CN106595551B - Ice covering thickness detection method in powerline ice-covering image based on deep learning - Google Patents
Ice covering thickness detection method in powerline ice-covering image based on deep learning Download PDFInfo
- Publication number
- CN106595551B CN106595551B CN201611145878.XA CN201611145878A CN106595551B CN 106595551 B CN106595551 B CN 106595551B CN 201611145878 A CN201611145878 A CN 201611145878A CN 106595551 B CN106595551 B CN 106595551B
- Authority
- CN
- China
- Prior art keywords
- image
- ice
- output
- layer
- icing
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B21/00—Measuring arrangements or details thereof, where the measuring technique is not covered by the other groups of this subclass, unspecified or not relevant
- G01B21/02—Measuring arrangements or details thereof, where the measuring technique is not covered by the other groups of this subclass, unspecified or not relevant for measuring length, width, or thickness
- G01B21/08—Measuring arrangements or details thereof, where the measuring technique is not covered by the other groups of this subclass, unspecified or not relevant for measuring length, width, or thickness for measuring thickness
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
Ice covering thickness detection method in the invention discloses a kind of powerline ice-covering image based on deep learning, belong to digital image understanding field, it aims to overcome that and existing answers pulling force monitoring medium sensitivity and the not high problem of reliability, improve the accuracy and the degree of automation of electric power line ice-covering thickness monitoring, ice covering thickness monitoring and transfinite alarm of the present invention for power system transmission line, comprising: (1) collect icing image；(2) pretreatment image and data set is established；(3) convolutional neural networks are established；(4) training and test model；(5) ice covering thickness information is extracted and five steps such as be transmitted back to control centre；Digital picture characteristic recognition method is introduced into the ice covering thickness detection of power transmission line and shaft tower by the present invention, thickness information is automatically extracted using the morphological feature of icing in image, deicing plan is formulated for operation maintenance personnel, to guarantee that safe and stable operation of power system provides a kind of new intuitive and intelligentized means.
Description
Technical field
The invention belongs to digital image understanding technical fields, and in particular to cover in a kind of image based on deep learning algorithm
Ice thickness condition detection method can be used for monitoring electric system power transmission network equipment icing and alarm of transfiniting.
Background technique
Power network safety operation is self-evident to the importance of the national economic development, deepen continuously with Power System Interconnection and
The gradually implementation of electricity market, the running environment of power grid is also more complicated, and the stability and reliability to power grid propose higher
Requirement.China has a vast territory, weather is various, with a varied topography, and power network throughout the country is often subject to various natural calamities
It destroys, China is most of often because extreme low temperature and icing lead to large area blackout.Icing disaster will lead to power grid hair
Raw mechanical breakdown and electric fault, as substation stops transport, shaft tower collapses, ice dodges tripping, line oscillation and substation equipment and damages
Etc. accidents.
Icing causes substantial equipment to lose and cause large area blackout in China's power grid, and ice-covering area occurs often
Weather conditions are severe, and transport and communication interrupts, and repairing difficulty is big, cause large area blackout, seriously affect power supply reliability.
The icing of transmission line of electricity and power equipment is objective reality, can not be inherently eliminated.It, must to reduce icing bring disaster
Electrical equipment icing in power grid must be protected, eliminate icing security risk in time.
Mainly by monitoring and inhibiting two methods, monitoring means is passed by installing on electrical equipment for icing protection at present
Sensor and video camera realize on-line monitoring ice coating state, or are patrolled by manual inspection mode critical circuits and find that failure is hidden
Suffer from；It is domestic often to be influenced the icing that area has been mounted with ice covering monitoring system monitoring power grid critical circuits and node by icing disaster
State, but the ice damage accident to take place frequently in recent years proves, and current several monitoring means are also unable to satisfy power system security, stabilization
Service requirement, for 2014, because icing cause tripping 597 times, trip-out rate be 0.103 time/hundred km years, be overlapped at
Power 46.4% increases by 376 times, amplification 170.1% compared with 2013 (221 times).Icing causes failure is non-to stop within 2014
320 times, the non-rate of stopping of failure was 0.055 time/hundred km years, and it is about 2013 (63 that icing in 2014, which causes the non-number that stops of failure,
It is secondary) 5 times.It analyzes current icing monitoring and is difficult to the reason of meeting operation of power networks requirement, may be summarized to be:
(1) sensor monitoring icing is affected by working environment, and harsh climate meteorological condition or the interference of electromagnetic field are equal
The measurement accuracy of sensor can be reduced；
(2) video camera monitoring judges current ice coating state by shooting and monitoring point shaft tower or transmission line icing image, but due to
Lack and is effectively treated and utilizes method to icing image, it can not be from the reliable icing information of acquisition；
(3) manual inspection mode or helicopter routing inspection are at high cost, low efficiency, it is difficult to be monitored to the whole network；
(4) system cannot find icing hidden danger in time, and ice-coating pre-warning is caused not issue in time, in addition the phase of anti-ice operation
To lag, icing can not be eliminated in time.
