CN113205063A - Visual identification and positioning method for defects of power transmission conductor - Google Patents
Visual identification and positioning method for defects of power transmission conductor Download PDFInfo
- Publication number
- CN113205063A CN113205063A CN202110546688.3A CN202110546688A CN113205063A CN 113205063 A CN113205063 A CN 113205063A CN 202110546688 A CN202110546688 A CN 202110546688A CN 113205063 A CN113205063 A CN 113205063A
- Authority
- CN
- China
- Prior art keywords
- image
- transmission conductor
- power transmission
- defect
- defects
- 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.)
- Pending
Links
- 230000007547 defect Effects 0.000 title claims abstract description 136
- 230000005540 biological transmission Effects 0.000 title claims abstract description 126
- 239000004020 conductor Substances 0.000 title claims abstract description 110
- 238000000034 method Methods 0.000 title claims abstract description 73
- 230000000007 visual effect Effects 0.000 title claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 31
- 238000003708 edge detection Methods 0.000 claims abstract description 23
- 238000001514 detection method Methods 0.000 claims abstract description 21
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims description 31
- 238000002372 labelling Methods 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 17
- 230000002708 enhancing effect Effects 0.000 claims description 13
- 238000001914 filtration Methods 0.000 claims description 12
- 238000003062 neural network model Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000009432 framing Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 3
- 230000004807 localization Effects 0.000 claims 2
- 230000009286 beneficial effect Effects 0.000 abstract 1
- 230000006870 function Effects 0.000 description 19
- 230000008569 process Effects 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000005764 inhibitory process Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- 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
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- 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
- G06V10/30—Noise filtering
-
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Abstract
The application relates to the technical field of computer vision, and provides a visual identification and positioning method for defects of a power transmission conductor, wherein the visual identification and positioning method comprises the steps of firstly utilizing an improved LBP characteristic value to carry out gray processing on an image, then carrying out denoising and enhancement on the image, and adopting Canny edge detection to obtain an edge image of the power transmission conductor; extracting the texture of the lead by using a Hough linear detection algorithm, fitting the central line of the lead by using a least square algorithm, and extracting a lead part as an interested area; and finally, automatically identifying the defects of the power transmission conductor through an improved YOLOv3 convolutional neural network model, and obtaining the types of the defects of the power transmission conductor and the position information of the defects in the images. The visual identification and positioning method can realize automatic identification and positioning of the defects of the transmission conductors, is beneficial to reducing the workload of visual identification, reduces the labor cost, improves the accuracy and efficiency of automatic identification, and further ensures the stable operation of a power grid.
Description
Technical Field
The application relates to the technical field of computer vision, in particular to a visual identification and positioning method for defects of a transmission conductor.
Background
The transmission line mainly undertakes the electric energy transmission task and is an important component of a national power grid. Because the power transmission line has a wide erection area and the terrain and climate of passing areas are complex and changeable, the transmission conductor is easy to generate faults such as conductor strand breakage, conductor damage and the like under the long-time influence of the external environment. If these faults are not maintained in time, the line break and short circuit faults are easily caused, and further a large-area unplanned power failure is caused. Therefore, the regular polling work is very critical, can ensure that the power transmission line stably and reliably runs, can avoid unnecessary economic loss, and plays a vital role in national power utilization safety and economic stable development.
Early patrol and examine work and mainly accomplish by artifical patrol and examine the mode, the staff is direct through the naked eye observation promptly, seeks and discern the fault point on the circuit, and this kind of mode of patrolling and examining is inefficient, the degree of difficulty is big and the reliability is relatively weak. Along with the progress of science and technology, the neotype mode of patrolling and examining replaces the manual work gradually and patrols and examines, and this kind of neotype mode of patrolling and examining mainly shoots transmission line image through carrying image acquisition instruments such as high definition camera or infrared camera, looks over and analyzes the video image who gathers by the staff again, observes the situation of transmission line typical part and finds out the fault point.
However, by adopting the novel inspection mode, the image data acquired by the video terminal is huge and high in repeatability, and the massive image data of the power transmission line still needs to be checked by the naked eyes of workers, so that the workload is huge, the situation of misjudgment or missed judgment is easy to occur, and the potential safety hazard on the power transmission line still cannot be timely and accurately found.
Disclosure of Invention
In order to overcome the defects of the prior art, the application aims to provide a visual identification and positioning method for the defects of the transmission conductor, and the automation degree, the data degree and the real-time performance of the transmission line inspection can be improved.
In order to achieve the above object, in one aspect, the present application provides a method for visually identifying and locating a defect of a power transmission conductor, which specifically includes:
and acquiring a high-definition image of the power transmission conductor.
Carrying out graying processing on the high-definition image by utilizing the improved local binary pattern characteristic value to obtain a grayed image; the improved local binary pattern characteristic value is obtained by expanding the contrast range of the pixel point to a circular area with any radius and according to a Laplacian operator.
And denoising and enhancing the grayed image to obtain an enhanced image, and performing edge detection on the enhanced image by using a Canny edge detection algorithm to obtain an edge image.
And after extracting the texture of the power transmission conductor from the edge image by using a Hough linear detection algorithm, fitting the texture center line of the power transmission conductor by using a least square method, obtaining a power transmission conductor fitting image by using the texture center line of the power transmission conductor, and taking the power transmission conductor fitting image as an interested area.
And automatically identifying the region of interest through a pre-established defect identification model to obtain the defect type of the region of interest, and framing the position information of the region of interest to which the defect type belongs.
The defect identification model comprises a transmission conductor defect type database, and can match the region of interest with the transmission conductor defect type database to determine whether the region of interest has defects and the defect types of the defects; the defect recognition model is obtained by training a modified YOLOv3 convolutional neural network, which comprises introducing a DropBlock layer and modifying a loss function.
Further, the graying processing is performed on the high-definition image by using the improved local binary pattern characteristic value, and the specific method includes:
step S21: and acquiring all pixel points of the high-definition image.
Step S22: selecting any pixel point as a central pixel point, determining a circular area of the central pixel point according to a preset radius, and selecting a set number of pixel points in the circular area as neighborhood pixel points.
Step S23: obtaining a Laplace operator of the central pixel point according to the gray value of the central pixel point; and solving the Laplacian operator of the neighborhood pixel points according to the gray value of the neighborhood pixel points.
Step S24: and calculating a local binary pattern characteristic value of the central pixel point according to the Laplacian of the central pixel point and the Laplacian of the neighborhood pixel points.
Step S25: and carrying out gray level normalization on the high-definition image by using the local binary pattern characteristic values of all the pixel points.
