CN117197415A - Method, device and storage medium for detecting target in inspection area of natural gas long-distance pipeline - Google Patents
Method, device and storage medium for detecting target in inspection area of natural gas long-distance pipeline Download PDFInfo
- Publication number
- CN117197415A CN117197415A CN202311474290.9A CN202311474290A CN117197415A CN 117197415 A CN117197415 A CN 117197415A CN 202311474290 A CN202311474290 A CN 202311474290A CN 117197415 A CN117197415 A CN 117197415A
- Authority
- CN
- China
- Prior art keywords
- model
- distance pipeline
- training
- long
- input image
- 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
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 37
- 239000003345 natural gas Substances 0.000 title claims abstract description 26
- 238000007689 inspection Methods 0.000 title claims abstract description 25
- 238000003860 storage Methods 0.000 title claims abstract description 8
- 238000001514 detection method Methods 0.000 claims abstract description 62
- 238000012549 training Methods 0.000 claims abstract description 45
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 230000004927 fusion Effects 0.000 claims description 25
- 238000012360 testing method Methods 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 2
- 238000005070 sampling Methods 0.000 description 9
- 230000006872 improvement Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000002457 bidirectional effect Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 239000007789 gas Substances 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 238000013523 data management Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The application discloses a method, a device and a storage medium for detecting targets in a natural gas long-distance pipeline inspection area, wherein the method comprises the following steps: s100: acquiring an input image of an area where a natural gas long-distance pipeline is located; s200: preprocessing the acquired image; s300: constructing a long-distance pipeline detection model and training; s400: and inputting the preprocessed long-distance pipeline image into a trained long-distance pipeline detection model to realize the detection of the natural gas long-distance pipeline. Compared with the traditional method, the method has higher efficiency, lower cost and better safety in the inspection area detection of the long-distance pipeline, and improves the overall inspection effect.
Description
Technical Field
The application belongs to the field of target detection, and particularly relates to a method and a device for detecting targets in a natural gas long-distance pipeline inspection area and a storage medium.
Background
With the acceleration of the industrialization process, long-distance pipelines have become a central point of modern economy as an important facility for energy transportation. The safety of natural gas long-distance pipelines relates to downstream users and upstream gas storages, pipelines, gas transmission stations and gas fields, and is a complex system engineering. Safety accidents occur at any position of the natural gas long-distance pipeline, so that the whole pipeline system cannot normally operate, and production and life of downstream users are seriously affected. The pipeline inspection is a basic work for effectively ensuring the safe and stable operation of the natural gas conveying pipeline and equipment thereof. The running condition of the pipeline and the change of the surrounding environment are mastered through inspection, the hidden danger of equipment defects and endangering the pipeline is discovered, the hidden danger is eliminated in time, the accident is prevented, or the faults are limited to the minimum range, and the safety and the stability of the conveying pipeline are ensured. The traditional pipeline inspection mode mainly comprises manual inspection and vehicle-mounted inspection, but hidden hazards shown by the two inspection modes are gradually prominent, and the pipeline inspection mode mainly comprises obvious defects of more human factors, high management cost, incapability of supervising the working state of inspection staff and the like, and meanwhile cannot adapt to the development requirement of management informatization of a conveying pipeline.
Disclosure of Invention
Aiming at the defects in the prior art, the application aims to provide a target detection method for a long-distance pipeline inspection area,
in order to achieve the above purpose, the present application provides the following technical solutions:
a target detection method for a natural gas long-distance pipeline inspection area comprises the following steps:
s100: acquiring an input image of an area where a natural gas long-distance pipeline is located;
s200: preprocessing the acquired input image;
s300: constructing a long-distance pipeline detection model and training;
s400: and inputting the preprocessed input image into a trained long-distance pipeline detection model to identify whether a natural gas long-distance pipeline exists in the input image.
Preferably, in step S200, the preprocessing the acquired input image includes the following steps: and carrying out convolution operation on the input image based on the Gaussian kernel function to obtain a binarized real density map.