Summary of the invention
In order to solve the above-mentioned technical problems, the present invention provides a kind of powerline ice-covering figures based on deep learning algorithm
It, can be according to the icing image automatic identification ice covering thickness of input, it is ensured that the timely transmission line icing of power department as recognition methods
Situation.
The technical scheme adopted by the invention is that: ice covering thickness in a kind of powerline ice-covering image based on deep learning
Detection method, which comprises the following steps:
Step 1: obtaining icing image data and corresponding answer pulling force monitoring data；
Step 2 pre-processes original icing image, and it is consistent that the size of original icing image is processed into size
Image, the ice covering thickness for using tension sensor to measure is as image tag；
Step 3: establishing deep learning convolutional neural networks model, for amount of images and size, establish corresponding model
The unit number and activation primitive of every layer network is arranged in parameter；
Step 4: adjustment weight and training pattern carry out feature extraction and combination to image, judge and export ice covering thickness；
Step 5: analysis model training classification results extract the ice covering thickness information of icing image.
China's electric icing monitors real-time and validity Shortcomings, as long as the reason is that improving sensor monitoring effect master
It to be improved from hardware aspect；The present invention starts with from icing image data, has studied a kind of more quick, accurate measurements icing
The method of state, eliminates the hidden trouble in time, improves the reliability of power grid ice covering monitoring system, ensures that power system security is steady
Fixed operation.
Digital picture characteristic recognition method is introduced into the ice covering thickness detection of power transmission line and shaft tower by the present invention, utilizes figure
The morphological feature of icing automatically extracts thickness information as in, formulates deicing plan for operation maintenance personnel, to guarantee power system security
Stable operation provides a kind of new intuitive and intelligentized means.
Figure of description
Fig. 1 is the flow diagram of the embodiment of the present invention；
Fig. 2 is the original icing image of the embodiment of the present invention；
Fig. 3 is the iterative method segmented image result of the embodiment of the present invention；
Fig. 4 is the LoG operator edge detection result of the automatic threshold of the embodiment of the present invention；
Fig. 5 is the Prewitt operator edge detection result of the automatic threshold of the embodiment of the present invention；
Fig. 6 is the Sobel operator edge detection result of the automatic threshold of the embodiment of the present invention；
Fig. 7 is the convolutional neural networks feature extraction schematic diagram of the embodiment of the present invention；
Fig. 8 is the convolution process schematic diagram of the embodiment of the present invention；
Fig. 9 is the pond process schematic of the embodiment of the present invention；
Figure 10 is 1 accuracy of identification comparison of convolutional neural networks ergodic data of each structure of the embodiment of the present invention；
Figure 11 is 5 accuracy of identification comparisons of convolutional neural networks ergodic data of each structure of the embodiment of the present invention；
Figure 12 is 10 accuracy of identification comparisons of convolutional neural networks ergodic data of each structure of the embodiment of the present invention；
Figure 13 is 15 accuracy of identification comparisons of convolutional neural networks ergodic data of each structure of the embodiment of the present invention；
Figure 14 is 20 accuracy of identification comparisons of convolutional neural networks ergodic data of each structure of the embodiment of the present invention；
Figure 15 is the convolutional neural networks difference model identification error comparison of the embodiment of the present invention.
Specific embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, with reference to the accompanying drawings and embodiments to this hair
It is bright to be described in further detail, it should be understood that implementation example described herein is merely to illustrate and explain the present invention, not
For limiting the present invention.
Referring to Fig.1, ice covering thickness detects in a kind of powerline ice-covering image based on deep learning provided by the invention
Method, comprising the following steps:
Step 1: from the ice covering monitoring system of power grid department collect icing image data and it is corresponding answer pulling force monitor number
According to；
The icing images such as the icing image of collection, including power transmission line, shaft tower, fitting, electrical equipment such as transformer, collection
Icing image should be clear as far as possible；The pulling force data of answering collected includes shaft tower model, position and the equivalence for answering tension sensor to measure
Ice covering thickness.
