CN113763327A - CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method - Google Patents
CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method Download PDFInfo
- Publication number
- CN113763327A CN113763327A CN202110914532.6A CN202110914532A CN113763327A CN 113763327 A CN113763327 A CN 113763327A CN 202110914532 A CN202110914532 A CN 202110914532A CN 113763327 A CN113763327 A CN 113763327A
- Authority
- CN
- China
- Prior art keywords
- cbam
- res
- pressure steam
- unet
- steam leakage
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000011156 evaluation Methods 0.000 claims abstract description 15
- 238000003709 image segmentation Methods 0.000 claims abstract description 11
- 238000005070 sampling Methods 0.000 claims description 17
- 230000004913 activation Effects 0.000 claims description 15
- 238000011176 pooling Methods 0.000 claims description 15
- 230000007246 mechanism Effects 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 abstract description 36
- 238000013135 deep learning Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 15
- 230000008569 process Effects 0.000 description 15
- 230000000694 effects Effects 0.000 description 7
- 238000002474 experimental method Methods 0.000 description 6
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 230000000052 comparative effect Effects 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008602 contraction Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000012360 testing method Methods 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
- G06T7/0004—Industrial image inspection
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F17—STORING OR DISTRIBUTING GASES OR LIQUIDS
- F17D—PIPE-LINE SYSTEMS; PIPE-LINES
- F17D5/00—Protection or supervision of installations
- F17D5/02—Preventing, monitoring, or locating loss
- F17D5/06—Preventing, monitoring, or locating loss using electric or acoustic means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Acoustics & Sound (AREA)
- Mechanical Engineering (AREA)
- Image Analysis (AREA)
- Examining Or Testing Airtightness (AREA)
Abstract
The invention relates to a CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method, which comprises the following steps: and acquiring a high-pressure steam leakage picture to be tested, adding a loss function and an evaluation standard into a CBAM-Res _ Unet network model, carrying out image segmentation on the high-pressure steam leakage picture, and finally outputting a detection result. Compared with the prior art, the method has the advantages that the deep learning network CBAM-Res _ Unet with stronger adaptability to high-pressure steam leakage is used, the stability of network training is improved after a Loss function Dice local function and a Focal local function are combined, the high-pressure steam of the pipeline of the power plant is monitored, the method has better universality on the high-pressure steam leakage of the pipeline of the power plant, the accuracy of detecting a steam leakage area is high, and the like.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method.
Background
Steam leakage easily occurs to pipelines of a power plant due to the influence of factors such as environment, human factors and the like, which not only wastes resources, but also threatens the life safety of field workers. And the long-term stable operation of power plant equipment is related to the sustainability of the power supply of the whole power system, so that the method has great practical significance for rapidly and accurately identifying whether the pipeline has steam leakage or not.
Most methods of gas leak detection can be divided into two categories: one is hardware based and the other is software based. The hardware-based method generally uses a sensor or a detector, but the method has a high requirement on leaked gas, and most of gas leakage detection hardware devices with high precision have high cost, so the software-based method has a wider application prospect at present, for example, in 2014, the first proposed FCN network draws a sequential curtain based on deep learning pixel-level semantic segmentation, and the subsequently proposed UNet and deep labv series continue the concept of FCN, but the prior art still has the problem of poor adaptability, and for the event that high-pressure steam leakage often has a complex situation, the detection accuracy is more difficult to guarantee.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a CBAM-Res _ Unet-based method and equipment for detecting high-pressure steam leakage of a power plant pipeline, so that the accurate detection of the high-pressure steam leakage condition of the power plant is realized
The purpose of the invention can be realized by the following technical scheme:
a CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method is characterized by comprising the following steps:
acquiring a high-pressure steam leakage picture to be tested, performing image segmentation on the high-pressure steam leakage picture by using a CBAM-Res _ Unet network model, and outputting a detection result;
the left side of the CBAM-Res _ Unet network model is provided with a down-sampling layer, each layer passes through two convolution layers of n x n and a maximum pooling layer of m x m, an activation function is a linear rectification function, the right side of the CBAM-Res _ Unet network model is provided with an up-sampling layer, each layer passes through two convolution layers of n x n and a convolution layer of p x p, the activation function is an S-shaped growth curve function, a residual error network module is added to each side of the activation function, and an attention mechanism module is added to the residual error network module; meanwhile, a Loss function and an evaluation standard are added into the CBAM-Res _ Unet network model, the Loss function is the combination of a Dice Loss function and a Focal Loss function, and the evaluation standard is F1_ score.
Further, the loss function L is expressed as:
where C denotes the number of samples input, i denotes the number of classes including regions, i.e., a total of both the steam leakage region and the steam non-leakage region, and TPp(i) Indicating true positives for i, FNp(i) Indicates false negatives of i, FPp(i) False positives for i; p is a radical ofn(i) Representing a predicted value for i; gn(i) Represents the true value of i; λ represents a weight; α and β are penalty values for false negatives and false positives, respectively, both set to 0.5; n represents all pixel values of the image.
Furthermore, the parameter settings of the two n × n convolutional layers of the residual error network module in the up-sampling process and the down-sampling process of the CBAM-Res _ Unet network model are the same, and the convolutional layer parameters are reduced by using the dropot.
Further, in the CBAM-Res _ Unet network model, the channel attention submodule in the attention mechanism module performs global average pooling and maximum pooling on the input feature map, performs addition operation by using the MLP multilayer perceptron, and generates an output feature map of the channel attention submodule by using an S-type growth curve activation function.
Further, the spatial attention submodule of the attention mechanism module in the CBAM-Res _ Unet network model takes a multiplication result of the input feature map and the output feature map of the channel attention module based on elements as an input feature map of the spatial attention module, changes the number of channels into 1 through convolution operation after global average pooling and maximum pooling, and generates an output feature map of the spatial attention submodule by using an S-type growth curve activation function, where the output feature map of the attention mechanism module is a multiplication result of the input feature map and the output feature map of the spatial attention module.
Furthermore, in the up-sampling process of the CBAM-Res _ Unet network model, the number of the characteristic channels is reduced by half through deconvolution, the input characteristic diagram is cut and copied, and then the characteristic diagrams with the same size in the down-sampling process and the up-sampling process are spliced.
Further, the method for training the CBAM-Res _ Unet network model includes the steps of:
s1, extracting a picture from every other multiframe of the video of the high-pressure steam leakage of the power plant pipeline to form a picture set,
s2, randomly extracting a set number of pictures from the picture set as initial sample pictures;
s3, marking the initial sample graph into a high-pressure steam leakage area and a high-pressure steam non-leakage area to obtain a labeled sample graph;
s4, preprocessing the initial sample graph and the labeled sample graph, and adjusting the sizes of the initial sample graph and the labeled sample graph to set values;
s5, increasing the number of multiple samples of the processed initial sample image and the labeled sample image by a data enhancement method;
and S6, taking the initial sample graph and the labeled sample graph after data enhancement as training data sets, and training the CBAM-Res _ Unet network model after adding a loss function.
Further, the calculation expression of the F1_ score is:
in the formula, TP represents true positive rate, FP represents false positive rate, P represents accuracy, R represents recall rate, and F represents F1_ score.
Further, the preprocessing in step S4 is a graying operation.
Further, the data enhancement method in step S5 includes rotating, mirroring, flipping, horizontally or vertically flipping, and the like, the picture and label data.
Compared with the prior art, the invention has the following advantages:
1. the CBAM-Res _ Unet network model is added with a residual error network module on the basis of UNet, so that the characteristic information beneficial to analyzing the high-pressure steam leakage condition is obtained, and meanwhile, an attention mechanism module is integrated, so that the method can adapt to the background which is changed continuously due to high-pressure steam leakage speed; meanwhile, aiming at the condition that the CBAM-Res _ Unet network model is unstable when a steam leakage area is small, the combination of Loss functions Dice Loss and Focal Loss is used in network training, so that the detection stability is improved; and the F1_ score is used as an evaluation criterion, so that the evaluation of the image segmentation result is more accurate. The algorithm is applied to the detection of the high-pressure steam leakage of the power plant pipeline for the first time at present, and has certain innovativeness and practicability.
2. Data enhancement, data preprocessing and other operations are carried out on the data in the acquired sample training set, so that the precision and the number of training samples are improved, the training quality is guaranteed, and the final detection efficiency of high-pressure steam leakage is higher.
Drawings
FIG. 1 is a schematic flow chart of the present invention.
Fig. 2 is a schematic diagram of a CBAM-Res _ UNet network structure used in the present invention.
FIG. 3 is a schematic diagram of a CBAM structure in a CBAM-Res _ UNet network model used in the present invention.
FIG. 4 is a schematic diagram of a channel attention module structure in a CBAM according to the present invention.
FIG. 5 is a schematic diagram of a spatial attention module in a CBAM according to the present invention.
FIG. 6 is a schematic image of a partial vapor leak sample used in the present invention.
FIG. 7 is a schematic illustration of a partial vapor leak sample image and its corresponding label sample used in the present invention.
FIG. 8 is a schematic diagram of a power plant pipeline high pressure steam leakage detection result of different network models obtained through a comparison experiment.
FIG. 9 is a schematic diagram of a detection result of high-pressure steam leakage of a power plant pipeline with different loss functions obtained through a comparative experiment.
FIG. 10 is a schematic diagram of a power plant pipeline high pressure steam leakage detection result with different evaluation criteria obtained by a comparative experiment according to the present invention.
FIG. 11 is a schematic diagram of evaluation standard values of different network models obtained by comparative experiments.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
The embodiment provides a CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method, which mainly comprises the following steps as shown in the steps indicated by solid arrows in FIG. 1:
and acquiring a high-pressure steam leakage picture to be tested, performing image segmentation on the high-pressure steam leakage picture by using the trained CBAM-Res _ Unet network model, outputting a detection result, and uploading the result to a software platform and giving an alarm if a leakage area occurs.
In the network model CBAM-Res _ Unet network model adopted in this embodiment, based on the Unet network, a residual error network module (residual _ block) structure is added to the feature extraction network to obtain feature information beneficial to high-pressure steam leakage detection, in order to adapt to a background which changes continuously due to a high-pressure steam leakage speed, an attention mechanism module (CBAM) is simultaneously incorporated, suitable weight coefficients are newly allocated to pixels of different levels of image features to form a new network structure as shown in fig. 2, a contraction path is on the left, which is called a down-sampling process, and is used for extracting high-resolution feature information, each layer is subjected to two 3 × 3 convolutions and a 2 × 2 maximum pooling layer, an activation function selects a linear rectification function (ReLU), and the number of feature channels is correspondingly increased by 4 times after 4 times of down-sampling; the expansion path is arranged on the right side, the up-sampling process is also used, the function is to extract low-resolution feature information, the number of feature channels is reduced by half through deconvolution ConvTranspose2D in each up-sampling process, then the feature graphs with the same size of the contraction path and the expansion path are spliced on the basis of the channels through copying and cutting the size of the feature graphs, the reduction of segmentation precision caused by direct up-sampling is reduced, each layer of up-sampling process is also subjected to two 3-times-3 convolution, the last step is subjected to 1-times-1 convolution layer, an S-shaped growth curve (Sigmoid) activation function is selected on the basis of a two-classification problem, and finally a two-bit feature vector is output.
The method comprises the steps that two residual _ block modules are adopted in each downsampling process, each residual _ block comprises two convolution layers, the size of each convolution core is 3 x 3, the number of the initial convolution cores is 16, the step length is set to be 1, based on the structural characteristics of CNN, ReLU is adopted as an activation function, and BatchNormal is used for input images so that the input of each convolution layer in network training can be distributed in the same mode.
Two residual _ blocks are also used in each upsampling process, and the convolutional layer parameter settings are the same as in the downsampling process. Because the depth of the network is deepened, unimportant network parameters are increased correspondingly, and dropout is used in the up-sampling and down-sampling processes to reduce some unnecessary parameters, so that the burden of network training is relieved.
Meanwhile, the CBAM is added into each residual _ block, so that the network can add different weights to different features. The CBAM is composed of a space attention mechanism and a channel attention mechanism module, the whole module is shown in figure 3, wherein the channel attention module is shown in figure 4, the structure of the channel attention module is used for carrying out global average pooling and maximum pooling on an input feature map of the CBAM, carrying out pixel-based addition operation on two feature maps of a multilayer perceptron MLP, and finally generating a final feature map of the module through an activation function Sigmoid. The structure of the spatial attention module is shown in fig. 5, and the input feature diagram of the spatial attention module is obtained by performing element-based multiplication operation on the output feature diagram and the input feature diagram of the channel attention module, performing channel-based addition operation on the feature diagrams subjected to global average pooling and maximum pooling, changing the number of channels into 1 through convolution operation, and finally generating the final feature diagram of the spatial attention module through an activation function Sigmoid. The output profile of the CBAM is the result of multiplying the output profile of the spatial attention module by the input profile.
The training step of the CBAM-Res _ Unet network model is shown as a step guided by a dotted line in fig. 1, and specifically includes:
step S1, extracting a picture from every ten frames of the video of the high-pressure steam leakage of the power plant pipeline to form a picture set, where the picture can refer to fig. 6.
And step 2, randomly extracting 50 pictures from the picture set as an initial sample picture.
Step S3, marking the initial sample map by using the software labelme, dividing the initial sample map into a high-pressure steam leakage area and a high-pressure steam non-leakage area, and obtaining a labeled sample map, which can refer to fig. 7.
And step S4, preprocessing the initial sample image and the labeled sample image, including graying, and batch cutting the sizes of the initial sample image and the labeled sample image into 512 x 512 by using a bilinear interpolation principle.
And step S5, increasing the number of data samples to 10 times of the original number by using data enhancement methods such as rotation, mirror image, flip, horizontal or vertical flip, and the like, for the processed initial sample image and the labeled sample image.
Step S6, taking the initial sample graph and the labeled sample graph after data enhancement as a training data set, and adding a loss function Dice + Focal loss, where the loss function has an expression:
where C denotes the number of samples input, and i denotes the number of categories including regions, i.e., a total of both the steam leakage region and the steam non-leakage region. TPp(i) Indicating true positives for i, FNp(i) Indicates false negatives of i, FPp(i) False positives for i; p is a radical ofn(i) Representing a predicted value for i; gn(i) Represents the true value of i; λ represents a weight; α and β are penalty values for false negatives and false positives, respectively, both set to 0.5; n represents all pixel values of the image.
After 400 times of iterative training, the CBAM-Res _ Unet network model used in the present embodiment is obtained.
After the CBAM-Res _ Unet network model is obtained, the segmentation capability of the network model is evaluated by using an evaluation standard F1_ score.
The computational expression for the F1_ score is:
in the formula, TP represents true positive rate, FP represents false positive rate, P represents accuracy, R represents recall rate, and F represents F1_ score. A closer P to 1 indicates a better partitioning capability of the network.