Further, the specific method for establishing the defect identification model comprises the following steps:
step S31: and acquiring a transmission conductor defect image in a transmission conductor defect type database.
Step S32: and dividing the transmission conductor defect image into a training set and a testing set.
Step S33: and repeatedly training the training set through a back propagation algorithm to establish a preliminary neural network model.
Step S34: and verifying the precision of the preliminary neural network model by using the test set, and if the precision of the preliminary neural network model meets the requirement, generating a defect identification model.
Further, the specific method for improving the loss function is as follows:
step S41: and calculating the intersection area and the union area of the prediction box and the marking box.
Step S42: and obtaining the intersection and union ratio of the prediction frame and the marking frame according to the intersection area and the union area, wherein the intersection and union ratio is the ratio of the intersection area and the union area of the prediction frame and the marking frame.
Step S43: and calculating the minimum closure area of the prediction frame and the labeling frame, wherein the minimum closure area is the area of the minimum frame simultaneously containing the prediction frame and the labeling frame.
Step S44: and calculating the proportion of the region which does not belong to the prediction frame or the labeling frame in the minimum closure region to the minimum closure region according to the area of the minimum closure region and the area of the union to obtain the non-odd proportion.
Step S45: obtaining a generalized intersection ratio according to the intersection ratio and the non-:
in the formula, GIoU is a generalized intersection ratio, IoU is an intersection ratio of a prediction box and a labeling box, D is a minimum closure area, and U is a union area of the prediction box and the labeling box.
Step S46: obtaining a loss function according to the generalized intersection ratio, wherein the loss function is as follows: l isGIoU＝1-GIoU。
Further, the DropBlock layer includes two parameters, which are a size Block _ size of a Block to be discarded and a number γ of active units to be discarded, respectively, and the calculation formula of γ is as follows:
where keep _ prob is the probability that the active cell in the conventional Dropout layer is reserved, and flat _ size is the size of the feature map.
Furthermore, a wiener filtering method is adopted for denoising the gray images.
Further, histogram equalization is adopted for enhancing the grayscale image.
Further, the Canny edge detection adopts a maximum inter-class difference method.
Further, the power conductor defect type database includes: single strand wire breakage, multiple strand wire breakage, and wire breakage.
In a second aspect, the present application further provides a system for visually identifying and positioning defects of a power transmission conductor, which specifically includes:
the image preprocessing unit is used for preprocessing a high-definition image and comprises a gray processing module, an image denoising and enhancing module and an edge detection module; the gray processing module performs gray processing on the high-definition image by using the improved local binary pattern characteristic value to obtain a gray image; the image denoising and enhancing module is used for carrying out filtering processing on the grayed image through a wiener filter and further enhancing the grayed image after the filtering processing by utilizing histogram equalization to obtain an enhanced image; and the edge detection module is used for carrying out edge detection on the enhanced image through a Canny edge detection algorithm to obtain an edge image.
The lead extraction unit is used for extracting a part of the power transmission lead as an interested area and comprises a straight line detection module, a fitting central line module and an interested area extraction module; the line detection module is used for extracting the texture of the power transmission conductor from the edge image through a Hough line detection algorithm; the fitting central line module is used for fitting the texture central line of the power transmission conductor by a least square method; and the interesting region extracting module is used for acquiring a transmission conductor fitting image according to the transmission conductor texture center line and taking the transmission conductor fitting image as an interesting region.
The classification and positioning unit is used for automatically identifying the region of interest through a pre-established defect identification model, obtaining the defect type of the region of interest and framing the position information of the region of interest to which the defect type belongs; the defect identification model comprises a transmission conductor defect type database, and can match the region of interest with the transmission conductor defect type database to determine whether the region of interest has defects and the defect types of the defects; the defect recognition model is obtained by training a modified YOLOv3 convolutional neural network, which comprises introducing a DropBlock layer and modifying a loss function.
The application provides a visual identification and positioning method for defects of a power transmission conductor, the types of the defects of the power transmission conductor are obtained through the combined application of an image processing technology and a neural network model, the position information of the defects in an image is framed, and meanwhile, the identification precision and the position information of the defects of the power transmission conductor are correspondingly stored, so that the automatic identification and positioning of the defects of the power transmission conductor can be realized, the workload of visual identification is favorably reduced, the labor cost is reduced, the automatic identification accuracy and efficiency are improved, and the stable operation of a power grid is further ensured.
Drawings
In order to more clearly explain the technical solution of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious to those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a method for visually identifying and locating a defect of a power transmission conductor according to an embodiment of the present application;
fig. 2 is a schematic diagram illustrating a transmission conductor image graying processing flow provided in an embodiment of the present application;
fig. 3 is a schematic flow chart of creating a defect identification model according to an embodiment of the present application;
fig. 4 is a schematic diagram illustrating a transmission conductor defect identification process according to an embodiment of the present application;
fig. 5 is a schematic diagram of changes in the clustering number and the sum of squares of errors according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be fully and clearly described below with reference to the drawings in the embodiments of the present application. It is to be understood that the embodiments described are only a few embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
In recent years, with the combination of a computer vision technology and a convolutional neural network becoming more and more compact, researchers increasingly apply the computer vision technology to line inspection work, and the automatic identification and positioning of the defects of the transmission conductor in the image are realized, so that the video image is detected instead of a worker.
Referring to fig. 1, a flow chart of a method for visually identifying and locating a power transmission conductor defect according to an embodiment of the present application is schematically shown. A first aspect of an embodiment of the present application provides a visual identification and positioning method for a power transmission conductor defect, which specifically includes:
step S1: and acquiring a high-definition image of the power transmission conductor.
Step S2: the method comprises the steps of utilizing improved local binary pattern characteristic values to conduct graying processing on a high-definition image to obtain a grayed image; the improved local binary pattern characteristic value is obtained by expanding the contrast range of the pixel point to a circular area with any radius and according to a Laplacian operator.
LBP (local Binary patterns), which is a commonly used descriptor of image local texture features in the field of visual inspection at present, has the principle that Binary numbers are obtained by comparing the gray values of each pixel point and its neighborhood pixel points, and then the Binary numbers are converted into decimal numbers to obtain the LBP value of the pixel.
In order to adapt to texture characteristics of various situations and meet the requirements of gray scale and rotation invariance, the embodiment of the application provides an improved LBP operator. The method comprises the steps of firstly calculating the Laplacian of each pixel point, further expanding the contrast range of the pixel points from a square 3 x 3 neighborhood to a circular arbitrary neighborhood, and finally obtaining the LBP characteristic values of P sampling points in the circular area with arbitrary radius. The gray level image obtained by the improved LBP characteristic is basically not influenced by brightness, and the robustness is stronger.