Step 2: original image being pre-processed, picture size is adjusted to identical, icing image data set is established；
Original image is pre-processed, including image segmentation and edge extracting；
Image segmentation is the technology and process for dividing the image into the region of each tool characteristic and extracting interesting target；
Threshold segmentation is a kind of common method in image segmentation, and the pixel that all gray scales are greater than or equal to threshold values is determined
To belong to object, gray value indicates prospect with " 255 ", and otherwise these pixels are excluded other than object area, threshold segmentation
Including Two-peak method and iterative method；
Two-peak method divides the image into foreground and background two parts, and the intensity profile curve approximation of image is considered by two just
State distribution functionWithBe formed by stacking, the histogram of image will will appear the peak values of two separation, it is bimodal it
Between trough at be exactly image threshold values where；
Iterative method is the improvement to Two-peak method, selects an approximate threshold values T first, divides the image into part R1And R2, meter
Calculate region R1And R2Mean μ1And μ2, select new partition threshold T=(μ1+μ2)/2, steps be repeated alternatively until μ1And μ2No longer
Until variation, the present invention carries out image dividing processing to image using iterative method；
Marginal point in edge extracting detection image first, then edge point is connected into profile, so that cut zone is constituted,
Since edge is the line of demarcation for the target and background extracted, target and background could be separated by extracting edge, and gradient-norm is calculated
Son has the property of shift invariant and isotropism matter, is suitable for edge detection, and the direction of grey scale change, the i.e. side on boundary
To then can be by θg=arctan (fy/fx) determine, wherein fxAnd fyIt is the direction mould of x and y, θ respectivelygIt is the inspection of consecutive image edge
Survey the direction of maximum of gradients；The present invention indicates operator in the form of differential operator, is then realized with fast convolution function.
Step 3: establishing deep learning convolutional neural networks model, for picture number and size, it is corresponding suitable to establish
The unit number and activation primitive of every layer network is arranged in model parameter；
Deep learning convolutional neural networks are established, are a kind of deep learning model based on BP neural network transformation, power
Value, which shares network structure, can reduce the complexity of network model, reduce the quantity of weight, which is in the input of network
What is showed when multidimensional image becomes apparent, and allows image directly as the input of network, avoids multiple in tional identification algorithm
Miscellaneous feature extraction and data reconstruction processes；
It is deconvoluted the input of this layer in convolutional layer with convolution kernel.First by upper one layer of each output characteristic pattern position phase
The convolution kernel of same data and this layer carries out convolution, then by all results addeds of same position convolution, obtains this layer output feature
The output of figure corresponding position.In order to reduce number of parameters, model training difficulty is reduced, deep learning uses weight shared mechanism.
Same output characteristic pattern uses the same convolution kernel, and all corresponding filter of convolution kernel is corresponding each time, and one
Convolution kernel only extracts a kind of feature, guarantees that aliasing does not occur for feature extraction；
After obtaining feature by convolution, go to train classifier with all obtained features of extracting, the present invention utilizes
Softmax classifier, each feature and image convolution can obtain the convolution feature of n dimension, due to there is n feature, learn
It practises the classifier that one has more than the input of 20,000,000 features to be inconvenient, is easy to appear overfitting；
Step 4: adjustment weight and training pattern carry out feature extraction and extraction to image, judge and export ice covering thickness,
Compare the thickness information of image tag；
The convolutional neural networks of foundation are trained and test, and the model for obtaining meeting required precision is for detecting in picture
Ice covering thickness, training process includes the following steps:
(1) sensitivity and error correction；
For C classification problem, N number of training sample is shared, the mean square deviation of model may be expressed as:
WhereinIndicate the kth class desired output of n-th of sample,Indicate the kth class reality output of n-th of sample.
For n-th of sample, reality output and the mean square deviation of ideal output be may be expressed as:
It is assumed that L is output layer, l is hidden layer, and 1 is input layer；L layers of activation output are as follows: xl=f (ul), wherein ul=
Wlxl-1+bl, f () is activation primitive, WlIt is l layers of weight, blIt is l layers of biasing；Define sensitivity are as follows:
WhereinTherefore the sensitivity of l layers and output layer respectively indicates are as follows:
Thus obtaining error correction (η is learning rate) may be expressed as:
(2) propagated forward
1) convolutional layer
Assuming that l layers are convolutional layers, then the characteristic pattern of this layer output and characteristic pattern size respectively indicate are as follows:
Output.size=input.size-ker nel.size+1
Wherein,It is l-1 layers of i-th of output,It is l layers of j-th of the convolution kernel inputted for i-th,It is
L layers of j-th of biasing, f () is activation primitive,It is l layers of j-th of output.