In the embodiment, a comparison test is performed on the design of a deep learning network model during the inventive conception, fig. 8 shows the image segmentation effect of the traditional Unet network model, the Unet network model only added with CBAM, the Res Unet network model and the CBAM-Res _ Unet network model used in the embodiment, and the comparison between the effect graph and the label graph in fig. 8 clearly shows that the image segmentation effect of the CBAM-Res _ Unet network model used in the embodiment on the high-pressure steam leakage of the pipeline is far better than that of other network models, thereby verifying the superiority of the embodiment on the design of the deep learning network model.
Meanwhile, the embodiment also performs a comparison experiment on the design of the Loss function and the evaluation standard, fig. 9 shows the effect of using the cross entropy Loss function, the Dice local function and the combination of the Dice local function and the Focal local function used in the embodiment respectively in the training process of the CBAM-Res _ Unet network model on image segmentation after the network model is established, and as can be seen from fig. 9, the combination of the Dice local function and the Focal local function used in the embodiment has the best effect on image segmentation of pipeline high-pressure steam leakage, and the superiority of the embodiment in the design of the Loss function on the training network model is verified. Fig. 10 shows the effect of image segmentation after model building under different evaluation criteria, and as can be seen from fig. 10, the evaluation criterion F1_ score used in this embodiment has better effect. In addition, different depth learning network models are compared under the evaluation standard, and as a result, as shown in fig. 11, the evaluation standard value of the CBAM-Res _ Unet network model used in the embodiment on the image segmentation of the high-pressure steam leakage of the pipeline is far higher than that of other network models, thereby further verifying the superiority of the embodiment.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (5)
1. A CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method is characterized by comprising the following steps:
acquiring a high-pressure steam leakage picture to be tested, performing image segmentation on the high-pressure steam leakage picture by using a CBAM-Res _ Unet network model, and outputting a detection result;
the left side of the CBAM-Res _ Unet network model is provided with a down-sampling layer, each layer passes through two convolution layers of n x n and a maximum pooling layer of m x m, an activation function is a linear rectification function, the right side of the CBAM-Res _ Unet network model is provided with an up-sampling layer, each layer passes through two convolution layers of n x n and a convolution layer of p x p, the activation function is an S-shaped growth curve function, a residual error network module is added to each side of the activation function, and an attention mechanism module is added to the residual error network module; meanwhile, a Loss function and an evaluation standard are added into the CBAM-Res _ Unet network model, the Loss function is the combination of a Dice Loss function and a Focal Loss function, and the evaluation standard is F1_ score.
2. The CBAM-Res _ Unet-based power plant pipeline high pressure steam leakage detection method according to claim 1, wherein the loss function L has an expression as follows:
where C denotes the number of samples input, i denotes the number of classes including regions, i.e., a total of both the steam leakage region and the steam non-leakage region, and TPp(i) Indicating true positives for i, FNp(i) Indicates false negatives of i, FPp(i) False positives for i; p is a radical ofn(i) Representing a predicted value for i; gn(i) Represents the true value of i; λ represents a weight; α and β are penalty values for false negatives and false positives, respectively, both set to 0.5; n represents all pixel values of the image.
3. The CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method according to claim 1, wherein a channel attention submodule in an attention mechanism module in the CBAM-Res _ Unet network model adopts global average pooling and maximum pooling for an input feature map, and performs addition operation through an MLP multi-layer sensor, and generates an output feature map of the channel attention submodule by using an S-type growth curve activation function.
4. The CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method according to claim 3, wherein a spatial attention submodule of an attention mechanism module in the CBAM-Res _ Unet network model takes a multiplication result of an input feature map and an output feature map of a channel attention module based on elements as an input feature map of the spatial attention module, changes the number of channels into 1 through convolution operation after global average pooling and maximum pooling, generates an output feature map of the spatial attention submodule by using an S-type growth curve activation function, and the output feature map of the attention mechanism module is a multiplication result of the input feature map and the output feature map of the spatial attention module.