Referring to fig. 2, a schematic diagram of a transmission conductor image graying processing flow provided in the embodiment of the present application is shown. In an embodiment of the present application, the graying processing is performed on the high-definition image by using an improved local binary pattern feature value, and a specific method includes:
step S201: and acquiring all pixel points of the high-definition image.
Step S202: selecting any pixel point as a central pixel point, and setting the central pixel point (x)c,yc) Has a gray value of icDetermining a circular area of the central pixel point according to a preset radius, and selecting a set number of pixel points in the circular area as neighborhood pixel points, wherein the gray value of each neighborhood pixel point is in(n＝0,1,...,p-1)。
Step S203: according to the gray value i of the central pixel pointcAnd calculating the Laplacian l of the central pixel pointc(ii) a And according to the gray value i of the neighborhood pixel pointn(n-0, 1.., p-1), and solving a laplacian operator l of the neighborhood pixel pointn(n＝0,1,...,p-1)。
The laplacian is a simpler isotropic differential operator with rotational invariance, and the laplacian of a two-dimensional image function f (x) is defined as follows:
in order to better process two-dimensional images, the discrete expression is as follows:
step S204: according to the centerLaplacian operator l of pixel pointscAnd the Laplacian l of the neighborhood pixelsn(n-0, 1.., p-1), calculating the LBP feature value of the center pixel point.
In the embodiment of the present application, the LBP value is calculated according to the following principle: and judging the Laplacian of the central pixel point, if the Laplacian of any neighborhood pixel point is smaller than the Laplacian of the central pixel point in the circular area, marking the neighborhood pixel point as 0, otherwise, marking as 1. Therefore, p-1 neighborhood pixel points in the circular area are marked to obtain p-1 bit binary numbers, and the LBP characteristic value of the central pixel point is obtained according to the p-1 bit binary numbers.
Calculating Laplacian operators of all neighborhood pixels by formula (2), so that a central pixel (x)c,yc) The LBP characteristic value of (A) can be obtained by the following formula:
wherein the content of the first and second substances,
step S205: and carrying out gray level normalization on the high-definition image by using the local binary pattern characteristic values of all the pixel points.
Step S3: and denoising and enhancing the grayed image to obtain an enhanced image, and performing edge detection on the enhanced image by using a Canny edge detection algorithm to obtain an edge image.
Further, in some embodiments of the present application, a wiener filtering method is used for denoising the grayed image.
The wiener filter, also called minimum mean square error filter, belongs to discrete FIR filter, and mainly solves the input image f (x, y) and output image
The optimal filter is designed according to the minimum mean square error, the noise existing in the image is eliminated, and the specific calculation formula is expressed as formula (5):
the noise intensity can be estimated by adopting a wiener filtering method, so that an estimated value of noise is obtained, an optimal filter is adaptively designed according to the noise, the image edge and other high-frequency parts can be reserved, and white Gaussian noise can be removed.
Further, in some embodiments of the present application, after performing denoising processing by using a wiener filtering method, the method further includes: and performing enhancement processing on the gray-scale image by histogram equalization.
The gray levels with dense distribution in the image are widened, and the gray levels with sparse distribution are compressed, so that the pixel values of the image are uniformly distributed in the whole gray range, the range of the pixel values is expanded, and the image contrast is improved. Let the histogram equalized gray scale mapping transform be as shown in equation (6):
s＝T(r)，0≤r≤L-1 (6)
where r is the input gray scale value of the image, s is the output gray scale value of the image, and T is the mapping of input and output, in the gray scale space [0, L-1 ].
The gray level of the image describes a random variable using a probability density function. Let p ber(r)，ps(r) probability density function representing random variables r and s, i.e. pr(r) is the probability density function of the gray level of the original image before histogram equalization, ps (r) is the probability density function of the gray level of the output image to be obtained, and the formula (7) can be obtained:
order:
then formula (9) is obtained:
the formula (9) is introduced into the formula (7) to obtain:
further, in some embodiments of the present application, the Canny edge detection uses the maximum inter-class difference method. The idea of the Canny operator is as follows: the method comprises the steps of firstly carrying out convolution filtering on an image by using a first-order directional derivative of a Gaussian equation, and then searching local maximum values of image gradient in the filtered image, wherein the maximum value points are edge points of the image. Since the threshold processing and the connection analysis are used to detect and connect the edges, the setting of the threshold determines the quality of the detection.
Specifically, in some embodiments of the present application, the threshold is set by an adaptive threshold-value-greater-body method. The self-adaptive threshold Otsu method is also called maximum inter-class difference method, and the principle is to divide the background and the foreground into two classes by utilizing the characteristics of a gray histogram, and obtain the optimal threshold by calculating the maximum inter-class difference. And setting the gray value of the image pixel point (x, y) as I (x, y) and the gray value t as an initial threshold. T is equal to or less than I (x, y) and accounts for the image proportion w0Mean value of gray level u0；I(x,y)>t is background pixel point, and the image proportion of t is w1Mean value of gray scale u1(ii) a The overall image gray level mean value u is calculated as shown in formula (11):
u＝w0(t)u0(t)+w1(t)u1(t) (11)
the inter-class error σ between background and foreground is calculated as shown in equation (12):
σ(t)＝w0(t)[u0(t)-u]2+w1(t)[u1(t)-u]2 (12)
the optimal threshold value T is calculated as shown in (13):
T＝argmax(σ(t))t∈[0,255] (13)
step S4: and after extracting the texture of the power transmission conductor from the edge image by using a Hough linear detection algorithm, fitting the texture center line of the power transmission conductor by using a least square method, obtaining a fitted image of the power transmission conductor by using the texture center line of the power transmission conductor, and taking the fitted image of the power transmission conductor as an interested area to remove the background interference of the power transmission line.
The background environment of the power transmission line is complex, in order to improve the detection accuracy, a region of interest (ROI), namely a lead part, needs to be extracted, and the method comprises the following specific steps:
step S401: and (4) performing Hough straight line detection on the edge detection image, setting constraint conditions for the straight line, removing the straight line of the image background edge, and screening the wire texture.
Hough straight line detection is an important geometric shape feature extraction technology in an image processing technology, and the principle of the Hough straight line detection is to convert a straight line detection problem of an image space into a voting problem of a midpoint of the Hough space. Each straight line of the rectangular coordinate system corresponds to a unique point (theta, rho) in the polar coordinate system, and the expression formula is shown as (14):
ρ＝x cosθ+y sinθ (14)
the Hough space takes (theta, rho) as a parameter, so that points on a straight line in the image space are all sinusoidal curves in the Hough space, and the points of the same straight line in the image space intersect with a unique point (theta, rho) in the Hough space. It is possible to detect a straight line in an image by setting a voting threshold value of a voting point in the Hough space.