There are two features for the feature extraction of convolutional neural networks:
I. by convolution, indicate that the pixel of the regional area of input feature vector figure is special with a pixel of output characteristic pattern
Sign, this is the feature extraction of convolutional neural networks, while also reducing data dimension；
II. weight is shared, and the same characteristic pattern uses the same convolution kernel, extracts a feature, it is possible to reduce parameter number
Amount reduces time complexity.
2) sub-sampling layer
Sub-sampling layer can be regarded as pond process and a reduction process, so that input feature vector figure does not exist overlappingly
It is indicated again on output characteristic pattern, i.e. the combination of feature；In addition pondization can also reduce data dimension, accelerate calculating speed.It is defeated
The expression of figure may be expressed as: out
Wherein down (x) is to carry out sampling operation to the pixel region of input picture n × n,It is controlling elements, sampling is tied
Fruit Numerical Control reduces noise jamming in colour element numberical range.
(3) backpropagation
1) convolutional layer
The sensitivity of convolutional layer is represented by shown in following formula, it is contemplated that is sub-sampling layer before and after convolutional layer, be can be rewritten as:
Wherein, β is weight,
Instead of ∑ δl+1。
The gradient of base and convolution kernel may be expressed as:
Wherein u, v are deconvoluted the corresponding part of a upper tomographic image with convolution kernel.Wherein,It is l j-th of convolution kernel of layer
Sensitivity,It is l-1 layers of factor of momentum；
2) sub-sampling layer
It has been saved in propagated forward:Therefore weight gradient are as follows:
Wherein,It is l layers of down-sampling handling function.
(4) feature combines
One important feature of deep learning network is automatic learning characteristic, and expression in the training process is to learn spy
When levying the combination of figure, weight is assigned for each feature of extraction, propagated forward and backpropagation more positive error is repeated simultaneously
Weight is adjusted, achievees the purpose that characteristic optimization combines.Feature weight αijIt indicates, indicates wherein the i-th of j-th of output characteristic pattern
A input feature vector figure weight or contribution, are usually indicated with following formula:
Wherein, cijIndicate i-th of input feature vector figure weight in j-th of output characteristic pattern, ∑kexp(ckj) indicate j-th it is defeated
Out all weights of characteristic pattern and.
Then j-th of feature output can be rewritten as:
Wherein, f () is activation primitive, αijIndicate the weight of output characteristic pattern,It is this layer input,It is convolution kernel,It is this layer biasing.
Above formula meets:0≤αij≤1；
For single output unit, ignore footmark j, due to meeting:
With
After obtaining base gradient, convolution kernel gradient, connection weight gradient and the feature weight gradient of hidden layer, error correction is then
For (η is learning rate) shown in table 1.
1 convolutional neural networks of table more positive value
Wherein, η is learning rate,It is the gradient of base,It is the gradient of convolution kernel,It is connection gradient, αiIt is special
Levy weight.
Step 5: analysis model training classification results extract the ice covering thickness information of icing image and pass back in regelation monitoring
The heart.
The step 1 of the present embodiment collects the original icing image data of power grid and answers tension sensor data, shown in attached drawing 2
Be powerline ice-covering image, icing image request is clear to icing photographing section, and no muddy object covers, and answers pulling force data packet
Model, position and the ice covering thickness for answering tension sensor to measure of shaft tower are included, to find the position of fault in time；
The step 2 of the present embodiment pre-processes icing image by image segmentation and edge extracting, by the size of original image
It is set as that size is consistent, the ice covering thickness for using tension sensor to measure is as image tag；Image segmentation is carried out to image
And edge processing improves feature extraction efficiency to remove noise, image segmentation and edge extracting are to exclude environment and unrelated
Interference of the object to icing feature extraction often has the environment such as trees, insect since ultra-high-tension power transmission line ambient enviroment is complex
Factor interference is needed to be rejected the irrelevant factor in icing image using image Segmentation Technology, avoids interfering feature extraction；It is attached
Fig. 3 is iterative method segmented image effect, and attached drawing 4-6 is the LoG operator edge detection of automatic threshold, Prewitt operator side respectively
Edge detection and Sobel operator edge detection result；
The step 3 of the present embodiment establishes convolutional neural networks, and convolutional neural networks need to learn using spatial relationship reduction
Number of parameters with to the training performance of, back-propagation algorithm, attached drawing 7 is the brief description to model structure before improving；
Convolutional neural networks include input layer, hidden layer and output layer, and the inside hidden layer of convolutional neural networks is convolution
Layer and pond stacking generation building, convolutional layer can extract data characteristics, also be feature extraction layer, substantially be convolution, and attached drawing 8 is volume
Product process schematic, sub-sampling layer is also Feature Mapping layer, by obtaining feature by activation primitive and reflecting to pixel weighted sum
It penetrates, attached drawing 9 is pond process schematic；
In convolutional neural networks, input of a part (local experiences area) of image as the lowermost layer of hierarchical structure,
Information is successively transferred to different layers again, and every layer is gone to obtain the most significant feature for observing data by a digital filter,
Therefore the notable feature to the observation data of translation, scaling and invariable rotary can be obtained,
Model parameter is defined in step 3 and initializes parameters, sets activation primitive as " sigmoid "；
The convolutional neural networks that the step 4 pair of the present embodiment is established are trained, according to the adjustable model of training effect
The relevant parameters such as structure, train epochs, batch processing size, activation primitive, every layer of neuron number, neuronal quantity；
Since model parameter amount is big, the present invention is trained using 15235 icing pictures, according to power grid to ice coating state
Monitoring mechanism is divided into no icing (0cm), slight icing (0~5cm), moderate icing (5 according to model judgment models monitoring thickness
~10cm), serious regelation (10~15cm), dangerous icing (15~20cm) and fault pre-alarming icing (20cm or more) six etc.
Judgement, alignment error be previously compared with image for grade, output.
The step 5 analysis model training classification results of the present embodiment, extract the ice covering thickness information of icing image and pass back
Icing monitoring center；
The present invention has built four models of convolutional neural networks, is (1) 6-4-12-2, (2) 12-4-24-2, (3) respectively
4-2-8-2-16-2-32-2, (4) 16-2-8-2-4-2-2-2.Wherein odd-level is convolutional layer, and even level is pond layer.Four kinds
Model is in ergodic data difference number, each model mean square deviation variation.The mean square deviation of model measures the training effect of model, works as mould
The mean square deviation of type is smaller, indicates that training effect is better, and parameter setting is more reasonable, obtained to export closer to desired result.
Step number is to represent adjustment weight number, in deep learning algorithm, is usually trained training sample in batches, after having trained a collection of sample
Weighed value adjusting is primary, to improve training speed and training effect；
The 4 kinds of model structures of expression of attached drawing 10: (1) 6-4-12-2, (2) 12-4-24-2, (3) 4-2-8-2-16-2-32-2,
(4) 16-2-8-2-4-2-2-2, ergodic data is primary respectively, the situation of change of the mean square deviation of training process, and abscissa indicates
Adjusting parameter number, ordinate indicate mean square deviation；Curve tendency can reflect the training effect of model, while according to training effect
Model parameter is adjusted, to reach optimal training effect.Similarly, attached drawing 11-14 respectively indicates four kinds of model structure ergodic datas 5
Secondary, 10 times, 15 times and at 20 times, the situation of change of mean square deviation usually traverse that number is more, and the classification performance of model is better.
Attached drawing 15 is to 4 kinds of model structures: (1) 6-4-12-2, (2) 12-4-24-2, (3) 4-2-8-2-16-2-32-2,
(4) comparative analysis of the classification results of 16-2-8-2-4-2-2-2, abscissa are ergodic data numbers, and ordinate is that classification misses
Difference.
It should be understood that the part that this specification does not elaborate belongs to the prior art.
It should be understood that the above-mentioned description for preferred embodiment is more detailed, can not therefore be considered to this
The limitation of invention patent protection range, those skilled in the art under the inspiration of the present invention, are not departing from power of the present invention
Benefit requires to make replacement or deformation under protected ambit, fall within the scope of protection of the present invention, this hair
It is bright range is claimed to be determined by the appended claims.
Claims (8)
1. ice covering thickness detection method in a kind of powerline ice-covering image based on deep learning, which is characterized in that including with
Lower step:
Step 1: obtaining icing image data and corresponding answer pulling force monitoring data；
Step 2 pre-processes original icing image, and the size of original icing image is processed into the consistent figure of size
Picture uses the ice covering thickness of tension sensor measurement as image tag；
Step 3: deep learning convolutional neural networks model is established, for amount of images and size, establishes corresponding model parameter,
The unit number and activation primitive of every layer network are set；
Step 4: adjustment weight and training pattern carry out feature extraction and combination to image, judge and export ice covering thickness；
The convolutional neural networks of described pair of foundation are trained and test, and training process includes following sub-step:
Step 4.1: meter sensitivity and error correction；
In step 4.1, for C classification problem, N number of training sample is shared,Indicate the kth class desired output of n-th of sample,
Indicate the kth class reality output of n-th of sample, then the mean square deviation table of the output layer error of model ideal output and reality output
It is shown as:
For n-th of sample, reality output and the mean square deviation of ideal output be may be expressed as:
It is assumed that L is output layer, l is hidden layer, and 1 is input layer；L layers of activation output are as follows: xl=f (ul), wherein ul=Wlxl-1
+bl, f () is activation primitive, WlIt is l layers of weight, blIt is l layers of biasing；Define sensitivity are as follows:
WhereinTherefore the sensitivity of l layers and output layer respectively indicates are as follows:
Thus error correction is obtained are as follows:
Wherein, η is learning rate；
Step 4.2: propagated forward；
Step 4.3: backpropagation；
Step 4.4: feature combination；Weight is assigned for each feature of extraction, propagated forward and backpropagation is repeated more
Positive error simultaneously adjusts weight, achievees the purpose that characteristic optimization combines；
Step 5: analysis model training classification results extract the ice covering thickness information of icing image.
2. the ice covering thickness detection method according to claim 1 in the powerline ice-covering image of deep learning, special
Sign is: icing image described in step 1, including ice covering on transmission lines image, shaft tower icing image, fitting icing image, electrical
Equipment icing image；It is described that answer pulling force monitoring data include shaft tower model, position, icing equivalent thickness.
3. the ice covering thickness detection method according to claim 1 in the powerline ice-covering image of deep learning, special
Sign is: pre-processing described in step 2 to original icing image, including image segmentation and edge extracting；Described image point
Cut is to carry out image dividing processing, gradient modules operator table in the form of differential operator in the edge extracting to image using iterative method
Show, is realized with fast convolution function.
4. the ice covering thickness detection method according to claim 1 in the powerline ice-covering image of deep learning, special
Sign is: establishing deep learning convolutional neural networks described in step 3, is a kind of deep learning based on BP neural network transformation
Model is deconvoluted the input of this layer in convolutional layer with convolution kernel, first that upper one layer of each output characteristic pattern position is identical
Data and the convolution kernel of this layer carry out convolution, then by all results addeds of same position convolution, obtain this layer output characteristic pattern
The output of corresponding position.
5. the ice covering thickness detection method according to claim 4 in the powerline ice-covering image of deep learning, special
Sign is: in order to reduce number of parameters, reducing model training difficulty, using weight shared mechanism, same output characteristic pattern makes
With the same convolution kernel, all corresponding filter of convolution kernel is corresponding each time, and a convolution kernel only extracts a kind of spy
Sign guarantees that aliasing does not occur for feature extraction.
6. the ice covering thickness detection method according to claim 1 in the powerline ice-covering image of deep learning, special
Sign is: in step 4.2, for convolutional layer, it is assumed that l layers are convolutional layers, then the characteristic pattern and characteristic pattern size of this layer output
It is respectively as follows:
Output.size=input.size-kernel.size+1
Wherein,It is l-1 layers of i-th of output,It is l layers of j-th of the convolution kernel inputted for i-th,It is l layers
J-th biasing, f () is activation primitive,It is l layers of j-th of output；
For sub-sampling layer, output figure are as follows:
Wherein, down (x) is to carry out sampling operation to the pixel region of input picture n × n,It is controlling elements, by sampled result
Numerical Control reduces noise jamming in colour element numberical range.
7. the ice covering thickness detection method according to claim 6 in the powerline ice-covering image of deep learning, special
Sign is: in step 4.3, the sensitivity of convolutional layer can be indicated:
Wherein, β is weight,Instead of ∑ δl+1；
The gradient of base and convolution kernel may be expressed as:
Wherein u, v are deconvoluted the corresponding part of a upper tomographic image with convolution kernel；
For sub-sampling layer, saved in propagated forwardTherefore weight gradient are as follows:
8. the ice covering thickness detection method according to claim 7 in the powerline ice-covering image of deep learning, special
Sign is: in step 4.4, feature weight αij, indicate j-th output characteristic pattern wherein i-th of input feature vector figure weight or
Contribution；
Then j-th of feature output can be rewritten as:
Above formula meets:
For single output unit, ignore footmark j, due to meeting:
With
After obtaining base gradient, convolution kernel gradient, connection weight gradient and the feature weight gradient of hidden layer, error is corrected respectively
Are as follows: base more positive value beConvolution kernel more positive value beConnection weight more positive value beFeature weight is more
Positive value be
Wherein, η is learning rate,It is the gradient of base,It is the gradient of convolution kernel,It is connection gradient, αiIt is feature power
Value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611145878.XA CN106595551B (en) | 2016-12-13 | 2016-12-13 | Ice covering thickness detection method in powerline ice-covering image based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611145878.XA CN106595551B (en) | 2016-12-13 | 2016-12-13 | Ice covering thickness detection method in powerline ice-covering image based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106595551A CN106595551A (en) | 2017-04-26 |
| CN106595551B true CN106595551B (en) | 2019-01-04 |
Family
ID=58802070
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611145878.XA Expired - Fee Related CN106595551B (en) | 2016-12-13 | 2016-12-13 | Ice covering thickness detection method in powerline ice-covering image based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106595551B (en) |
Families Citing this family (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107230205A (en) * | 2017-05-27 | 2017-10-03 | 国网上海市电力公司 | A kind of transmission line of electricity bolt detection method based on convolutional neural networks |
| CN109003254B (en) * | 2018-05-28 | 2021-04-27 | 南方电网科学研究院有限责任公司 | Method, device, equipment, system and medium for detecting icing based on logistic regression |
| CN109325956A (en) * | 2018-08-15 | 2019-02-12 | 国网河南省电力公司电力科学研究院 | A kind of transmission pressure icing section feature extracting method based on image procossing |
| CN109631794A (en) * | 2018-12-28 | 2019-04-16 | 天津大学 | Target object curvature measurement method based on convolutional neural networks |
| CN109886321B (en) * | 2019-01-31 | 2021-02-12 | 南京大学 | Image feature extraction method and device for fine-grained classification of icing image |
| CN110070530B (en) * | 2019-04-19 | 2020-04-10 | 山东大学 | Transmission line icing detection method based on deep neural network |
| CN110555400A (en) * | 2019-08-23 | 2019-12-10 | 国网山东省电力公司泗水县供电公司 | icing detection method and system based on unmanned aerial vehicle inspection overhead transmission line |
| CN110702015B (en) * | 2019-09-26 | 2021-09-03 | 中国南方电网有限责任公司超高压输电公司曲靖局 | Method and device for measuring icing thickness of power transmission line |
| CN110782431B (en) * | 2019-10-08 | 2022-03-22 | 浙江大学 | High-voltage wire icing area detection method based on deep learning |
| CN110929611A (en) * | 2019-11-12 | 2020-03-27 | 中国地质大学（武汉） | Modeling method of power transmission line icing thickness prediction model based on PR-KELM |
| CN111160424B (en) * | 2019-12-16 | 2022-12-13 | 南方电网科学研究院有限责任公司 | NFC equipment fingerprint authentication method and system based on CNN image identification |
| CN112232591A (en) * | 2020-11-02 | 2021-01-15 | 国网湖南省电力有限公司 | Icing thickness intelligent early warning method based on meteorological factors |
| CN112595276B (en) * | 2020-11-27 | 2022-07-05 | 哈尔滨工程大学 | Power transmission line icing thickness detection method based on deep learning |
| CN113222892B (en) * | 2021-04-02 | 2024-05-14 | 贵州电网有限责任公司 | Power transmission line icing monitoring and early warning method, system and readable storage medium |
| CN113486936A (en) * | 2021-06-28 | 2021-10-08 | 国网宁夏电力有限公司电力科学研究院 | Icing detection method, device and system for power transmission line equipment and storage medium |
| CN113255239B (en) * | 2021-07-12 | 2021-09-28 | 南京信息工程大学 | Deep learning method for predicting icing thickness of ultrahigh voltage transmission line tower |
| CN114354674A (en) * | 2021-12-06 | 2022-04-15 | 西安交通大学 | Experimental device for water drop impact supercooling surface and application |
| CN115423810B (en) * | 2022-11-04 | 2023-03-14 | 国网江西省电力有限公司电力科学研究院 | Blade icing form analysis method for wind generating set |
| CN116000945B (en) * | 2022-12-05 | 2024-04-12 | 湖北工业大学 | Intelligent control method of cable deicing robot |
| CN115863204B (en) * | 2023-02-27 | 2023-06-02 | 青岛芯康半导体科技有限公司 | On-line thickness monitoring and measuring method and system for wafer processing |
| CN116306893B (en) * | 2023-05-24 | 2023-07-28 | 华东交通大学 | Contact net icing early warning method |
| CN118089854A (en) * | 2024-04-19 | 2024-05-28 | 国网山西省电力公司晋城供电公司 | Overhead transmission line icing observation system and method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101793501A (en) * | 2010-04-14 | 2010-08-04 | 华中科技大学 | Transmission line ice coating status detection method based on image |
| CN103400141A (en) * | 2013-07-24 | 2013-11-20 | 华南理工大学 | Method for calculating thickness of ice coated on transmission line on basis of improved image method |
| CN103453867A (en) * | 2013-09-09 | 2013-12-18 | 国家电网公司 | Electric transmission line ice coating thickness monitoring method |
| CN104978580A (en) * | 2015-06-15 | 2015-10-14 | 国网山东省电力公司电力科学研究院 | Insulator identification method for unmanned aerial vehicle polling electric transmission line |
| CN105095670A (en) * | 2015-08-21 | 2015-11-25 | 国家电网公司 | Drawing method of electric transmission line icing-thickness growth rate distribution diagram based on number of continuous freezing rain days |
| CN105138976A (en) * | 2015-08-16 | 2015-12-09 | 东北石油大学 | Power transmission line icing thickness identification method based on genetic wavelet neural network |
-
2016
- 2016-12-13 CN CN201611145878.XA patent/CN106595551B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101793501A (en) * | 2010-04-14 | 2010-08-04 | 华中科技大学 | Transmission line ice coating status detection method based on image |
| CN103400141A (en) * | 2013-07-24 | 2013-11-20 | 华南理工大学 | Method for calculating thickness of ice coated on transmission line on basis of improved image method |
| CN103453867A (en) * | 2013-09-09 | 2013-12-18 | 国家电网公司 | Electric transmission line ice coating thickness monitoring method |
| CN104978580A (en) * | 2015-06-15 | 2015-10-14 | 国网山东省电力公司电力科学研究院 | Insulator identification method for unmanned aerial vehicle polling electric transmission line |
| CN105138976A (en) * | 2015-08-16 | 2015-12-09 | 东北石油大学 | Power transmission line icing thickness identification method based on genetic wavelet neural network |
| CN105095670A (en) * | 2015-08-21 | 2015-11-25 | 国家电网公司 | Drawing method of electric transmission line icing-thickness growth rate distribution diagram based on number of continuous freezing rain days |
Non-Patent Citations (1)
| Title |
|---|
| 基于数字图像处理的输电线路状态智能识别技术;金华等;《微计算机信息》;20120430;第28卷(第4期);第91-92页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106595551A (en) | 2017-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106595551B (en) | Ice covering thickness detection method in powerline ice-covering image based on deep learning | |
| WO2022077605A1 (en) | Wind turbine blade image-based damage detection and localization method | |
| CN110070530A (en) | A kind of powerline ice-covering detection method based on deep neural network | |
| CN108898077B (en) | Power equipment infrared chart identification method and power equipment infrared identification system | |
| CN107392247B (en) | Real-time detection method for ground object safety distance below power line | |
| CN108734143A (en) | A kind of transmission line of electricity online test method based on binocular vision of crusing robot | |
| CN105718842B (en) | A kind of machine vision detection method of power transmission line line-broken malfunction | |
| CN108537154A (en) | Transmission line of electricity Bird's Nest recognition methods based on HOG features and machine learning | |
| CN102446390A (en) | Method and system for safety detection and early warning of monitoring areas near power transmission lines | |
| CN106991666B (en) | A kind of disease geo-radar image recognition methods suitable for more size pictorial informations | |
| CN109214308A (en) | A kind of traffic abnormity image identification method based on focal loss function | |
| CN108734117A (en) | Cable machinery external corrosion failure evaluation method based on YOLO | |
| CN108680833B (en) | Composite insulator defect detection system based on unmanned aerial vehicle | |
| CN109376676A (en) | Highway engineering site operation personnel safety method for early warning based on unmanned aerial vehicle platform | |
| CN115909093A (en) | Power equipment fault detection method based on unmanned aerial vehicle inspection and infrared image semantic segmentation | |
| CN108416299A (en) | A kind of power circuit intelligent inspection system | |
| CN106250835A (en) | Bird's Nest recognition methods on the transmission line of electricity of feature based identification | |
| CN115620239B (en) | Point cloud and video combined power transmission line online monitoring method and system | |
| CN113947555A (en) | Infrared and visible light fused visual system and method based on deep neural network | |
| CN111951520A (en) | System and method for detecting icing thickness of transmission line conductor | |
| CN106250920A (en) | The insulator state detection merged based on multicharacteristic information and diagnostic method | |
| Prabhakaran et al. | Deep Learning-Based Model for Defect Detection and Localization on Photovoltaic Panels. | |
| CN113361686A (en) | Multilayer heterogeneous multi-mode convolutional neural network integrated robot inspection method | |
| CN108664886A (en) | A kind of fast face recognition method adapting to substation's disengaging monitoring demand | |
| CN116824517B (en) | Substation operation and maintenance safety control system based on visualization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190104 Termination date: 20191213 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |