CN112418241A - Power equipment identification method based on infrared imaging technology - Google Patents
Power equipment identification method based on infrared imaging technology Download PDFInfo
- Publication number
- CN112418241A CN112418241A CN201910779829.9A CN201910779829A CN112418241A CN 112418241 A CN112418241 A CN 112418241A CN 201910779829 A CN201910779829 A CN 201910779829A CN 112418241 A CN112418241 A CN 112418241A
- Authority
- CN
- China
- Prior art keywords
- image
- power equipment
- edge
- infrared
- equipment identification
- 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
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000005516 engineering process Methods 0.000 title claims abstract description 21
- 238000003331 infrared imaging Methods 0.000 title claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 35
- 238000001514 detection method Methods 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 11
- 238000003708 edge detection Methods 0.000 claims abstract description 9
- 238000003707 image sharpening Methods 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 33
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 230000009466 transformation Effects 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 8
- 230000000877 morphologic effect Effects 0.000 claims description 8
- 238000005352 clarification Methods 0.000 claims description 4
- 238000012706 support-vector machine Methods 0.000 description 30
- 230000005540 biological transmission Effects 0.000 description 6
- 238000000354 decomposition reaction Methods 0.000 description 6
- 238000013507 mapping Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000013519 translation Methods 0.000 description 4
- 238000007689 inspection Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 201000004569 Blindness Diseases 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 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 1
- 108010074864 Factor XI Proteins 0.000 description 1
- 208000025274 Lightning injury Diseases 0.000 description 1
- 206010034719 Personality change Diseases 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 239000002510 pyrogen Substances 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Images
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
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V10/507—Summing image-intensity values; Histogram projection analysis
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
- G06V10/12—Details of acquisition arrangements; Constructional details thereof
- G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
- G06V10/143—Sensing or illuminating at different wavelengths
-
- 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
- G06V10/443—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 by matching or filtering
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C1/00—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people
- G07C1/20—Checking timed patrols, e.g. of watchman
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Water Supply & Treatment (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an electric power equipment identification method based on an infrared imaging technology. The method mainly comprises the steps of image sharpening, edge extraction, feature extraction and power equipment detection. According to the method, firstly, infrared images are subjected to sharpening processing to obtain high-quality infrared images, then, image edges are extracted through edge detection, characteristics of the edge images are extracted, SVM training is carried out on the characteristics of all positive and negative samples to obtain a training model, and finally, electric power equipment identification in the infrared images is achieved through the training model. The power equipment identification method based on the infrared imaging technology can automatically identify the power equipment, overcomes the problem that the power equipment identification in a visible light image is easily interfered by the background, provides a foundation for the thermal fault detection of the power equipment, and has high practical value.
Description
Technical Field
The invention relates to the field of power transmission and transformation line inspection, in particular to an electric power equipment identification method based on an infrared imaging technology.
Background
With the rapid development of social economy and the rapid increase of the demand of society on electric power, the routing inspection of power transmission and transformation lines becomes an important means for guaranteeing the safety of national electricity utilization. In the power industry, power transmission and transformation lines are important components of power systems, and are exposed to the natural environment for a long time, so that the power transmission and transformation lines not only bear the internal pressure of mechanical load and power load, but also are damaged by external factors such as dirt, lightning stroke, strong wind, bird damage and the like. These factors accelerate aging and fatigue of the components on the line, and if the potential hazards are not discovered and eliminated in time, various faults may develop and even serious accidents may occur, which threatens the safety and stability of the power system.
The infrared thermal fault detection technology of the charged equipment is an emerging technology. The method is a comprehensive technology which utilizes the pyrogenicity effect of charged equipment, adopts a special instrument to obtain infrared radiation information emitted from the surface of the equipment and further judges the condition and defect property of the equipment. The infrared detection technology has the advantages of no power failure, long distance, accuracy, high efficiency and the like, overcomes the blindness of regular scheduled maintenance, and has high safety and economic value. The existing technology can diagnose the fault position and cannot confirm the type of the fault equipment only by utilizing thermal fault detection, and the equipment cannot be maintained in a targeted manner according to the type of the power equipment.