The following constraints are empirically designed:
1) the threshold range of the straight line voting point is: t is more than 65 and less than 90;
2) the slope of the line ranges from (0.2, 0.5), i.e., the angle of the line with the horizontal axis ranges from about (11.4, 28.7);
thereby screening out the straight lines at the wire texture.
Step S402: and performing straight line fitting on the midpoint of the texture straight line by using a least square method to obtain a wire center line, wherein the specific calculation process is as follows:
the high-voltage transmission line central line equation is shown as the formula (15):
x＝a0+a1y (15)
in the formula, a0、a1Are the equation parameters.
Minimizing discrete points (x)i,yiThe weighted sum of squares of the horizontal deviation from the centerline, as shown in equation (16):
obtaining the following formulae (17) and (18):
formulae (19) and (20) are available:
finishing to obtain a0And a1Is estimated value of
And
the method specifically comprises the following steps:
and fitting the central point of the line texture straight line by using the formula (21) and the formula (22) to obtain the central line of the line.
Step S403: and taking the height of the upper and lower 50 pixels of the central line of the conducting wire as the height and the width of the original image as the width to obtain an ROI rectangular frame with the resolution of 704 x 100, taking an area image framed by the rectangular frame as a power transmission conducting wire fitting image, and taking the power transmission conducting wire fitting image as an ROI of the interested area, thereby finishing the extraction of the power transmission conducting wire.
Step S5: and automatically identifying the region of interest through a pre-established defect identification model to obtain the defect type of the region of interest, and framing the position information of the region of interest to which the defect type belongs.
The defect identification model comprises a transmission conductor defect type database, and can match the region of interest with the transmission conductor defect type database to determine whether the region of interest has defects and the defect types of the defects; the defect recognition model is obtained by training a modified YOLOv3 convolutional neural network, which comprises introducing a DropBlock layer and modifying a loss function.
Further, in some embodiments of the present application, referring to fig. 3, a method for establishing the defect identification model specifically includes:
step S501: and acquiring a transmission conductor defect image in a transmission conductor defect type database.
Step S502: and dividing the transmission conductor defect image into a training set and a testing set.
Step S503: and repeatedly training the training set through a back propagation algorithm to establish a preliminary neural network model.
Step S504: and verifying the precision of the preliminary neural network model by using the test set, and if the precision of the preliminary neural network model meets the requirement, generating a defect identification model.
After the training, testing the defect images of the power transmission conductors in the training set, wherein the test result comprises: 1) category and location information for each defect in the transmission conductor defect type database; 2) and calculating the accuracy of the test result, and judging whether the accuracy of the test result is greater than a preset threshold value. And finally, using the training set and the test set for training the network, checking the training error and the test error, and if the training error and the test error are reduced, representing reasonable convergence to obtain an optimal defect identification model.
Referring to fig. 4, a schematic diagram of a transmission conductor defect identification process provided in the embodiment of the present application is shown, where the automatic conductor defect identification and location model based on the improved YOLOv3 convolutional neural network in the embodiment of the present application specifically includes:
step S51: and establishing a power transmission conductor defect database.
Before training with the YOLOv3 convolutional neural network, we needed to build a transmission line defect database by themselves because there was no database of transmission line defects on the network. Each image in the sample set has a rating label for the target box and the component. The component level labels are
Wherein i represents the sequence number of the target box, CiIndicating the class of the power conductor defect,
indicating the location point of the transmission conductor defect target box.
Further, in some embodiments of the present application, three types of marked power conductor defects are selected, which are: single strand wire breakage, multiple strand wire breakage, and wire breakage.
Step S52: and (4) improving the prior anchor box by using a K-means + + algorithm.
The priori anchor frame mechanism leads in anchor frames with different sizes and different widths and heights as the priori frame, so that the boundary frame does not need to be predicted directly from zero in the training process, but the boundary frame is predicted on the basis of a certain anchor frame. Since the embodiment of the application identifies the three types of the conductor defects in the transmission conductor defect database, a more applicable aspect ratio needs to be selected for the characteristics of the target to be detected. In order to find the best default box of the sample, the embodiment of the application derives the labeling information of the sample database, calculates the aspect ratio of each labeling box, and obtains the aspect ratio clustering center which can represent the characteristics of the target most by using a K-means + + algorithm.
The K-means + + is an improved method of a classic clustering algorithm K-means, the first step of an original K-means algorithm is to randomly select a clustering center from a database, the K-means + + algorithm firstly randomly selects one clustering center, the distances between the rest samples and the clustering center are calculated, and the probability that the farther away sample is selected as the next clustering center is higher. By the method, a certain distance exists between the generated initial clustering centers, so that a better effect is achieved, and the result error is obviously reduced. The specific algorithm process is as follows:
step S521: one sample was randomly selected as the initial cluster center.
Step S522: after the initial clustering centers are selected, the shortest distance between the n samples and the current existing clustering centers (namely the distance between the n samples and the nearest clustering center) is calculated and is represented by D (x) (0 < n < K).
Step S523: using formulas
Calculate each sample xiProbability P of being selected as n +1 th cluster centeri。
Step S524: p of the sampleiAccumulating and summing in sequence to obtain multi-interval probability PiAnd' then, selecting the n +1 th clustering center by using a roulette method.
Step S525: and repeating the steps S522 to S524 until the selection of the K cluster centers is completed.
Step S526: and calculating the Euclidean distance between each sample and the K clustering centers, and if the distance between the sample and the nth clustering center is smaller than that of other centers, distributing the sample to the class to which the clustering center belongs.
Step S527: according to the formula
A weighted average is calculated for each class sample.
Step S528: and repeating the steps S526 and S527, stopping iteration if the clustering center is not changed any more or the change amplitude is smaller than the threshold value, and otherwise, continuing to iterate to obtain the optimal clustering result.
The optimal number of clusters is determined using the Sum of Squared Error (SSE) as a metric. The mathematical expression of SSE is shown in equation (23):
in the formula: a. theiTable ith cluster, μiThe average value of all samples in the ith cluster is obtained, the classification of each cluster is finer along with the increase of the center number K, the clustering effect is better, the SSE value is reduced along with the increase of the center number K, and finally the SSE value gradually approaches to a stable value.