5. The CBAM-Res _ Unet-based power plant pipeline high pressure steam leakage detection method according to claim 1, wherein the CBAM-Res _ Unet network model training method comprises the steps of:
s1, extracting a picture from every other multiframe of the video of the high-pressure steam leakage of the power plant pipeline to form a picture set,
s2, randomly extracting a set number of pictures from the picture set as initial sample pictures;
s3, marking the initial sample graph into a high-pressure steam leakage area and a high-pressure steam non-leakage area to obtain a labeled sample graph;
s4, preprocessing the initial sample graph and the labeled sample graph, and adjusting the sizes of the initial sample graph and the labeled sample graph to set values;
s5, increasing the number of multiple samples of the processed initial sample image and the labeled sample image by a data enhancement method;
and S6, taking the initial sample graph and the labeled sample graph after data enhancement as training data sets, and training the CBAM-Res _ Unet network model after adding a loss function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110914532.6A CN113763327B (en) | 2021-08-10 | 2021-08-10 | Power plant pipeline high-pressure steam leakage detection method based on CBAM-Res_Unet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110914532.6A CN113763327B (en) | 2021-08-10 | 2021-08-10 | Power plant pipeline high-pressure steam leakage detection method based on CBAM-Res_Unet |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113763327A true CN113763327A (en) | 2021-12-07 |
| CN113763327B CN113763327B (en) | 2023-11-24 |
Family
ID=78788912
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110914532.6A Active CN113763327B (en) | 2021-08-10 | 2021-08-10 | Power plant pipeline high-pressure steam leakage detection method based on CBAM-Res_Unet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113763327B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114187293A (en) * | 2022-02-15 | 2022-03-15 | 四川大学 | Oral cavity palate part soft and hard tissue segmentation method based on attention mechanism and integrated registration |
| CN115170579A (en) * | 2022-09-09 | 2022-10-11 | 之江实验室 | Metal corrosion image segmentation method and device |
| CN116206221A (en) * | 2023-01-16 | 2023-06-02 | 北京师范大学 | Water flare detection method and system |
| CN116576405A (en) * | 2023-07-12 | 2023-08-11 | 上海电机学院 | Air duct leakage signal detection method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA3010163A1 (en) * | 2017-07-01 | 2019-01-01 | Algolux Inc. | Method and apparatus for joint image processing and perception |
| CN110120055A (en) * | 2019-04-12 | 2019-08-13 | 浙江大学 | Fundus fluorescein angiography image based on deep learning is without perfusion area automatic division method |
| CN111709886A (en) * | 2020-05-27 | 2020-09-25 | 杭州电子科技大学 | Image highlight removing method based on U-shaped cavity residual error network |
| CN113012172A (en) * | 2021-04-09 | 2021-06-22 | 杭州师范大学 | AS-UNet-based medical image segmentation method and system |
| WO2021134871A1 (en) * | 2019-12-30 | 2021-07-08 | 深圳市爱协生科技有限公司 | Forensics method for synthesized face image based on local binary pattern and deep learning |
| WO2021139069A1 (en) * | 2020-01-09 | 2021-07-15 | 南京信息工程大学 | General target detection method for adaptive attention guidance mechanism |
-
2021
- 2021-08-10 CN CN202110914532.6A patent/CN113763327B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA3010163A1 (en) * | 2017-07-01 | 2019-01-01 | Algolux Inc. | Method and apparatus for joint image processing and perception |
| CN110120055A (en) * | 2019-04-12 | 2019-08-13 | 浙江大学 | Fundus fluorescein angiography image based on deep learning is without perfusion area automatic division method |
| WO2021134871A1 (en) * | 2019-12-30 | 2021-07-08 | 深圳市爱协生科技有限公司 | Forensics method for synthesized face image based on local binary pattern and deep learning |
| WO2021139069A1 (en) * | 2020-01-09 | 2021-07-15 | 南京信息工程大学 | General target detection method for adaptive attention guidance mechanism |
| CN111709886A (en) * | 2020-05-27 | 2020-09-25 | 杭州电子科技大学 | Image highlight removing method based on U-shaped cavity residual error network |
| CN113012172A (en) * | 2021-04-09 | 2021-06-22 | 杭州师范大学 | AS-UNet-based medical image segmentation method and system |
Non-Patent Citations (2)
| Title |
|---|
| 谭雨婷;李家琨;金伟其;王霞: "气体泄漏的单点探测器与红外成像检测的灵敏度模拟分析", 红外与激光工程, vol. 43, no. 8 * |
| 金耀;徐阳;韩飞杨;何少阳;王俊博;: "基于深度学习语义分割的桥梁病害图像像素级识别方法", 公路交通科技(应用技术版), no. 01 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114187293A (en) * | 2022-02-15 | 2022-03-15 | 四川大学 | Oral cavity palate part soft and hard tissue segmentation method based on attention mechanism and integrated registration |
| CN115170579A (en) * | 2022-09-09 | 2022-10-11 | 之江实验室 | Metal corrosion image segmentation method and device |
| CN116206221A (en) * | 2023-01-16 | 2023-06-02 | 北京师范大学 | Water flare detection method and system |
| CN116206221B (en) * | 2023-01-16 | 2023-09-05 | 北京师范大学 | Water flare detection method and system |
| CN116576405A (en) * | 2023-07-12 | 2023-08-11 | 上海电机学院 | Air duct leakage signal detection method and system |
| CN116576405B (en) * | 2023-07-12 | 2023-10-31 | 上海电机学院 | Air duct leakage signal detection method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113763327B (en) | 2023-11-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113763327A (en) | CBAM-Res _ Unet-based power plant pipeline high-pressure steam leakage detection method | |
| CN111639692B (en) | Shadow detection method based on attention mechanism | |
| CN111768388B (en) | Product surface defect detection method and system based on positive sample reference | |
| CN112861729B (en) | Real-time depth completion method based on pseudo-depth map guidance | |
| CN111524135A (en) | Image enhancement-based method and system for detecting defects of small hardware fittings of power transmission line | |
| WO2021088101A1 (en) | Insulator segmentation method based on improved conditional generative adversarial network | |
| CN115131797B (en) | Scene text detection method based on feature enhancement pyramid network | |
| CN110020658B (en) | Salient object detection method based on multitask deep learning | |
| CN110751649A (en) | Video quality evaluation method and device, electronic equipment and storage medium | |
| CN113066089B (en) | Real-time image semantic segmentation method based on attention guide mechanism | |
| CN111798469A (en) | Digital image small data set semantic segmentation method based on deep convolutional neural network | |
| CN113554032B (en) | Remote sensing image segmentation method based on multi-path parallel network of high perception | |
| CN115223009A (en) | Small target detection method and device based on improved YOLOv5 | |
| CN114639102B (en) | Cell segmentation method and device based on key point and size regression | |
| CN115375999A (en) | Target detection model, method and device applied to dangerous chemical vehicle detection | |
| CN115063833A (en) | Machine room personnel detection method based on image layered vision | |
| CN113052103A (en) | Electrical equipment defect detection method and device based on neural network | |
| CN112418229A (en) | Unmanned ship marine scene image real-time segmentation method based on deep learning | |
| CN117079260A (en) | Text detection method based on mixed attention and feature enhancement | |
| CN115587967B (en) | Fundus image optic disk detection method based on HA-UNet network | |
| CN113723414B (en) | Method and device for dividing mask face shielding object | |
| CN114897765A (en) | Bolt abnormity detection method in complex scene based on depth feature reconstruction | |
| CN113888424A (en) | Historical relic photo color restoration method and device, electronic equipment and storage medium | |
| CN113298097A (en) | Feature point extraction method and device based on convolutional neural network and storage medium | |
| CN113222012A (en) | Automatic quantitative analysis method and system for lung digital pathological image |
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 |