The method utilizes an infrared imaging technology to combine Edge-Oriented Histogram (EOH) features and a Support Vector Machine (SVM) classifier to classify and train the types of the electric power equipment, so as to realize the identification of the types of the electric power equipment. The image edge characteristics can reflect the main characteristics of the target in the image to a certain extent, and the Canny edge detection algorithm as a classical edge detection algorithm can effectively detect a large amount of edge information in the image, but cannot highlight the edge of the obvious target. And the morphological-based image sharpening algorithm can effectively enhance edges and smooth noise, so that the image significance can be improved to a certain extent. Therefore, the method adopts the morphology-based image sharpening algorithm and the Canny edge detection algorithm to extract the significant edge characteristics, obtains the training model by combining the SVM, and extracts the power equipment from the image by using the detection model. The algorithm can rapidly screen the power equipment from the complex background, reduces the influence of the complex background on the extraction of the power equipment, and has wider application range.
Disclosure of Invention
The invention aims to provide a power equipment identification method based on an infrared imaging technology, aiming at solving the problem that the existing unmanned aerial vehicle power transmission and transformation line inspection cannot accurately identify the type of a thermal fault power equipment.
In order to achieve the purpose, the invention adopts the technical scheme that:
the method comprises the steps of detecting and identifying the power equipment according to the processes of image sharpening, edge extraction, feature extraction and target identification, wherein the image sharpening is used for obtaining a high-quality infrared image, the edge extraction is used for extracting the edge of the image through edge detection, the feature extraction is used for extracting the feature of the edge image, the target identification is used for carrying out SVM training on the features of all positive and negative samples to obtain a training model, and then the training model is used for realizing the power equipment identification in the infrared image.
Preferably, the image sharpening further comprises:
a. normalizing the image, namely normalizing the gray value of the infrared image of the power equipment to [0,1], and eliminating the gray value deviation of the image;
b. wavelet denoising, namely eliminating the noise of the image with normalized gray value by utilizing wavelet transformation to improve the target characteristics;
c. and c, morphological filtering, if the noise of the image is still larger than the maximum allowable noise after the processing of the steps a and b, performing background filtering by using image morphological processing to further eliminate a sharpening noise point and an independent noise point.
Preferably, the edge extraction further comprises: and acquiring the salient edge contour of the image by using a Canny operator.
Further, in the region segmentation process, after image preprocessing is completed, a dog (difference of gaussian) algorithm and a Canny edge detection algorithm are used to separate candidate regions from background and other noise.
Preferably, the feature extraction further comprises: five different detection windows are utilized to slide in the edge image, and the edge direction histogram feature of the specified detection window is extracted as the feature of the detection window.
Further, in the feature extraction process, the Sobel operator is used for extracting edge features, and the features are improved, such as symmetric information, so that the edge information of the candidate area can be better captured.
Preferably, in the sliding process of the window, five types of windows with the same area and different lengths and widths are used for sliding. This is to reduce the false detection caused by different directions of the power equipment.
Preferably, the target recognition further comprises: putting the significant edge EOH characteristics of all positive and negative samples into an SVM classifier for training to obtain an SVM training model; and classifying the EOH characteristics of the infrared image detection window by using an SVM training model to realize the identification of the power equipment in the infrared image.
Preferably, the image EOH characteristics are used for an SVM classifier, a linear inner product is selected as a kernel function of the SVM classifier, and the training speed is improved by utilizing a self-adaptive modified learning rate and a momentum batch gradient descent algorithm during training to obtain the structure of the designed SVM classifier.
Compared with the prior art, the invention has the beneficial effects that: according to the method, firstly, infrared images are subjected to sharpening processing to obtain high-quality infrared images, then, image edges are extracted through edge detection, characteristics of the edge images are extracted, SVM training is carried out on the characteristics of all positive and negative samples to obtain a training model, and finally, electric power equipment identification in the infrared images is achieved through the training model. The power equipment identification method based on the infrared imaging technology can automatically identify the power equipment, overcomes the problem that the power equipment identification in a visible light image is easily interfered by the background, provides a foundation for the thermal fault detection of the power equipment, and has high practical value.
Drawings
FIG. 1 is a schematic flow diagram of the method of the present invention according to an embodiment.
FIG. 2 is a schematic view of a sliding window of the present invention, according to an embodiment.
FIG. 3 is a two-dimensional profile of the present invention according to an embodiment.
Fig. 4 is a diagram of an SVM mapping architecture of the present invention according to an embodiment.
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the accompanying drawings, and it is obvious that the described embodiments are only some embodiments of the present invention, 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 invention.
The invention provides an electric power equipment identification method based on an infrared imaging technology, which is used for extracting electric power equipment of an electric power tower by adopting infrared light electric power equipment detection on the basis of an electric power tower image.
Fig. 1 is a flow chart of the system of the present invention, which first performs sharpening and feature extraction on an infrared image to be recognized. The clarification process comprises the following steps: normalizing the image, namely normalizing the gray value of the infrared image of the power equipment to [0,1], and eliminating the gray value deviation of the image; wavelet denoising, namely eliminating the noise of an image by utilizing wavelet transformation and improving the target characteristics; and (4) morphological filtering, wherein if the image noise does not meet the requirement, background filtering is carried out by using image morphological processing, and a sharpening noise point and an independent noise point are further eliminated to obtain an image contour. After the image is clarified, edge extraction is carried out, and the significant edge contour of the image is obtained by using a Canny operator. And finally, sliding the sliding window in the edge image, extracting EOH characteristics of the image area in the window, and inputting the EOH characteristics of all positive and negative samples into an SVM classifier for classification to obtain a classification model. And the trained classifier is used for identifying the power equipment in the infrared image.
The clarification process is described in detail below:
1. image sharpening:
(1) the image is subjected to intensity normalization processing, a transformation function is determined by using moments which have invariance to affine transformation in the image, and then the original image is transformed into a gray-scale image by using the transformation function.
Assuming that the original image is an 8-bit grayscale image, the maximum value of the read pixel matrix is 256, the minimum value is 1, and the original image matrix is defined as I, then:
J=I/256 (1)
where J is the normalized image matrix, i.e. all pixel values after normalization are in the [0,1] interval.
(2) The wavelet transform is a mathematical tool widely used in recent years, and has been paid much attention in image processing. In contrast to fourier transforms, windowed fourier transforms, wavelet transforms are local transforms in space (time) and frequency, and thus can efficiently extract information from a signal. The multi-scale detailed analysis is carried out on the function or the signal through the operation functions of stretching, translation and the like, and a plurality of difficult problems which cannot be solved by Fourier transform are solved. A related definition of the wavelet transform is given below.
Giving a basic wavelet function psi (t) E L2(R),L2(R) represents the next one-bit square integrable function space of the Hilbert metric space if the condition is satisfied(i.e. the) The psi (t) function is called the basic wavelet. The wavelet ψ (t) is scaled and translated to obtain:
weighing Ψa,b(t) is a wavelet function dependent on the parameters a, b. Wherein a is a scale factor, b is a translation factor, and a, b and t are continuous variables. Then the square integrable function x (t) e L2The Wavelet Transform (WT) of (R) is defined as
Wherein psi*(t) complex conjugate expressed as ψ (t), < x (t), ψa,b(t)>Denotes x (t) and ψa,b(t) inner product.
The corresponding inverse transform:
wherein a, b and t are continuous variables, a is a scale factor, b is a translation factor, cψIs a tolerable condition for wavelets and is therefore called Continuous Wavelet Transform (CWT).
The continuous wavelet transform contains much redundant information, which is not beneficial to the analysis and processing of signals. In practical applications, particularly in the implementation of computer signals and image processing methods, only discrete samples are considered for determining an effective algorithm. The above wavelet transform can be discretized by sampling its scaling factor a and translation factor b, which can be taken as: j, k are integers, a0Is a constant greater than 1, b0A and b are chosen to be constants greater than 0 depending on the particular form of the wavelet. The discrete wavelet function is expressed as
The corresponding discrete wavelet transform may be represented as
When a is0=2,b0When 1, the discrete wavelet transform is called a dyadic discrete wavelet transform.
According to the fast algorithm of discrete dyadic wavelet transform proposed by Mallat, the wavelet decomposition process can use orthogonal mirror image digital filters H (low pass) and G (high pass) to realize layer-by-layer decomposition of the original image. Assuming that the impulse responses of H and G are H (n) and G (n), respectively, they are related to:
g(n)=(-1)nh(1-n) (7)
the output of the low pass filter is called the approximation and the output of the high pass filter is called the detail. The digital filter coefficients n are related to the wavelet shape employed. The approximation image obtained by decomposition can be further decomposed into new approximation and detail signals, namely, a higher-level decomposition is formed, the decomposition process is reversible, and the original signals can be reconstructed by the approximation and detail signals of any one-level decomposition.
(3) Morphological filtering: and if the noise is still larger than the maximum allowable noise, performing expansion, erosion and opening and closing operations on the image to eliminate sharp noise points and independent noise points.
2. Edge extraction:
and extracting a Gaussian difference saliency map of the image. The DoG (difference of gaussian) means that feature detection on a certain scale can obtain a response value image of DoG by subtracting images of two adjacent gaussian scale spaces, so that the effects of exciting a local central region and inhibiting surrounding neighborhoods are achieved, the image saliency can be reflected to a certain extent, and an initial gaussian difference saliency map of the image is calculated according to the following formula:
wherein x and y are respectively the abscissa and ordinate of the image pixel point, sigma1And σ2Representing excitation and suppression bandwidth, respectively, the value σ in the present invention1=0.7,σ20.9, I is a grayscale image, symbolRepresenting the sliding frequency filtering of the image, DoG (x, y) is the derived saliency measure.
Wherein, count (DoG >0) represents the number of significant points with significant value greater than 0 in DoG, and sum (DoG >0) represents the sum of significant values greater than 0.
Finally, a saliency map D (x, y) is calculated:and separates the object from the background and other noise.
3. Feature extraction:
and sliding the sliding window in the target candidate window, and extracting the EOH characteristics of the image area in the window.
The invention designs five windows with the same area and different sizes so as to reduce the missing detection caused by the condition that the size, the shape and the direction of the power equipment are different, as shown in figure 2. FIG. 2(a) (b) (c) (d) (e) shows five windows designed, after determining the window area S
S=x1y1=x2y2=x3y3=x4y4=x5y5 (9)
Fig. 2(f) shows the overlapping effect of five windows, which is the length and width of the corresponding window, respectively, wherein the window in fig. 2(c) is square.
The EOH feature extraction process is as follows:
the edge strength of a point (x, y) can be defined as
Gx(x,y)2,Gy(x,y)2Respectively, the gradient in the x, y directions.
To remove the effect of noise, a threshold value may be set
The value of T in practical application is usually set to 80-100.
The edge direction of the point (x, y) can be defined as
Then theta (0)0~1800) Equally divided into K intervals (K value is 4 to 8)
Now define the edge strength of a certain direction k in the sub-window W asFrom the above definitions, 3 types of EOH features are available (ξ is a random variable):
the first method comprises the following steps: comparison of edge intensities in the horizontal and vertical directions of the image,
and the second method comprises the following steps: the symmetry of the image symmetry feature is
And the third is that: the intensity in the horizontal direction is dominant, i.e.
4. Target identification:
and (4) putting the significant edge EOH characteristics of all the positive and negative samples into an SVM classifier for training to obtain an SVM detection model. And classifying the EOH characteristics of the infrared image detection window by using an SVM detection model to realize the identification of the power equipment in the infrared image.
SVMs are used to solve the following problems: find a hyperplane (hyperplane) that separates two different sets, as shown by the two-dimensional features in FIG. 3, for a given training set { (x)i,yi)},i=1,2,…,n,x∈Rn,yiE { +1, -1}, there is a straight line g (x) ═ wTx+b(wTIs slope and T ═ 1) let all yiPoint of ═ 1 falls in g(x)Side of < 0, all yiPoint +1 falls on g(x)>One side of 0. And the maximum distance from the two side boundaries (margin) becomes the optimal hyperplane. The standard support vector classifier at this time is:
yi[w·x+b]-1≥0i=1,2,…N (17)
the distance from the support hyperplane to the optimal hyperplane is as follows: d 1/| w |, the distance margin between two classification hyperplanes is twice d, i.e.: margin 2d 2/| | w | | | then the problem of solving the optimal hyperplane becomes:
s.t.yi[w·xi+b]-1≥0,i=1,2,…,N
however, when the sample data is more complex and changeable, the optimal hyperplane is difficult to divide all samples accurately, in order to deal with the wrong division problem of the SVM algorithm, a learner introduces the concept of the soft edge optimal hyperplane and introduces a relaxation factor ξ, which aims to relax the restriction conditions and allow a certain wrong division condition, so that the classifier model in the formula (18) is changed into:
yi[w·xi+b]-1+ξi≥0i=1,2,…,N (18)
at this time, solving the quadratic programming problem of the optimal hyperplane becomes:
s.t.yi[w·xi+b]-1+ξi≥0,i=1,2,…,N
the samples appearing in practical engineering are often non-linear, and the input space x needs to be mapped to a high-dimensional feature space through some non-linear mappingIn this space, linear classification rules exist, and a linear optimal classification hyperplane can be constructed. This mapping is typically implemented in the way the kernel function is designed. Different kernel functions are adopted to obtain different forms of nonlinear support vector machines. The following four types of kernel functions are commonly used:
(1) linear kernel function: k (x, x ') (x · x')
(2) Polynomial kernel function: k (x, x ') +1 ((x · x'))q
(4) sigmoid function: k (x, x ') ═ tanh (ν (x · x') + c)
The classifier model constructed in the space after nonlinear mapping is as follows;
the solution of the objective function at this time becomes:
in summary, the basic idea of the support vector machine can be summarized as transforming the input space to a high-dimensional space by the non-linear change defined by the inner product function, and then finding the optimal classification plane in the new space. The SVM classification function is formally similar to a neural network, the output of which is a linear combination of intermediate level nodes, each of which corresponds to the inner product of an input sample and a support vector, and is therefore also referred to as a support vector network, and the SVM mapping structure is shown in FIG. 3.
When the SVM classifier is trained on the power equipment target sample base under the complex interference. The perfection of the sample base used for training and the selection of different kernel functions and penalty weights all affect the performance of the SVM classifier.
The reasonable selection of the samples can not only improve the learning speed of the classifier, but also enable the classifier to have good identification precision. The invention follows the following three principles when constructing a sample set for machine learning training:
(1) avoiding unbalanced sample sets; (2) the target power equipment sample is representative; (3) the background sample is rich in types.
The training sample library constructed in the invention comprises positive sample target power equipment and negative sample background interference, and the sample database basically covers various possible interferences. The positive sample comprises a large power equipment target and a small power equipment target, and the attitude change of each direction of the target: negative examples include various backgrounds and interferences such as terrestrial background, natural interference and artificial interference.
For the kernel function in the SVM. A linear kernel function is typically used because it has a smaller model selection complexity: the polynomial kernel function and the RBF kernel function need to adjust a parameter q and a parameter sigma respectively; in addition, the Sigmoid function has more parameters v and c, and the function may have an illegal situation under some parameters.
The training precision of the linear kernel function is highest, the iteration times are minimum, and the model calculation complexity is low, so that the linear kernel inner product is selected as the kernel function when the SVM classifier is designed.