When K is less than the optimal cluster number, the aggregation degree of the clusters is greatly influenced when a cluster center is added, so that SSE is greatly reduced; however, when K is greater than the optimal clustering number, the aggregation degree cannot be greatly improved by adding the clustering center, the SEE drop amplitude is reduced, the curve of the relationship with K tends to be flat and takes the shape of an elbow, so the method is called an elbow method, and the K value corresponding to the elbow is the optimal clustering number of the data. Referring to fig. 5, a diagram of variation of the sum of squares of the error and the cluster number provided in the embodiment of the present application is shown. As can be seen from the figure, when K is 3, the SSE amplitude decreases rapidly, and the optimal cluster number is 3, where the cluster centers are 3.28053427, 1.07720264 and 2.27266057, and the percentage of clusters with the center of 1.07720264 is the largest. In summary, the optimized default frame width-to-height ratio is set as
Step S53: an improved YOLOv3 convolutional neural network was constructed.
The YOLOv3 convolutional neural network replaces the softmax algorithm with an independent logical classifier, predicts bounding boxes using a spatial pyramid structure, and designs a darknet-53 neural network structure. The system applies the YOLOv3 convolutional neural network to transmission conductor defect classification, thereby helping to improve efficiency and reduce cost.
1) Introducing DropBlock layer
Since the number of pictures in the power transmission conductor defect database is far less than that of the parameters in the model relative to the main network (Darknet-53), the trained model is easy to generate an overfitting phenomenon. To effectively solve this problem, a DropBlock layer with a structure is added after the convolution layer in the YOLOv3 convolutional neural network of the embodiment of the present application. The principle of the DropBlock layer is that other feature information must be searched by a network to fit data by abandoning adjacent areas in a feature map, so that more spatial distributions can be learned, and the robustness of a model is improved.
In this embodiment of the application, the DropBlock layer includes two parameters, which are a size Block _ size of a Block to be discarded and a number γ of active units to be discarded, respectively, and a calculation formula of γ is as follows:
in the formula (24), keep _ prob is the probability that the activation unit in the conventional Dropout layer is reserved, and is 0.8 in the embodiment of the present application. The flat _ size is the size of the feature map.
In the embodiment of the application, a DropBlock layer is added behind a residual unit in a YOLOv3 convolutional neural network.
2) Improving the loss function
The loss function of YOLOv3 mainly includes coordinate error, confidence error, and classification error. Wherein the confidence error comprises a confidence error of a bounding box containing the object and a confidence error of a bounding box containing no object. In the YOLOv3 convolutional neural network, sum-squared error is used to calculate these losses and regress the predictions. For the coordinate error, the calculation formula is as follows:
in the formula (25), the number S of network cell partitions in each layer of prediction is 13, 26, 52; b is the predicted number of the boundary frames of each network cell, which is 3 in the embodiment of the application;
judging whether the jth boundary box in the ith network cell is responsible for predicting the object, wherein in the three predicted boundary boxes, only the boundary box with the maximum intersection ratio with the object marking box is responsible for predicting the object; lambda [ alpha ]coordWeights are lost for the coordinates to increase the weight of the coordinate error.
However, the L2 Loss method is sensitive to the scale change of the target when performing border regression, and the method uses four values in coordinates
Treated separately, but in practice these four values are highly correlated and together they form a bounding box, thus making the positioning not very accurate. The intersection and union ratio is the ratio of the intersection and union of the prediction frame and the marking frame, the area of the prediction frame is A, the area of the marking frame is B, and the area of the intersection region is C, then the intersection and union ratio of the prediction frame and the marking frame is shown as formula (26):
when IoU is 1, the prediction box fits the label box completely, and when IoU is 0, the prediction box does not fit the label box at all. However, when IoU is used as the loss function, if the prediction box does not intersect the annotation box, i.e. IoU is 0, this does not reflect the distance between the two bounding boxes, and there is no gradient back-pass when loss is 0.
To address this problem, the giou (generalized interaction over union) is introduced in the embodiment of the present application to evaluate the distance between the prediction box and the annotation box. Firstly, calculating the minimum closure area D of the prediction frame and the labeling frame, then calculating the intersection ratio IoU of the prediction frame and the labeling frame and the union area U of the prediction frame and the labeling frame, wherein the calculation formula of the GIoU is as follows:
the GIoU pays attention to not only the overlapping region of the two frames but also the non-overlapping region, when the prediction frame is completely attached to the labeling frame, the GIoU is 1, and when the prediction frame is infinitely far away from the labeling frame, the GIoU is-1, so that the GIoU can well reflect the distance between the two bounding frames, and is insensitive to target scale change. In order to meet the requirement of the loss function, the larger the distance is, the larger the error is, and the coordinate loss function in the embodiment of the present application is LGIoU1-GIoU, while the loss function of classification error remains unchanged. By improving the coordinate Loss function, the convergence process of Loss can be optimized, and the positioning accuracy is improved.
Step S54: and (5) training.
The embodiment of the application uses a keras framework based on a YOLOv3 convolutional neural network, and takes 90% of the lead defect images in the database as a training set and 10% of the images as a test set. In the testing process, the image which is not marked is directly transmitted through the forward propagation of the neural network, and the detection result of the position image can be obtained through non-maximum inhibition. In the training process, the training speed of the network is improved by adopting an Adam optimization algorithm, the number of training iterations is 150 epochs (all samples are trained for 150 times), the batch sample training number (batch _ size) of the first 100 epochs is 16 (namely 16 pictures are trained and optimized at the same time in the training process), and the learning rate is 1e-3The batch _ size of the last 50 epochs is 8, and learning is performedThe ratio is 1e-4. After training is finished, the system can automatically save the parameters of the neural network detected by the electric power facility.
Step S55: and (6) testing.
And performing non-maximum inhibition on all the predicted frames, selecting the predicted frames as final prediction results according to the confidence coefficient, and performing model performance test by using 1000 pictures in the test set. The improved YOLOv3 target detection model can accurately identify and locate single-strand wire strands, multi-strand wire strands and wire breakage positions in the image of the power transmission line, and outputs coordinate information (namely the coordinates of the upper left vertex and the coordinates of the lower right vertex of the bounding box) and category information of the targets in the image.
Compared with the prior art, the automatic identification precision of the transmission conductor defects can be obviously improved, and the automatic identification speed of the system which is trained by using the improved YOLOv3 convolutional neural network is obviously improved compared with that of other neural networks.
In summary, the embodiment of the application obtains the position information of the transmission conductor defect in the image through the image processing technology and the neural network model, and simultaneously, the identification precision and the position information of the transmission conductor defect are correspondingly stored, so that the identification process of the transmission conductor defect is automatically realized, the workload of manual identification is reduced, the identification precision and efficiency are improved, and reliable data support is provided for the state modeling and scientific operation and maintenance of the transmission line.
