155A (en) * | 2017-11-28 | 2018-06-01 | 东北林业大学 | A kind of automatic threshold edge detection method based on improvement Canny |
| KR20210090489A (en) * | 2020-01-10 | 2021-07-20 | 한국전력공사 | CAMERA DRONE for INSPECTING TRANSMISSION TOWER and SHOOTING METHOD of POWER LINE or TRANSMISSION TOWER |
| CN111444809A (en) * | 2020-03-23 | 2020-07-24 | 华南理工大学 | Power transmission line abnormal target detection method based on improved YO L Ov3 |
| CN113762276A (en) * | 2021-08-17 | 2021-12-07 | 湖南大学 | Method for detecting angle of ladder in electric power operation |
| CN114022761A (en) * | 2021-10-26 | 2022-02-08 | 国网辽宁省电力有限公司经济技术研究院 | Detection and positioning method and device for power transmission line tower based on satellite remote sensing image |
| CN114114314A (en) * | 2021-11-08 | 2022-03-01 | 国网福建省电力有限公司漳州供电公司 | Power transmission line inspection detection system and detection method based on laser point cloud |
| CN115239667A (en) * | 2022-07-21 | 2022-10-25 | 淮阴工学院 | Tire qualification detection method based on multi-view camera |
| CN115731164A (en) * | 2022-09-14 | 2023-03-03 | 常州大学 | Insulator defect detection method based on improved YOLOv7 |
| CN115564725A (en) * | 2022-09-29 | 2023-01-03 | 四川省公路规划勘察设计研究院有限公司 | Slope monitoring method and device based on intelligent image identification |
| CN115841669A (en) * | 2022-09-30 | 2023-03-24 | 国网山西省电力公司大同供电公司 | Pointer instrument detection and reading identification method based on deep learning technology |
| CN115861186A (en) * | 2022-11-15 | 2023-03-28 | 佳源科技股份有限公司 | Electric power tower detection method and device based on deep learning and unmanned aerial vehicle equipment |
| CN115761404A (en) * | 2022-11-18 | 2023-03-07 | 云南电网有限责任公司电力科学研究院 | Training method, using method and training device of YOLOv5s model |
| CN115841633A (en) * | 2022-12-05 | 2023-03-24 | 深圳市优智创芯科技有限公司 | Power tower and power line associated correction power tower and power line detection method |
Non-Patent Citations (6)
| Title |
|---|
| Development of power transmission line detection technology based on unmanned aerial vehicle image vision;Baoyu Xu 等;《SN Applied Sciences》;1-15 * |
| UAV Transmission Line Inspection Object Recognition based on Mask R-CNN;Yi Liu 等;《CISAT 2019》;1-9 * |
| 基于Hough变换的目标数据处理算法研究;陈镜 等;《无线电工程》;第50卷(第5期);357-361 * |
| 基于图像处理的无人机电力巡线航向偏差检测;杨坚 等;《浙江电力》;第40卷(第9期);105-110 * |
| 改进的Canny算法在工业机器人视觉系统中的应用;刘亮亮 等;《河南科技》;第42卷(第3期);第31-34页，第2-3.2节 * |
| 无人机影像中直线类电力线检测算法研究;李朋月 等;《电子测量技术》;第42卷(第19期);148-153 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116843909A (en) | 2023-10-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| DE102020100684B4 (en) | MARKING OF GRAPHICAL REFERENCE MARKERS | |
| WO2021000702A1 (en) | Image detection method, device, and system | |
| CN110569756B (en) | Face recognition model construction method, recognition method, device and storage medium | |
| CN112686812B (en) | Bank card inclination correction detection method and device, readable storage medium and terminal | |
| CN111275129A (en) | Method and system for selecting image data augmentation strategy | |
| CN107103320B (en) | Embedded medical data image identification and integration method | |
| CN109086714A (en) | Table recognition method, identifying system and computer installation | |
| CN111723721A (en) | Three-dimensional target detection method, system and device based on RGB-D | |
| CN109657612B (en) | Quality sorting system based on facial image features and application method thereof | |
| CN107229917A (en) | A kind of several remote sensing image general character well-marked target detection methods clustered based on iteration | |
| CN108764358A (en) | A kind of Terahertz image-recognizing method, device, equipment and readable storage medium storing program for executing | |
| CN111383244B (en) | Target detection tracking method | |
| CN109711268B (en) | Face image screening method and device | |
| CN104217459B (en) | A kind of spheroid character extracting method | |
| CN107506795A (en) | A kind of local gray level histogram feature towards images match describes sub- method for building up and image matching method | |
| CN108875504B (en) | Image detection method and image detection device based on neural network | |
| CN112883881B (en) | Unordered sorting method and unordered sorting device for strip-shaped agricultural products | |
| CN111597845A (en) | Two-dimensional code detection method, device and equipment and readable storage medium | |
| CN111709305B (en) | Face age identification method based on local image block | |
| CN116543325A (en) | Unmanned aerial vehicle image-based crop artificial intelligent automatic identification method and system | |
| CN107392211B (en) | Salient target detection method based on visual sparse cognition | |
| CN113436735A (en) | Body weight index prediction method, device and storage medium based on face structure measurement | |
| Lou et al. | Smoke root detection from video sequences based on multi-feature fusion | |
| CN104268550A (en) | Feature extraction method and device | |
| CN116843909B (en) | Power line extraction method and device, storage medium and computer equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |