CN113989257A - Electric power comprehensive pipe gallery settlement crack identification method based on artificial intelligence technology - Google Patents
Electric power comprehensive pipe gallery settlement crack identification method based on artificial intelligence technology Download PDFInfo
- Publication number
- CN113989257A CN113989257A CN202111319884.3A CN202111319884A CN113989257A CN 113989257 A CN113989257 A CN 113989257A CN 202111319884 A CN202111319884 A CN 202111319884A CN 113989257 A CN113989257 A CN 113989257A
- Authority
- CN
- China
- Prior art keywords
- image
- pipe gallery
- artificial intelligence
- electric power
- output
- 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 29
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 20
- 238000005516 engineering process Methods 0.000 title claims abstract description 19
- 238000013528 artificial neural network Methods 0.000 claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 17
- 238000005286 illumination Methods 0.000 claims abstract description 13
- 238000012937 correction Methods 0.000 claims abstract description 10
- 238000001514 detection method Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 210000002569 neuron Anatomy 0.000 claims description 32
- 230000006870 function Effects 0.000 claims description 22
- 238000000605 extraction Methods 0.000 claims description 14
- 230000005284 excitation Effects 0.000 claims description 8
- 230000001629 suppression Effects 0.000 claims description 6
- 238000009499 grossing Methods 0.000 claims description 5
- 230000002441 reversible effect Effects 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000011478 gradient descent method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 231100001261 hazardous Toxicity 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012905 input function Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30184—Infrastructure
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an artificial intelligence technology-based electric power comprehensive pipe gallery settlement crack identification method, which comprises the following steps: (1) acquiring a wall image of a pipe gallery; (2) preprocessing the wall image of the pipe gallery, analyzing the uneven illumination condition by utilizing the gray value of the image, and performing histogram equalization correction processing; (3) extracting image edge characteristics of the preprocessed image; (4) carrying out BP neural network training by adopting a back propagation algorithm; (5) and inputting the extracted image edge characteristics into the trained BP neural network and outputting a detection and identification result. The invention provides an improved back propagation algorithm suitable for detecting pipe gallery wall cracks in a complex environment, a large amount of image data collected in a power comprehensive pipe gallery, including crack pictures and normal pictures of walls in an internal structure of the pipe gallery, are learned and trained, and finally detection and identification of pipe gallery settlement cracks are achieved.
Description
Technical Field
The invention relates to the technical field of electric power comprehensive pipe rack crack identification, in particular to an electric power comprehensive pipe rack settlement crack identification method based on an artificial intelligence technology.
Background
Underground electric power utility tunnel or underground electric power utility pipeline are in long-term service in-process, because surrounding ground environment is complicated, receive different geological environment, ground building, the river, highway, the railway, piping lane structural material performance, and the long-term effect of ground load, the combined action of factors such as fatigue effect and sudden change effect, structural damage accumulation will inevitably appear in piping lane or pipe wall, can cause the piping lane to subside or the condition such as crack to a certain extent, lead to piping lane structural failure and piping lane to collapse even under extreme condition, cause major safety accident. Consequently, subside and take place the pipe wall crack that appears after the piping lane subsides and carry out intelligent monitoring to utility power pipe gallery, have the great significance to utility power pipe gallery safety operation and reduce the emergence that the piping lane subsides the accident.
According to experience, when the power utility tunnel takes place to subside, the crack can appear in the inside wall structure of general piping lane. With the rapid development of the industry 4.0 and the smart manufacturing industry, the artificial intelligence technology is increasingly applied to industrial safety production, robot smart inspection and other important fields. Wherein, intelligence inspection robot is to underground electric power utility tunnel safety monitoring, is an important application of artificial intelligence technique in the electric power industry, but current crack identification method is difficult to more accurate observe the crack, and monitoring effect can not obtain better assurance, uses to have the drawback, and can't be adapted to under the complex environment piping lane wall crack detection.
Disclosure of Invention
In order to solve the defects in the prior art, the invention aims to provide the electric power comprehensive pipe gallery settlement crack identification method based on the artificial intelligence technology, a back propagation algorithm is adopted for BP neural network training, the trained BP neural network is used for detecting and identifying the image edge characteristics of the acquired pipe gallery wall image, the crack of the pipe gallery wall is detected, and the real-time monitoring of the settlement of the internal structure of the electric power comprehensive pipe gallery is realized.
The invention adopts the following technical scheme.
An electric power comprehensive pipe gallery settlement crack identification method based on an artificial intelligence technology comprises the following steps:
(1) acquiring a wall image of a pipe gallery;
(2) preprocessing the wall image of the pipe gallery, analyzing the uneven illumination condition by utilizing the gray value of the image, and performing image histogram equalization correction processing;
(3) extracting image edge characteristics of the preprocessed image;
(4) carrying out BP neural network training by adopting a back propagation algorithm;
(5) and inputting the extracted image edge characteristics into the trained BP neural network and outputting a detection and identification result.
Further, in the step (2), the image gray-scale value analysis method includes: and converting the acquired color image into a binary gray image, respectively projecting pixels of the gray image row by row and column by column, calculating the sum of all pixels in each row and each column in the gray image, and obtaining the uneven distribution of illumination through the step change of gray projection values in the horizontal direction and the vertical direction.
Further, in the step (2), an image histogram equalization method is adopted to perform correction processing of illumination uneven distribution on the gray level image;
the mathematical expression for the image histogram equalization is:
where n is the sum of the image pixels, niIs the sum of the pixels of the ith gray level, riIs the ith gray level, i-0, 1,2, …, L-1.
Further, in the step (3), the extraction of the image edge features is performed by extracting Canny edge features of the image from the pipe gallery wall image.
Further, the Canny edge feature extraction step is as follows:
(3.1) performing Gaussian smoothing on the input image by using a Gaussian kernel function;
(3.2) calculating the gradient amplitude and direction of the image by using a Sobel operator;
(3.3) according to the gradient direction, carrying out non-maximum suppression on the gradient amplitude;
and (3.4) carrying out double-threshold processing on the image after the non-extreme value large suppression, linking edges into the contour in the high-threshold image, searching points meeting a low threshold in the neighborhood points of the broken points when the end point of the contour is reached, and collecting new edges according to the points until the edge of the whole image is closed.
Further, in the step (4), historically collected crack pictures and normal pictures containing the internal structure of the pipe gallery are subjected to image preprocessing and edge feature extraction to form a training sample; and training the BP neural network by adopting the sample data through a back propagation algorithm.
Further, the back propagation algorithm starts from the output unit and propagates the weight correction caused by the total error back to the hidden layer unit; using output layersPartial derivative of each neuron deltao(k) And the output h of each neuron of the hidden layerho(k) Calculating partial derivative delta w of error function to each neuron of hidden layerho(k) Using the output delta of each neuron of the hidden layerh(k) And input parameters of neurons of the input layerAnd correcting the connection weight value.
Further, learning the weight of the multilayer BP neural network by using a back propagation algorithm, weighting a plurality of output units comprehensively, and recalculating the reverse error E, so as to add the error weights output by all the networks, wherein the formula is as follows:
where D is the total number of training samples, outputs are the set of output units, wiIs a weighting coefficient, tkdAnd OkdThe output values associated with the training sample d and the kth output unit are determined by the corresponding input values, the hidden function, and the excitation function.
Further, a gradient descent algorithm is adopted in the back propagation algorithm, and the square error between the output value of the BP neural network output unit and the target value is minimized.
Further, in the step (5), the extracted edge image is converted into an initial value of neuron input, where each neuron unit has a certain number of real-valued inputs and generates a single real-valued output, and a final output result of the neuron is shown as follows:
wherein f is the excitation function in the neural network, and the input units in the neurons are x respectivelyiCorresponding to a weighting coefficient of wi,i=1,2,3,…,n-1。
Compared with the prior art, the method has the beneficial effects that the improved back propagation algorithm suitable for detecting the pipe gallery wall cracks in the complex environment is provided, a large amount of image data collected in the power comprehensive pipe gallery, including the crack pictures and normal pictures of the wall in the internal structure of the pipe gallery, are learned and trained, and finally, the detection and identification of the pipe gallery settlement cracks are realized.
Considering the relevance among all units, the back propagation algorithm adopts a gradient descent method, reduces the square error between the network output value and the target value, comprehensively weights a plurality of output units, recalculates the reverse error E, and thus weights and adds the errors of all network outputs.
The uneven illumination of the pipe gallery wall image is analyzed by utilizing the image gray value, the uneven illumination is weakened through image light correction, and an excessively dark or excessively bright area can be corrected to the original natural color.
Drawings
FIG. 1 is a flow chart of an electric power comprehensive pipe gallery settlement crack identification method based on an artificial intelligence technology;
FIG. 2 is a schematic diagram of a three-layer BP neural network;
FIG. 3 is a schematic diagram of a neuron structure in a BP neural network;
FIG. 4 is a picture of a tube lane crack;
FIG. 5 is a schematic diagram of tube lane crack edge feature extraction;
figure 6 is electric utility tunnel settlement crack detection effect schematic diagram.
Detailed Description
The present application is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present application is not limited thereby.
As shown in fig. 1, the method for identifying settlement cracks of the power utility tunnel based on the artificial intelligence technology comprises the following steps:
(1) acquiring a wall image of the pipe gallery as image input;
patrol and examine the robot through the intelligence and carry out real-time shooting to pipe gallery wall under the complicated hazardous environment, acquire pipe gallery wall image, as shown in figure 4.
(2) Preprocessing the wall image of the pipe gallery;
and analyzing the acquired pipe gallery wall image by using the image gray value and carrying out correction processing. Analyzing the illumination distribution condition of the image on the wall of the pipe gallery by utilizing the gray value of the image, acquiring areas with uneven illumination, including areas with over-dark or over-bright areas, correcting the uneven condition of the image light by a gradient histogram method, and correcting the areas with over-dark or over-bright areas to the original natural color.
Pipe rack wall surrounding environment is complicated, and various power equipment install inside the pipe rack, cause the robot to shelter from because the light of replenishing the light source is sheltered from by the power equipment of installation in the pipe rack when patrolling and examining, lead to appearing light on the wall and shelter from the edge, influence the extraction at the true edge of wall.
And judging whether the region of interest has gray shade or not by analyzing the gray projection value of the edge of the region of interest by using an image gray value analysis method.
Firstly, converting the obtained color image into a binary gray image, then respectively projecting pixels of the image row by row and column by column, and calculating the sum of all pixels in each row and each column in the image. For a region of interest of an m × n size image I (x, y), a pixel grayscale projection value in the I-th row horizontal direction is calculated:
pixel gray projection value in the j-th column in the vertical direction:
by analyzing the gray projection values in the horizontal direction and the vertical direction, if the pixel gray projection values in any direction have obvious step changes, the step changes of the image gray values caused by the shielding of the projection light can be considered to exist.
Based on the analysis, an image histogram method is utilized to perform histogram equalization on the image, and the interference edge caused by the fact that the light is shielded is eliminated as much as possible.
The image histogram is a statistical relationship representing the frequency of occurrence of each gray level in the digital image. The histogram can give a general description of the gray scale range of the image, the frequency of each gray scale and the distribution of the gray scale, the average brightness and contrast of the whole image, and the like. Assuming that the gray scale range of the gray image is [0, L-1], its histogram is defined as:
p(rk)=nk/n
where n is the sum of the image pixels, nkIs the sum of pixels of the kth gray level, rkIs the kth gray level, k being 0,1,2, …, L-1.
The mathematical expression for histogram equalization of an image is as follows:
the histogram equalization is carried out on the image by using the above formula, namely, the equalization of the gray value of the image can be realized, and the influence of the interference edge caused by light shielding on the extraction of the real crack edge of the pipe gallery wall is reduced.
(3) Extracting image edge characteristics of the preprocessed image;
as shown in fig. 5, the image edge feature extraction is performed on the preprocessed pipe gallery wall image, and the extracted edge feature is used as the input of the BP neural network by analyzing the edge feature of the pipe gallery wall structure.
The extraction of image edge features is carried out on the pipe gallery wall image by extracting Canny edge features of the image, and the extracted edge features are used as input of a BP neural network input unit.
The Canny edge extraction algorithm is as follows:
firstly, Gaussian smoothing is carried out on an input image, smoothing of noise points is achieved by utilizing a Gaussian kernel function, interference of various noises on image feature extraction is reduced, and edge error rate generated by image noises is reduced. The gaussian kernel function is defined as follows:
where σ is the standard deviation, the size is set to 0.6 for the canal corridor image, and the gaussian smoothing window size is set to 7 × 7 pixels.
Then, the gradient magnitude and direction are calculated to estimate the edge strength and direction at each point of the image. Convolving the input image by using a Sobel operator, assuming that the original image is I (x, y), and Sobel gradient descriptors in the horizontal direction and the vertical direction are respectively:
then, with the convolution operation on the image I (x, y):
where is the sign of the convolution.
The Sobel gradient magnitude of image I (x, y) is then:
G=|Gx|+|Gy|
secondly, according to the gradient direction, the non-maximum value suppression is carried out on the gradient amplitude. For each pixel point, the gradient direction of the pixel point is approximate to one of the following values (0,45,90,135,180,225,270,315), the gradient strength of the pixel point and the gradient direction of the pixel point in the positive and negative directions is compared, if the gradient strength of the pixel point is maximum, the pixel point is kept, and if the gradient strength of the pixel point is maximum, the pixel point is inhibited (deleted, namely set to be 0).
Finally, edges are processed and connected with dual thresholds. After the non-extreme value large suppression, a plurality of noise points still exist in the image. And (3) processing by using double thresholds, namely setting an upper threshold and a lower threshold, considering that pixel points in the image are bound if the pixel points are larger than the upper threshold, considering that the pixel points are not bound if the pixel points are smaller than the lower threshold, and considering that the pixel points are candidates between the pixel points and the lower threshold. By a weak boundary connected to a strong boundary being considered a boundary, other weak boundaries are suppressed. The edges are linked into the contour in the high-threshold image, when the end point of the contour is reached, the algorithm searches for a point meeting the low threshold in the neighborhood of the broken point, and then collects new edges according to the point until the edge of the whole image is closed.
(4) Carrying out BP neural network training by adopting a back propagation algorithm;
a large amount of historically acquired image data in the power comprehensive pipe rack, including crack pictures and normal pictures in the internal structure of the pipe rack, are subjected to image preprocessing and edge feature extraction to form a training sample D; and training the BP neural network by adopting the sample data through a back propagation algorithm.
The BP neural network is composed of a plurality of neuron units to form a huge multi-layer BP neural network system. The neuron unit comprises an input unit, an output unit and one or more hidden layer units. Taking three-layer BP neural network as an example, the structure is shown in FIG. 2, xiIs an input value, yiThe output value is i ═ 1,2, and 3.
The back propagation algorithm starts from the output unit and propagates the weight correction caused by the total error back to the hidden layer unit.
First, the partial derivative delta of each neuron in the output layer is usedo(k) And the output h of each neuron of the hidden layerho(k) Calculating partial derivative delta w of error function to each neuron of hidden layerho(k):
Δwho(k)=-μδo(k)hho(k)
Wherein mu is a corresponding proportionality coefficient.
Using the output delta of each neuron of the hidden layerh(k) And input parameters of neurons of the input layerCorrecting the connection weight:
wherein eta is a corresponding proportionality coefficient.
Considering the relevance among the units of the BP neural network, learning the weight of the multilayer BP neural network by using a back propagation algorithm, comprehensively weighting a plurality of output units, and recalculating the reverse error E, thereby weighting and adding the errors output by all the networks, wherein the formula is as follows:
where D is the total number of training samples, outputs are the set of output units, wiIs a weighting coefficient, tkdAnd OkdThe output values associated with the training sample d and the kth output unit are determined by the corresponding input values, the hidden function, and the excitation function.
The hidden function and the excitation function used in the present invention are defined as follows, respectively:
suppose we want to hide the parameter W of a layer to the k-th layer(k)And partial derivative b(k). Suppose z(k)The input representing the k-th layer neurons, i.e., defining the hidden function at this time, is:
z(k)=W(k)*n(k-1)+b(k)
sigmoid excitation function:
and a gradient descent method is adopted in the back propagation algorithm, so that the square error between the output value of the BP neural network output unit and the target value is reduced.
In the gradient descent algorithm, the error function is set as:
J=0.5(y-o)2=0.5(y-f(∑Wixi))2
where y is the actual result, o is the predicted result, and o ═ f (∑ W)ixi) Results of weighted outputs for various input functions in a neural network, WiWeighting coefficients, x, for corresponding input channelsiThe neuron inputs a value.
The gradient descent algorithm process is to minimize the function J and solve the corresponding WiThe process of (1). Defining the rate of change as α, then:
by using the above formula, W can be obtained through a plurality of operationsiThe purpose of reducing errors is achieved by using the gradient descent algorithm.
(5) Inputting the extracted image edge characteristics into a trained BP neural network and outputting a detection recognition result;
and detecting and identifying the obtained image edge characteristics by using the trained BP neural network, judging whether the wall has cracks, and finally outputting a detection result and early warning information, as shown in FIG. 6.
As shown in fig. 3, the BP neural network converts the extracted edge image into an initial neuron input value according to the edge feature extracted from the image, where each neuron unit has a certain number of real inputs and generates a single real value output, and the final output result of the neuron is shown as follows:
wherein f is the excitation function in the neural network, and the input units in the neurons are x respectivelyiCorresponding to a weighting coefficient of wi,i=1,2,3,…,n-1。
Compared with the prior art, the method has the beneficial effects that the improved back propagation algorithm suitable for detecting the pipe gallery wall cracks in the complex environment is provided, a large amount of image data collected in the power comprehensive pipe gallery, including the crack pictures and normal pictures of the wall in the internal structure of the pipe gallery, are learned and trained, and finally, the detection and identification of the pipe gallery settlement cracks are realized.
Considering the relevance among all units, the back propagation algorithm adopts a gradient descent method, reduces the square error between the network output value and the target value, comprehensively weights a plurality of output units, recalculates the reverse error E, and thus weights and adds the errors of all network outputs.
The uneven illumination of the pipe gallery wall image is analyzed by utilizing the image gray value, the uneven illumination is weakened through image light correction, and an excessively dark or excessively bright area can be corrected to the original natural color.
The present applicant has described and illustrated embodiments of the present invention in detail with reference to the accompanying drawings, but it should be understood by those skilled in the art that the above embodiments are merely preferred embodiments of the present invention, and the detailed description is only for the purpose of helping the reader to better understand the spirit of the present invention, and not for limiting the scope of the present invention, and on the contrary, any improvement or modification made based on the spirit of the present invention should fall within the scope of the present invention.
Claims (10)
1. The method for identifying the settlement cracks of the power comprehensive pipe gallery based on the artificial intelligence technology is characterized by comprising the following steps of:
(1) acquiring a wall image of a pipe gallery;
(2) preprocessing the wall image of the pipe gallery, analyzing the uneven illumination condition by utilizing the gray value of the image, and performing image histogram equalization correction processing;
(3) extracting image edge characteristics of the preprocessed image;
(4) carrying out BP neural network training by adopting a back propagation algorithm;
(5) and inputting the extracted image edge characteristics into the trained BP neural network and outputting a detection and identification result.
2. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 1,
in the step (2), the image gray-scale value analysis method includes: and converting the acquired color image into a binary gray image, respectively projecting pixels of the gray image row by row and column by column, calculating the sum of all pixels in each row and each column in the gray image, and obtaining the uneven distribution of illumination through the step change of gray projection values in the horizontal direction and the vertical direction.
3. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 2,
in the step (2), an image histogram equalization method is adopted to correct uneven illumination distribution of the gray level image;
the mathematical expression for the image histogram equalization is:
where n is the sum of the image pixels, niIs the sum of the pixels of the ith gray level, riIs the ith gray level, i-0, 1,2, …, L-1.
4. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 1,
in the step (3), the extraction of the image edge features is carried out on the pipe gallery wall image by extracting the Canny edge features of the image.
5. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 4,
the Canny edge feature extraction steps are as follows:
(3.1) performing Gaussian smoothing on the input image by using a Gaussian kernel function;
(3.2) calculating the gradient amplitude and direction of the image by using a Sobel operator;
(3.3) according to the gradient direction, carrying out non-maximum suppression on the gradient amplitude;
and (3.4) carrying out double-threshold processing on the image after the non-extreme value large suppression, linking edges into the contour in the high-threshold image, searching points meeting a low threshold in the neighborhood points of the broken points when the end point of the contour is reached, and collecting new edges according to the points until the edge of the whole image is closed.
6. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 1,
in the step (4), historically collected crack pictures and normal pictures in the internal structure of the pipe gallery are subjected to image preprocessing and edge feature extraction to form a training sample; and training the BP neural network by adopting the sample data through a back propagation algorithm.
7. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 6,
the back propagation algorithm starts from the output unit and propagates the weight correction caused by the total error back to the hidden layer unit; using partial derivatives delta of neurons in the output layero(k) And the output h of each neuron of the hidden layerho(k) Calculating partial derivative delta w of error function to each neuron of hidden layerho(k) Using the output delta of each neuron of the hidden layerh(k) And input parameters of neurons of the input layerAnd correcting the connection weight value.
8. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 7,
learning the weight of the multilayer BP neural network by using a back propagation algorithm, weighting a plurality of output units comprehensively, and recalculating the reverse error E, thereby weighting and adding the errors output by all the networks, wherein the formula is as follows:
where D is the total number of training samples, outputs are the set of output units, wiIs a weighting coefficient, tkdAnd OkdThe output values associated with the training sample d and the kth output unit are determined by the corresponding input values, the hidden function, and the excitation function.
9. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 7,
a gradient descent algorithm is adopted in the back propagation algorithm, and the square error between the output value of the BP neural network output unit and the target value is minimized.
10. The artificial intelligence technology-based electric power utility tunnel settlement fracture identification method according to claim 1,
in the step (5), the extracted edge image is converted into an initial neuron input value, wherein each neuron unit has a certain number of real value inputs and generates a single real value output, and a final output result of the neuron is shown as the following formula:
wherein f is the excitation function in the neural network, and the input units in the neurons are x respectivelyiCorresponding to a weighting coefficient of wi,i=1,2,3,…,n-1。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111319884.3A CN113989257A (en) | 2021-11-09 | 2021-11-09 | Electric power comprehensive pipe gallery settlement crack identification method based on artificial intelligence technology |
PCT/CN2021/138598 WO2023082418A1 (en) | 2021-11-09 | 2021-12-16 | Power utility tunnel settlement crack identification method based on artificial intelligence technology |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111319884.3A CN113989257A (en) | 2021-11-09 | 2021-11-09 | Electric power comprehensive pipe gallery settlement crack identification method based on artificial intelligence technology |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113989257A true CN113989257A (en) | 2022-01-28 |
Family
ID=79747328
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111319884.3A Pending CN113989257A (en) | 2021-11-09 | 2021-11-09 | Electric power comprehensive pipe gallery settlement crack identification method based on artificial intelligence technology |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN113989257A (en) |
WO (1) | WO2023082418A1 (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116630321B (en) * | 2023-07-24 | 2023-10-03 | 铁正检测科技有限公司 | Intelligent bridge health monitoring system based on artificial intelligence |
CN116630813B (en) * | 2023-07-24 | 2023-09-26 | 青岛奥维特智能科技有限公司 | Highway road surface construction quality intelligent detection system |
CN117037006B (en) * | 2023-10-09 | 2023-12-15 | 山东中宇航空科技发展有限公司 | Unmanned aerial vehicle tracking method with high endurance capacity |
CN117237368B (en) * | 2023-11-16 | 2024-01-26 | 临沂市公路事业发展中心兰陵县中心 | Bridge crack detection method and system |
CN117593651B (en) * | 2024-01-18 | 2024-04-05 | 四川交通职业技术学院 | Tunnel crack segmentation recognition method |
CN117975374B (en) * | 2024-03-29 | 2024-06-18 | 山东天意机械股份有限公司 | Intelligent visual monitoring method for double-skin wall automatic production line |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108629775A (en) * | 2018-05-14 | 2018-10-09 | 华中科技大学 | A kind of hot high-speed rod surface image processing method |
CN109685786A (en) * | 2018-12-20 | 2019-04-26 | 龙口味美思环保科技有限公司 | A kind of non-destructive testing of birds, beasts and eggs face crack and automatic grading method |
CN110119687A (en) * | 2019-04-17 | 2019-08-13 | 浙江工业大学 | Detection method based on the road surface slight crack defect that image procossing and convolutional neural networks combine |
CN112233067A (en) * | 2020-09-21 | 2021-01-15 | 武汉钢铁有限公司 | Hot rolled steel coil end face quality detection method and system |
CN112766460A (en) * | 2021-01-12 | 2021-05-07 | 上海电机学院 | Weld joint identification method based on sub-region and BP neural network |
WO2021118463A1 (en) * | 2019-12-09 | 2021-06-17 | Agency For Science, Technology And Research | Defect detection in image space |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109034530A (en) * | 2018-06-19 | 2018-12-18 | 天津大学 | A kind of piping lane planing method based on neural network algorithm |
CN110044268B (en) * | 2019-03-19 | 2021-08-31 | 天津大学前沿技术研究院有限公司 | Shield tunnel joint opening and slab staggering monitoring system based on optical fiber reflection principle |