And selecting the linear inner product as a kernel function of the SVM classifier, learning and training the whole sample database, and improving the training speed by utilizing a self-adaptive modification learning rate and momentum batch gradient descent algorithm during training to obtain the structure of the designed SVM classifier.
The invention completes the automatic identification of the power equipment, overcomes the blindness of the regular scheduled maintenance of the power transmission and transformation circuit, and has higher efficiency, safety and economic value.
According to the method, firstly, infrared images are subjected to sharpening processing to obtain high-quality infrared images, then, image edges are extracted through edge detection, characteristics of the edge images are extracted, SVM training is carried out on the characteristics of all positive and negative samples to obtain a training model, and finally, electric power equipment identification in the infrared images is achieved through the training model. The power equipment identification method based on the infrared imaging technology can automatically identify the power equipment, overcomes the problem that the power equipment identification in a visible light image is easily interfered by the background, provides a foundation for the thermal fault detection of the power equipment, and has high practical value.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (7)
1. The method is characterized in that the power equipment is detected and identified according to the processes of image clarification, edge extraction, feature extraction and target identification, wherein the image clarification is used for obtaining high-quality infrared images, the edge extraction is used for extracting image edges through edge detection, the feature extraction is used for extracting edge image features, the target identification is used for carrying out SVM training on the features of all positive and negative samples to obtain a training model, and then the training model is used for realizing the power equipment identification in the infrared images.
2. The power equipment identification method based on the infrared imaging technology as claimed in claim 1, wherein the image sharpening further comprises:
a. normalizing the image, namely normalizing the gray value of the infrared image of the power equipment to [0,1], and eliminating the gray value deviation of the image;
b. wavelet denoising, namely eliminating the noise of the image with normalized gray value by utilizing wavelet transformation to improve the target characteristics;
c. and c, morphological filtering, if the noise of the image is still larger than the maximum allowable noise after the processing of the steps a and b, performing background filtering by using image morphological processing to further eliminate a sharpening noise point and an independent noise point.
3. The power equipment identification method based on the infrared imaging technology as claimed in claim 1, wherein the edge extraction further comprises: and acquiring the salient edge contour of the image by using a Canny operator.
4. The power equipment identification method based on the infrared imaging technology as claimed in claim 1, wherein the feature extraction further comprises: five different detection windows are utilized to slide in the edge image, and the edge direction histogram feature of the specified detection window is extracted as the feature of the detection window.
5. The infrared imaging technology-based power equipment identification method as claimed in claim 4, wherein in the window sliding process, five windows with the same area and different lengths and widths are used for sliding.
6. The power equipment identification method based on the infrared imaging technology as claimed in claim 1, wherein the target identification further comprises: putting the significant edge EOH characteristics of all positive and negative samples into an SVM classifier for training to obtain an SVM training model; and classifying the EOH characteristics of the infrared image detection window by using an SVM training model to realize the identification of the power equipment in the infrared image.