A second aspect of the present embodiment provides a system for visually recognizing and positioning a defect of a power transmission line, which is used to execute the method for visually recognizing and positioning a defect of a power transmission line provided in the first aspect of the present embodiment.
The visual identification and positioning system for the defects of the power transmission conductors specifically comprises an image preprocessing unit, a conductor extracting unit and a classifying and positioning unit.
The image preprocessing unit is used for preprocessing a high-definition image and comprises a gray processing module, an image denoising and enhancing module and an edge detection module; the gray processing module performs gray processing on the high-definition image by using the improved local binary pattern characteristic value to obtain a gray image; the image denoising and enhancing module is used for carrying out filtering processing on the grayed image through a wiener filter and further enhancing the grayed image after the filtering processing by utilizing histogram equalization to obtain an enhanced image; and the edge detection module is used for carrying out edge detection on the enhanced image through a Canny edge detection algorithm to obtain an edge image.
The lead extraction unit is used for extracting a part of the power transmission lead as an interested area and comprises a straight line detection module, a fitting central line module and an interested area extraction module; the line detection module is used for extracting the texture of the power transmission conductor from the edge image through a Hough line detection algorithm; the fitting central line module is used for fitting the texture central line of the power transmission conductor by a least square method; and the interesting region extracting module is used for acquiring a transmission conductor fitting image according to the transmission conductor texture center line and taking the transmission conductor fitting image as an interesting region.
The classification and positioning unit is used for automatically identifying the region of interest through a pre-established defect identification model, obtaining the defect type of the region of interest and framing the position information of the region of interest to which the defect type belongs; the defect identification model comprises a transmission conductor defect type database, and can match the region of interest with the transmission conductor defect type database to determine whether the region of interest has defects and the defect types of the defects; the defect recognition model is obtained by training a modified YOLOv3 convolutional neural network, which comprises introducing a DropBlock layer and modifying a loss function.
According to the technical scheme, the visual identification and positioning method for the defects of the power transmission conductors based on the convolutional neural network is characterized in that the types of the defects of the power transmission conductors are obtained through the combined application of an image processing technology and a neural network model, the position information of the defects in the images is framed, and the identification precision and the position information of the defects of the power transmission conductors are correspondingly stored, so that the automatic identification and positioning of the defects of the power transmission conductors can be realized, the workload of visual identification is favorably reduced, the labor cost is reduced, the automatic identification accuracy and efficiency are improved, and the stable operation of a power grid is further ensured.
The present application has been described in detail with reference to specific embodiments and illustrative examples, but the description is not intended to limit the application. Those skilled in the art will appreciate that various equivalent substitutions, modifications or improvements may be made to the presently disclosed embodiments and implementations thereof without departing from the spirit and scope of the present disclosure, and these fall within the scope of the present disclosure. The protection scope of this application is subject to the appended claims.
Claims (10)
1. A visual identification and positioning method for a defect of a power transmission conductor is characterized by comprising the following steps:
acquiring a high-definition image of a power transmission conductor;
carrying out graying processing on the high-definition image by utilizing the improved local binary pattern characteristic value to obtain a grayed image; the improved local binary pattern characteristic value is obtained by expanding the contrast range of the pixel point to a circular area with any radius and according to a Laplacian operator;
denoising and enhancing the grayed image to obtain an enhanced image, and performing edge detection on the enhanced image by using a Canny edge detection algorithm to obtain an edge image;
after the texture of the power transmission conductor is extracted from the edge image by using a Hough linear detection algorithm, fitting the texture center line of the power transmission conductor by using a least square method, obtaining a fitted image of the power transmission conductor by using the texture center line of the power transmission conductor, and taking the fitted image of the power transmission conductor as an interested area;
automatically identifying the region of interest through a pre-established defect identification model to obtain the defect type of the region of interest, and framing the position information of the region of interest to which the defect type belongs;
the defect identification model comprises a transmission conductor defect type database, and can match the region of interest with the transmission conductor defect type database to determine whether the region of interest has defects and the defect types of the defects; the defect recognition model is obtained by training a modified YOLOv3 convolutional neural network, which comprises introducing a DropBlock layer and modifying a loss function.
2. The visual identification and location method for the defects of the power transmission conductor according to claim 1, wherein the graying processing is performed on the high-definition image by using the improved local binary pattern characteristic value, and the specific method comprises the following steps:
step S21: acquiring all pixel points of the high-definition image;
step S22: selecting any pixel point as a central pixel point, determining a circular area of the central pixel point according to a preset radius, and selecting a set number of pixel points in the circular area as neighborhood pixel points;
step S23: obtaining a Laplace operator of the central pixel point according to the gray value of the central pixel point; obtaining a Laplacian operator of the neighborhood pixel points according to the gray value of the neighborhood pixel points;
step S24: calculating a local binary pattern characteristic value of the central pixel point according to the Laplacian of the central pixel point and the Laplacian of the neighborhood pixel points;
step S25: and carrying out gray level normalization on the high-definition image by using the local binary pattern characteristic values of all the pixel points.
3. The visual identification and location method for the defects of the power transmission conductor according to claim 1, wherein the specific method for establishing the defect identification model is as follows:
step S31: acquiring a transmission conductor defect image in a transmission conductor defect type database;
step S32: dividing the transmission conductor defect image into a training set and a test set;
step S33: repeatedly training the training set through a back propagation algorithm to establish a preliminary neural network model;
step S34: and verifying the precision of the preliminary neural network model by using the test set, and if the precision of the preliminary neural network model meets the requirement, generating a defect identification model.
4. The method according to claim 1, wherein the specific method for improving the loss function is:
step S41: calculating the intersection area and the union area of the prediction frame and the marking frame;
step S42: obtaining the intersection ratio of the prediction frame and the marking frame according to the intersection area and the union area, wherein the intersection ratio is the ratio of the intersection area of the prediction frame and the marking frame to the union area;
step S43: calculating the minimum closure area of the prediction frame and the labeling frame, wherein the minimum closure area is the area of the minimum frame which simultaneously comprises the prediction frame and the labeling frame;
step S44: calculating the proportion of the region which does not belong to the prediction frame or the labeling frame in the minimum closure region to the minimum closure region according to the area of the minimum closure region and the area of the union to obtain the non-intersection proportion;
step S45: obtaining a generalized intersection ratio according to the intersection ratio and the non-:
in the formula, GIoU is a generalized intersection-to-parallel ratio, IoU is an intersection-to-parallel ratio of a prediction box and a labeling box, D is the area of a minimum closure area, and U is the union area of the prediction box and the labeling box;
step S46: obtaining a loss function according to the generalized intersection ratio, wherein the loss function is as follows: l isGIoU＝1-GIoU。
5. A method for visual identification and location of power transmission conductor defects according to claim 1, characterized in that said DropBlock layer comprises two parameters, respectively the size Block _ size of the Block to be discarded and the number γ, γ of active cells to be discarded, the formula for which is as follows:
where keep _ prob is the probability that the active cell in the conventional Dropout layer is reserved, and flat _ size is the size of the feature map.
6. The method as claimed in claim 1, wherein said de-noising of said grayed images is performed by wiener filtering.
7. The method according to claim 1, wherein the enhancement of the grayed image is performed by histogram equalization.
8. A method according to claim 1, wherein the Canny edge detection uses the maximum inter-class difference method.
9. A method for visual identification and location of power transmission conductor defects according to claim 1, wherein said database of power transmission conductor defect types comprises: single strand wire breakage, multiple strand wire breakage, and wire breakage.
10. A visual identification and localization system of a power transmission conductor defect, characterized in that it is adapted to perform a method of visual identification and localization of a power transmission conductor defect according to any one of claims 1-9, comprising:
the image preprocessing unit is used for preprocessing a high-definition image and comprises a gray processing module, an image denoising and enhancing module and an edge detection module; the gray processing module performs gray processing on the high-definition image by using the improved local binary pattern characteristic value to obtain a gray image; the image denoising and enhancing module is used for carrying out filtering processing on the grayed image through a wiener filter and further enhancing the grayed image after the filtering processing by utilizing histogram equalization to obtain an enhanced image; the edge detection module is used for carrying out edge detection on the enhanced image through a Canny edge detection algorithm to obtain an edge image;
the lead extraction unit is used for extracting a part of the power transmission lead as an interested area and comprises a straight line detection module, a fitting central line module and an interested area extraction module; the line detection module is used for extracting the texture of the power transmission conductor from the edge image through a Hough line detection algorithm; the fitting central line module is used for fitting the texture central line of the power transmission conductor by a least square method; the interesting region extracting module is used for acquiring a transmission conductor fitting image according to the transmission conductor texture center line and taking the transmission conductor fitting image as an interesting region;
the classification and positioning unit is used for automatically identifying the region of interest through a pre-established defect identification model, obtaining the defect type of the region of interest and framing the position information of the region of interest to which the defect type belongs; the defect identification model comprises a transmission conductor defect type database, and can match the region of interest with the transmission conductor defect type database to determine whether the region of interest has defects and the defect types of the defects; the defect recognition model is obtained by training a modified YOLOv3 convolutional neural network, which comprises introducing a DropBlock layer and modifying a loss function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110546688.3A CN113205063A (en) | 2021-05-19 | 2021-05-19 | Visual identification and positioning method for defects of power transmission conductor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110546688.3A CN113205063A (en) | 2021-05-19 | 2021-05-19 | Visual identification and positioning method for defects of power transmission conductor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113205063A true CN113205063A (en) | 2021-08-03 |
Family
ID=77031829
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110546688.3A Pending CN113205063A (en) | 2021-05-19 | 2021-05-19 | Visual identification and positioning method for defects of power transmission conductor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113205063A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113450372A (en) * | 2021-08-27 | 2021-09-28 | 海门裕隆光电科技有限公司 | Power transmission line image intelligent enhancement method and system based on artificial intelligence |
| CN113569981A (en) * | 2021-08-13 | 2021-10-29 | 桂林电子科技大学 | Power inspection bird nest detection method based on single-stage target detection network |
| CN113640308A (en) * | 2021-08-31 | 2021-11-12 | 郑州铁路职业技术学院 | Track abnormity monitoring system based on machine vision |
| CN113640307A (en) * | 2021-08-31 | 2021-11-12 | 郑州铁路职业技术学院 | Track condition monitoring method adopting machine vision |
| CN114022554A (en) * | 2021-11-03 | 2022-02-08 | 北华航天工业学院 | Massage robot acupuncture point detection and positioning method based on YOLO |
| CN114291531A (en) * | 2022-02-24 | 2022-04-08 | 深圳市扬帆威志实业有限公司 | Method for controlling speed of discharging material by using mechanical conveyor belt for marinated food |
| CN114882043A (en) * | 2022-07-12 | 2022-08-09 | 南通三信塑胶装备科技股份有限公司 | Injection molding defect positioning method and system based on image processing |
| CN116168034A (en) * | 2023-04-25 | 2023-05-26 | 深圳思谋信息科技有限公司 | Method, device, equipment and storage medium for detecting defect of knitted fabric |
| CN116664571A (en) * | 2023-07-31 | 2023-08-29 | 湖南湘江电缆有限公司 | Cable quality detection method and device based on machine vision |
| CN116778263A (en) * | 2023-08-22 | 2023-09-19 | 四川坤鸿电子科技有限公司 | Sorting apparatus control method, electronic apparatus, and computer-readable medium |
| CN117423001A (en) * | 2023-12-18 | 2024-01-19 | 东北大学 | Rock lithology recognition method and system based on improved SSD, storage medium and equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542243A (en) * | 2010-12-17 | 2012-07-04 | 北京理工大学 | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method |
| CN106525027A (en) * | 2016-11-02 | 2017-03-22 | 上海航天控制技术研究所 | Star sensor star point extracting method based on local binary pattern |
| CN109961460A (en) * | 2018-12-26 | 2019-07-02 | 国网浙江省电力有限公司 | A kind of multiple target method for inspecting based on improvement YOLOv3 model |
| US20200175352A1 (en) * | 2017-03-14 | 2020-06-04 | University Of Manitoba | Structure defect detection using machine learning algorithms |
| CN111999308A (en) * | 2020-08-27 | 2020-11-27 | 银河水滴科技（北京）有限公司 | Defect detection control method, control device and control system |
-
2021
- 2021-05-19 CN CN202110546688.3A patent/CN113205063A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542243A (en) * | 2010-12-17 | 2012-07-04 | 北京理工大学 | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method |
| CN106525027A (en) * | 2016-11-02 | 2017-03-22 | 上海航天控制技术研究所 | Star sensor star point extracting method based on local binary pattern |
| US20200175352A1 (en) * | 2017-03-14 | 2020-06-04 | University Of Manitoba | Structure defect detection using machine learning algorithms |
| CN109961460A (en) * | 2018-12-26 | 2019-07-02 | 国网浙江省电力有限公司 | A kind of multiple target method for inspecting based on improvement YOLOv3 model |
| CN111999308A (en) * | 2020-08-27 | 2020-11-27 | 银河水滴科技（北京）有限公司 | Defect detection control method, control device and control system |
Non-Patent Citations (5)
| Title |
|---|
| QINGPENG LI: "Hough transform guided deep feature extraction for dense building detection in remote sensing images", 《IEEE》, 13 September 2018 (2018-09-13), pages 1 - 6 * |
| 孙凤杰: "输电导线图像目标识别方法", 《中国图像图形学报》, no. 2012, 16 March 2012 (2012-03-16), pages 5 - 7 * |
| 张迪: "基于卷积神经网络的输电线路典型部件视觉检测方法研究", 《中国优秀硕士学位论文全文数据库》, no. 2020, 15 July 2020 (2020-07-15), pages 23 - 29 * |
| 王浩宇: "融合曲面形状和纹理特征的三维人脸识别研究", 《中国优秀硕士学位论文全文数据库》, no. 2019, 15 February 2019 (2019-02-15), pages 23 * |
| 黄东芳: "复杂背景下输电导线图像特征的提取", 《数字技术与应用》, no. 2015, 15 November 2015 (2015-11-15), pages 1 - 2 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113569981A (en) * | 2021-08-13 | 2021-10-29 | 桂林电子科技大学 | Power inspection bird nest detection method based on single-stage target detection network |
| CN113450372B (en) * | 2021-08-27 | 2021-11-16 | 海门裕隆光电科技有限公司 | Power transmission line image intelligent enhancement method and system based on artificial intelligence |
| CN113450372A (en) * | 2021-08-27 | 2021-09-28 | 海门裕隆光电科技有限公司 | Power transmission line image intelligent enhancement method and system based on artificial intelligence |
| CN113640307B (en) * | 2021-08-31 | 2023-10-10 | 郑州铁路职业技术学院 | Rail condition monitoring method adopting machine vision |
| CN113640308A (en) * | 2021-08-31 | 2021-11-12 | 郑州铁路职业技术学院 | Track abnormity monitoring system based on machine vision |
| CN113640307A (en) * | 2021-08-31 | 2021-11-12 | 郑州铁路职业技术学院 | Track condition monitoring method adopting machine vision |
| CN113640308B (en) * | 2021-08-31 | 2024-03-29 | 夏冰心 | Rail anomaly monitoring system based on machine vision |
| CN114022554A (en) * | 2021-11-03 | 2022-02-08 | 北华航天工业学院 | Massage robot acupuncture point detection and positioning method based on YOLO |
| CN114022554B (en) * | 2021-11-03 | 2023-02-03 | 北华航天工业学院 | Massage robot acupoint detection and positioning method based on YOLO |
| CN114291531A (en) * | 2022-02-24 | 2022-04-08 | 深圳市扬帆威志实业有限公司 | Method for controlling speed of discharging material by using mechanical conveyor belt for marinated food |
| CN114882043B (en) * | 2022-07-12 | 2022-09-06 | 南通三信塑胶装备科技股份有限公司 | Injection molding part defect positioning method and system based on image processing |
| CN114882043A (en) * | 2022-07-12 | 2022-08-09 | 南通三信塑胶装备科技股份有限公司 | Injection molding defect positioning method and system based on image processing |
| CN116168034B (en) * | 2023-04-25 | 2023-07-18 | 深圳思谋信息科技有限公司 | Method, device, equipment and storage medium for detecting defect of knitted fabric |
| CN116168034A (en) * | 2023-04-25 | 2023-05-26 | 深圳思谋信息科技有限公司 | Method, device, equipment and storage medium for detecting defect of knitted fabric |
| CN116664571A (en) * | 2023-07-31 | 2023-08-29 | 湖南湘江电缆有限公司 | Cable quality detection method and device based on machine vision |
| CN116664571B (en) * | 2023-07-31 | 2023-10-10 | 湖南湘江电缆有限公司 | Cable quality detection method and device based on machine vision |
| CN116778263A (en) * | 2023-08-22 | 2023-09-19 | 四川坤鸿电子科技有限公司 | Sorting apparatus control method, electronic apparatus, and computer-readable medium |
| CN116778263B (en) * | 2023-08-22 | 2023-11-14 | 四川坤鸿电子科技有限公司 | Sorting apparatus control method, electronic apparatus, and computer-readable medium |
| CN117423001A (en) * | 2023-12-18 | 2024-01-19 | 东北大学 | Rock lithology recognition method and system based on improved SSD, storage medium and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113205063A (en) | Visual identification and positioning method for defects of power transmission conductor | |
| CN108961235B (en) | Defective insulator identification method based on YOLOv3 network and particle filter algorithm | |
| CN111428748B (en) | HOG feature and SVM-based infrared image insulator identification detection method | |
| CN109118479B (en) | Capsule network-based insulator defect identification and positioning device and method | |
| CN112199993B (en) | Method for identifying transformer substation insulator infrared image detection model in any direction based on artificial intelligence | |
| CN110060237B (en) | Fault detection method, device, equipment and system | |
| CN110264448B (en) | Insulator fault detection method based on machine vision | |
| CN115082683A (en) | Injection molding defect detection method based on image processing | |
| CN112967243A (en) | Deep learning chip packaging crack defect detection method based on YOLO | |
| CN111814686A (en) | Vision-based power transmission line identification and foreign matter invasion online detection method | |
| CN113724231B (en) | Industrial defect detection method based on semantic segmentation and target detection fusion model | |
| CN109615609A (en) | A kind of solder joint flaw detection method based on deep learning | |
| CN110838112A (en) | Insulator defect detection method based on Hough transform and YOLOv3 network | |
| CN115330767B (en) | Method for identifying production abnormity of corrosion foil | |
| Ni et al. | An improved Faster R-CNN for defect recognition of key components of transmission line | |
| CN110766016B (en) | Code-spraying character recognition method based on probabilistic neural network | |
| CN109447036A (en) | A kind of segmentation of image digitization and recognition methods and system | |
| CN113538503A (en) | Solar panel defect detection method based on infrared image | |
| CN113240623A (en) | Pavement disease detection method and device | |
| CN110276747B (en) | Insulator fault detection and fault rating method based on image analysis | |
| CN116664565A (en) | Hidden crack detection method and system for photovoltaic solar cell | |
| CN113421223B (en) | Industrial product surface defect detection method based on deep learning and Gaussian mixture | |
| CN108734709B (en) | Insulator flange shape parameter identification and damage detection method | |
| CN111695373A (en) | Zebra crossing positioning method, system, medium and device | |
| CN116363136A (en) | On-line screening method and system for automatic production of motor vehicle parts |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210803 |
|
| RJ01 | Rejection of invention patent application after publication |