CN111401131A (en) * | 2020-02-13 | 2020-07-10 | 深圳供电局有限公司 | Image processing method and device for tunnel pipe gallery, computer equipment and storage medium |
CN112861681B (en) * | 2021-01-29 | 2023-04-07 | 长兴云尚科技有限公司 | Pipe gallery video intelligent analysis method and system based on cloud processing |
-
2021
- 2021-11-09 CN CN202111319884.3A patent/CN113989257A/en active Pending
- 2021-12-16 WO PCT/CN2021/138598 patent/WO2023082418A1/en unknown
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108629775A (en) * | 2018-05-14 | 2018-10-09 | 华中科技大学 | A kind of hot high-speed rod surface image processing method |
CN109685786A (en) * | 2018-12-20 | 2019-04-26 | 龙口味美思环保科技有限公司 | A kind of non-destructive testing of birds, beasts and eggs face crack and automatic grading method |
CN110119687A (en) * | 2019-04-17 | 2019-08-13 | 浙江工业大学 | Detection method based on the road surface slight crack defect that image procossing and convolutional neural networks combine |
WO2021118463A1 (en) * | 2019-12-09 | 2021-06-17 | Agency For Science, Technology And Research | Defect detection in image space |
CN112233067A (en) * | 2020-09-21 | 2021-01-15 | 武汉钢铁有限公司 | Hot rolled steel coil end face quality detection method and system |
CN112766460A (en) * | 2021-01-12 | 2021-05-07 | 上海电机学院 | Weld joint identification method based on sub-region and BP neural network |
Non-Patent Citations (2)
Title |
---|
人工智能技术干货: "一种基于灰度图积分投影和高光区域处理的光照方向估计方法", 《HTTPS://ZHUANLAN.ZHIHU.COM/P/272657255》, pages 17 - 21 * |
张硕: "利用图像处理技术的隧道裂缝检测方法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》, pages 1 - 89 * |
Also Published As
Publication number | Publication date |
---|---|
WO2023082418A1 (en) | 2023-05-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113989257A (en) | Electric power comprehensive pipe gallery settlement crack identification method based on artificial intelligence technology | |
CN113469177B (en) | Deep learning-based drainage pipeline defect detection method and system | |
CN111899224A (en) | Nuclear power pipeline defect detection system based on deep learning attention mechanism | |
CN110726725A (en) | Transmission line hardware corrosion detection method and device | |
CN111899225A (en) | Nuclear power pipeline defect detection method based on multi-scale pyramid structure | |
CN109781737B (en) | Detection method and detection system for surface defects of hose | |
CN113313107B (en) | Intelligent detection and identification method for multiple types of diseases on cable surface of cable-stayed bridge | |
CN112508245A (en) | Method and system for predicting power generation capacity based on artificial intelligence photovoltaic cell panel glass breakage | |
CN116665011A (en) | Coal flow foreign matter identification method for coal mine belt conveyor based on machine vision | |
CN113379737A (en) | Intelligent pipeline defect detection method based on image processing and deep learning and application | |
CN117036255A (en) | Pipeline defect detection and evaluation method and device based on deep learning | |
US20230222643A1 (en) | Semantic deep learning and rule optimization for surface corrosion detection and evaluation | |
CN115147418A (en) | Compression training method and device for defect detection model | |
CN116910752A (en) | Malicious code detection method based on big data | |
CN111091554B (en) | Railway wagon swing bolster fracture fault image identification method | |
CN113313678A (en) | Automatic sperm morphology analysis method based on multi-scale feature fusion | |
CN116310358A (en) | Method, storage medium and equipment for detecting bolt loss of railway wagon | |
CN112785594B (en) | Automatic bridge structure crack identification method based on image two-dimensional amplitude estimation | |
Wang et al. | Rust defect detection and segmentation method for tower crane | |
CN114882375A (en) | Intelligent identification method and device for tailing pond | |
CN114757941A (en) | Transformer substation equipment defect identification method and device, electronic equipment and storage medium | |
CN114821098A (en) | High-speed pavement damage detection algorithm based on gray gradient fusion characteristics and CNN | |
CN114120061A (en) | Small target defect detection method and system for power inspection scene | |
CN114648507A (en) | Method for carrying out online detection on aggregate, readable storage medium and terminal equipment | |
CN111767815A (en) | Tunnel water leakage identification method |
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 |