7. The infrared imaging technology-based power equipment identification method as claimed in claim 6, wherein the image EOH features are used for an SVM classifier, a linear inner product is selected as a kernel function of the SVM classifier, and a training speed is increased by using a self-adaptive modified learning rate and a momentum batch gradient descent algorithm during training to obtain a designed SVM classifier structure.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910779829.9A CN112418241A (en) | 2019-08-22 | 2019-08-22 | Power equipment identification method based on infrared imaging technology |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910779829.9A CN112418241A (en) | 2019-08-22 | 2019-08-22 | Power equipment identification method based on infrared imaging technology |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112418241A true CN112418241A (en) | 2021-02-26 |
Family
ID=74780074
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910779829.9A Pending CN112418241A (en) | 2019-08-22 | 2019-08-22 | Power equipment identification method based on infrared imaging technology |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112418241A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115188091A (en) * | 2022-07-13 | 2022-10-14 | 国网江苏省电力有限公司泰州供电分公司 | Unmanned aerial vehicle grid inspection system and method integrating power transmission and transformation equipment |
CN116580290A (en) * | 2023-07-11 | 2023-08-11 | 成都庆龙航空科技有限公司 | Unmanned aerial vehicle identification method, unmanned aerial vehicle identification device and storage medium |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104700066A (en) * | 2013-12-11 | 2015-06-10 | 杭州海康威视数字技术股份有限公司 | Method and device for detecting whether driver wears safety belt or not |
-
2019
- 2019-08-22 CN CN201910779829.9A patent/CN112418241A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104700066A (en) * | 2013-12-11 | 2015-06-10 | 杭州海康威视数字技术股份有限公司 | Method and device for detecting whether driver wears safety belt or not |
Non-Patent Citations (2)
Title |
---|
张迪飞;张金锁;姚克明;成明伟;吴永国;: "基于SVM分类的红外舰船目标识别", 红外与激光工程, no. 01 * |
郭文诚;崔昊杨;马宏伟;秦伦明;: "基于Zernike矩特征的电力设备红外图像目标识别", 激光与红外, no. 04 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115188091A (en) * | 2022-07-13 | 2022-10-14 | 国网江苏省电力有限公司泰州供电分公司 | Unmanned aerial vehicle grid inspection system and method integrating power transmission and transformation equipment |
CN115188091B (en) * | 2022-07-13 | 2023-10-13 | 国网江苏省电力有限公司泰州供电分公司 | Unmanned aerial vehicle gridding inspection system and method integrating power transmission and transformation equipment |
CN116580290A (en) * | 2023-07-11 | 2023-08-11 | 成都庆龙航空科技有限公司 | Unmanned aerial vehicle identification method, unmanned aerial vehicle identification device and storage medium |
CN116580290B (en) * | 2023-07-11 | 2023-10-20 | 成都庆龙航空科技有限公司 | Unmanned aerial vehicle identification method, unmanned aerial vehicle identification device and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108154519A (en) | Dividing method, device and the storage medium of eye fundus image medium vessels | |
Tahseen et al. | Binarization Methods in Multimedia Systems when Recognizing License Plates of Cars | |
Dixit et al. | Image texture analysis-survey | |
Thanikkal et al. | Advanced plant leaf classification through image enhancement and canny edge detection | |
CN109523529A (en) | A kind of transmission line of electricity defect identification method based on SURF algorithm | |
Rabee et al. | License plate detection and recognition in complex scenes using mathematical morphology and support vector machines | |
Ye et al. | Overhead ground wire detection by fusion global and local features and supervised learning method for a cable inspection robot | |
CN105512622A (en) | Visible remote-sensing image sea-land segmentation method based on image segmentation and supervised learning | |
CN112418241A (en) | Power equipment identification method based on infrared imaging technology | |
CN116823725A (en) | Aeroengine blade surface defect detection method based on support vector machine | |
CN116934725A (en) | Method for detecting sealing performance of aluminum foil seal based on unsupervised learning | |
CN110348307B (en) | Path edge identification method and system for crane metal structure climbing robot | |
Abdelkader et al. | Evaluation of spalling in bridges using machine vision method | |
Wang et al. | Fingerprint image segmentation by energy of Gaussian-Hermite moments | |
Guo et al. | Fault diagnosis of power equipment based on infrared image analysis | |
WO2015168362A1 (en) | Method and apparatus for processing block to be processed of urine sediment image | |
CN112132104B (en) | ISAR ship target image domain enhancement identification method based on loop generation countermeasure network | |
Qiu et al. | An optimized license plate recognition system for complex situations | |
Bin et al. | Automatic inspection of surface crack in solar cell images | |
CN113093164A (en) | Translation-invariant and noise-robust radar image target identification method | |
Rani et al. | Object Detection in Natural Scene Images Using Thresholding Techniques | |
Uthaib et al. | Vehicle plate localization and extraction based on hough transform and bilinear operations | |
Chaiyakhan et al. | Traffic sign classification using support vector machine and image segmentation | |
CN111950320A (en) | Iris identification method based on relative total variation and probability collaborative representation | |
Mao et al. | Change detection in semantic level for SAR images |
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 |