Preferably, in step S300, the long-distance pipeline detection model is trained by the following steps:
s301: constructing a data set, and dividing the data set into a training set and a verification set;
s302: initializing the weight of the model, training the model through a training set, calculating a loss function loss in the training process, carrying out back propagation to optimize the network weight, and completing the model training when the loss function loss is converged;
s303: testing the trained model through a test set, evaluating the model through average detection precision in the test process, and passing the model test when the average detection precision reaches 0.95; otherwise, the training parameters are adjusted to train the model again.
The application also provides a target detection device for the natural gas long-distance pipeline inspection area, which comprises the following components:
the acquisition module is used for acquiring an input image;
the preprocessing module is used for preprocessing the acquired input image;
the model construction and training module is used for constructing a long-distance pipeline detection model and training; the model comprises a backbone network, and depth separable convolution DWC3 is introduced into the backbone network, so that the model can ensure the accuracy and simultaneously improve the feature extraction speed; the model also comprises a feature fusion network, and the feature fusion network enables the model to automatically learn and pay attention to the most important features in the image by introducing a attention module CBAM of a convolution layer;
the detection module is used for inputting the preprocessed input image into a trained long-distance pipeline detection model so as to identify whether the input image contains a natural gas long-distance pipeline or not.
Preferably, the model building and training module includes:
the data set constructing sub-module is used for constructing a data set and dividing the data set into a training set and a verification set;
the training sub-module is used for training the model through the training set;
and the test sub-module is used for testing the trained model through the test set.
The present application also provides an electronic device including:
a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein,
the processor, when executing the program, implements a method as described in any of the preceding.
The application also provides a computer storage medium storing computer executable instructions for use in a method as described in any one of the preceding claims.
Compared with the prior art, the application has the beneficial effects that:
1. the detection efficiency is improved: the method can automatically and rapidly detect the target object by utilizing computer vision and a machine learning algorithm, and greatly improves the inspection efficiency.
2. The labor cost is reduced: the traditional inspection method requires a large amount of manpower input, and the method can reduce the manpower demand and the manpower cost.
3. The safety is improved: the long-distance pipeline inspection area generally has a certain safety risk, and the method can reduce the requirement of personnel entering the dangerous area and improve the inspection safety.
4. Accuracy is improved: the method utilizes advanced algorithm and technology, can improve the detection accuracy and reduce the conditions of missed detection and false detection.
5. And (3) data management: the method can carry out data management on the detection result, is convenient for subsequent analysis and processing, and provides better decision basis.
Drawings
FIG. 1 is a flow chart of a method for detecting targets in a long-distance pipeline inspection area according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a long-distance pipeline detection model according to another embodiment of the present application.
Detailed Description
Specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. While specific embodiments of the application are shown in the drawings, it should be understood that the application may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.
It should be noted that certain terms are used throughout the description and claims to refer to particular components. Those of skill in the art will understand that a person may refer to the same component by different names. The specification and claims do not identify differences in terms of components, but rather differences in terms of the functionality of the components. As used throughout the specification and claims, the terms "include" and "comprise" are used in an open-ended fashion, and thus should be interpreted to mean "include, but not limited to. The description hereinafter sets forth a preferred embodiment for practicing the application, but is not intended to limit the scope of the application, as the description proceeds with reference to the general principles of the description. The scope of the application is defined by the appended claims.
For the purpose of facilitating an understanding of the embodiments of the present application, reference will now be made to the drawings, by way of example, and specific examples of which are illustrated in the accompanying drawings.
In one embodiment, as shown in fig. 1, the application provides a method for detecting targets in a natural gas long-distance pipeline inspection area, which comprises the following steps:
s100: acquiring an input image of an area where a natural gas long-distance pipeline is located;
s200: preprocessing the acquired image;
s300: constructing a long-distance pipeline detection model and training;
s400: and inputting the preprocessed long-distance pipeline image into a trained long-distance pipeline detection model to realize the detection of the natural gas long-distance pipeline.
In another embodiment, in step S200, the preprocessing the acquired image includes the following steps: and carrying out convolution operation on the image based on the Gaussian kernel function to obtain a binarized real density map.
In another embodiment, in step S300, as shown in fig. 2, the long-distance pipeline detection model includes a backbone network, a feature fusion network, and an output detection head.
In this embodiment, the backbone network includes the following layers connected in sequence:
focus layers (3, 64) (wide, high);
a CONV layer (64, 128);
DWC3 layer (depth separable convolution dwconv3×3) (128 );
a CONV layer (128, 256);
DWC3 layers (256 );
a CONV layer (256, 512);
SPPF layer (spatial pyramid pooling layer) (512 );
DWC3 layers (512 );
a CONV layer (512, 1024);
SPPF layers (1024 );
DWC3 layers (1024 ).
The feature fusion network comprises three branches from bottom to top, wherein,
the first branch includes:
CBAM layer (1024 )
DWC3 layers (1024, 256);
the second branch comprises:
a CONCAT layer (1024);
DWC3 layers (1024, 256);
a CBAM layer (512, 256);
the third branch includes:
a CONCAT layer (512);
DWC3 layers (512, 256);
CBAM layer (256 ).
In addition, a CONV layer (1024, 512) and a lightweight general up-sampling operator (DUsample) (512) are sequentially arranged between the CBAM layer of the first branch and the CONCAT layer of the second branch, and the DWC3 layer of the first branch is connected with the CBAM layer of the second branch (by adding the CBAM layer after the DWC3 layer, the expression capability and the perception capability of the model can be improved, and important features of the model can be focused better); a DWC3 layer (1024, 512), a CONV layer (512, 256) and a lightweight general up-sampling operator (sampling) (256) are sequentially arranged between the CONCAT layer of the second branch and the CONCAT layer of the third branch.
It should be noted that the CBAM layer (Convolutional Block Attention Module, the attention module of the convolution layer) is an attention mechanism, and mainly includes two attention sub-modules: a channel attention module (Channel Attention Module) and a spatial attention module (Spatial Attention Module), wherein,
the channel attention module selectively enhances or suppresses the feature representation of different channels in an adaptive manner by learning the importance weights of each channel. This helps the model to better focus on important feature channels, reduces redundant information, and improves the model's ability to distinguish between objects.
The spatial attention module then selectively enhances or suppresses the feature representation of the different locations in an adaptive manner by learning the importance weights for each spatial location. This helps the model to better focus on important spatial locations, improving the model's ability to locate and perceive regions of the target.
By introducing a CBAM attention mechanism, the model can automatically learn and pay attention to the most important features in the image, and the expression capability and the perception capability of the model are improved, so that better performance is achieved in various computer vision tasks.
Furthermore, the long-distance pipeline has smaller size and low resolution in the acquired input image, so that the detection of the long-distance pipeline belongs to small target detection, and the characteristic information of the long-distance pipeline is not obvious enough in the acquired input image, but the information is further lost due to downsampling and other operations adopted in the characteristic processing process in the conventional target detection method based on the deep convolutional neural network, so that the characteristic information for subsequent prediction is reduced, and the network detection precision is greatly limited. The existing target detection network adopts nearest neighbor up-sampling, but the characteristic information loss is relatively serious in the small target detection process, and the key information of the long-distance pipeline cannot be extracted better. Aiming at the problem, the embodiment introduces a finer lightweight general upsampling operator DUsample in the feature fusion network, the upsampling operator has a larger receptive field, the information around the long-distance pipeline can be better utilized, and the upsampling core of the upsampling operator is related to the semantic information of the feature map. Meanwhile, as the reconstruction capability of the sampling operator is greatly improved, and the decoder based on the operator can flexibly utilize the combination of different layer characteristics of any CNN decoder, the model can recover the size of the feature map and improve the detection precision, and the calculation complexity is reduced. Therefore, the sampling operator is introduced in the characteristic processing process, so that the loss of the characteristic information such as texture, color and the like of the long-distance pipeline extracted by the main network can be reduced.
In addition, as an improvement to the existing feature fusion network, the feature fusion network is designed into a bidirectional weighted feature structure, namely, on the basis of the structure that three branches of the feature fusion network are vertically connected from bottom to top, a CBAM layer in a first branch of the feature fusion network is transversely connected with a DWC3 layer (1024) in a backbone network; transversely connecting the CONCAT layer in the second branch of the feature fusion network with the DWC3 layer (512 ) in the backbone network; the CONCAT layer in the first branch of the feature fusion network is transversely connected with the DWC3 layer (256) in the backbone network, so that a bidirectional feature fusion structure for vertical feature fusion and transverse feature fusion is formed, the structure can fuse feature images from different scales, retain detailed information of the feature images to the greatest extent, and increase the receptive field of the network, so that targets with different sizes and different positions can be detected, the network can better understand the whole image, more useful features are extracted, and finally the whole network can improve the robustness and the detection precision of target detection.
The specific steps of the processing of the model on the input image are as follows: first, for an input feature mapWhich belongs to the network structureThen to other layersIs characterized by (a)Up-sampling and down-sampling operations are performed and adjusted to be equal toThe same scale. Then, spatial weights of the feature map are adaptively learned from each input layer while synchronizing the respective input layer feature map dimensions. Is provided withIs to adjust from level 0 to level l feature mappingThe feature vector of the position corresponds toThe feature fusion results of the layers are as follows:
（1）
wherein,the output characteristic diagram is shown in the representation,、andrespectively, represent the parameters that can be learned,、andrespectively represent slaveThe weights learned in the layer profile.
The problem that the image features are difficult to effectively fuse under the same scale can be solved by adaptively learning the spatial weight of the feature map from each input layer.
After the fused long-distance pipeline characteristics are obtained, the target size of the long-distance pipeline is estimated through the depth information of the image, and the size range of the target is determined according to the size range, which scale region in the characteristic extraction module the target is positioned in, so that anchor point information of a detection frame is conveniently set in a corresponding decoding layer, and the size of the anchor point information is initialized. The anchor points of a typical detector are all located at the higher decoding layer, whereas for smaller sizes the anchor points should be located at the lower decoding layer.
The generated anchor point information is shown as formulas (2) and (3), wherein the formula (2) is used for calculating IoU values between the target frame and the prediction frame; equation (3) is used to calculate the value of the adjustment coefficient of the prediction frame.
（2）
（3）
In the formula (2), m represents an index of a prediction frame, n represents an index of a target frame, γ is a constant, and d (m, n) represents a distance (typically, euclidean distance or distance between center points) between the prediction frame m and the target frame n. The meaning of the formula (2) is that the larger IoU between the predicted frame and the target frame is, the smaller the value of H (m, n) is, which means that the matching degree of the two is higher.
In the formula (3), σ represents an adjustment coefficient, β is a constant, r represents the size of the prediction frame, rf represents the size of the anchor, and d represents the distance between the prediction frame and the target frame. From these two formulas, the dimensions and aspect ratio of the anchor frame can be calculated from the distances of the different decoding layers and feature extractors. In this way, anchor frames adapting to different target scales can be generated on the feature maps with different scales so as to perform target detection and positioning. By adjusting the two parameters of gamma and beta, the variation range of the dimension and the length-width ratio of the anchor frame can be controlled.
For the nearest neighbor method adopted by the position with the invalid depth value to generate a detection boundary frame, the proposed method can be easily expanded to the long-distance pipeline target detection of the RGB image, and the long-distance pipeline input target detection network predicted by the depth separation cavity convolutional neural network model is directly used for classifying pipelines and non-pipelines because of no depth information, so that the detection of the long-distance pipeline is completed.
The improved feature fusion network provided by the embodiment can enhance the feature extraction capability in the model and effectively reduce the calculation amount, wherein the up-down sampling directly uses bilinear interpolation instead of global pooling operation, and the implementation method comprises the following steps:
1. the new design of the bidirectional weighting feature structure is also adopted in the back part by using CSPRepResNet as a back, and both the back and the back are based on CSPRepResstage provided by the embodiment. The novel backspace and the novel enhance the model characterization capability and simultaneously promote the reasoning speed of the model, and the size of the model can be flexibly configured through a width multiplexer and a depth multiplexer.
2. In the embodiment, a DWC3 layer is introduced into a backbone network, so that the model improves the characteristic extraction speed while ensuring the constant precision.
Finally, in the embodiment, a learnable weight alpha is added on the CONV layer of the backbone network, so that the characterization capability of the backbone network is further improved, and the improvement of the feature extraction effect is obtained. To cope with NMS, the anchor point allocation of the training instance should satisfy the following rules: (1) The aligned anchor points should be able to predict high classification scores and perform accurate joint positioning; (2) The misaligned anchor points should have a low classification score and be subsequently suppressed. Based on these two objectives, the present embodiment designs a new anchor point alignment metric to explicitly measure the task alignment at the anchor point level. The alignment metric is integrated into the sample allocation and loss function to dynamically improve the prediction at each anchor point, as shown in equation (4):
（4）
wherein,andrepresenting the classification score and the IOU value, respectively. Alpha and beta are used to control the impact of these two tasks in the anchor point alignment metric. For each instance, the anchor point with the largest value is selected as the positive sample, and the remaining anchor points are selected as the negative samples. Likewise, training is performed by calculating new penalty functions specifically designed for tuning classification and positioning tasks.
The output detection Head employs a coupled output to output three different heads for multi-scale detection, wherein a first Head connects CBAM (256 ) layers in the feature fusion network, a second Head connects CBAM (512, 256) layers in the feature fusion network, and a third Head connects DWC3 (1024,256) layers in the feature fusion network. Three heads are used to detect targets of different scales, respectively, because: in the image, the sizes of the targets may have great difference, some targets may be small, some targets may be large, and in order to accurately detect targets with different scales, a plurality of heads need to be introduced to carry out differential detection on the targets with different scales, so that the accuracy and the diversity of the detection of the targets with different scales by the model can be improved, and the higher detection speed can be maintained.
In another embodiment, in step S300, the long-distance pipeline detection model is trained by:
s301: constructing a data set, and dividing the data set into a training set and a verification set;
s302: initializing the weight of the model, training the model through a training set, calculating a loss function loss in the training process, carrying out back propagation to optimize the network weight, and completing the model training when the loss function loss is converged;
s303: testing the trained model through a test set, evaluating the model through average detection precision (mAP) in the test process, and passing the model test when the average detection precision reaches 0.95; otherwise, the training parameters are adjusted to train the model again.
In another embodiment, the application further provides a device for detecting targets in a natural gas long-distance pipeline inspection area, which comprises:
the acquisition module is used for acquiring an input image of the region where the natural gas long-distance pipeline is located;
the preprocessing module is used for preprocessing the acquired input image;
the model construction and training module is used for constructing a long-distance pipeline detection model and training; the model comprises a backbone network, and depth separable convolution DWC3 is introduced into the backbone network, so that the model can ensure the accuracy and simultaneously improve the feature extraction speed; the model also comprises a feature fusion network, and the feature fusion network enables the model to automatically learn and pay attention to the most important features in the image by introducing a attention module CBAM of a convolution layer;
the detection module is used for inputting the preprocessed input image into a trained long-distance pipeline detection model so as to identify whether the input image contains a natural gas long-distance pipeline or not.
In another embodiment, the model building and training module comprises:
the data set constructing sub-module is used for constructing a data set and dividing the data set into a training set and a verification set;
the training sub-module is used for training the model through the training set;
and the test sub-module is used for testing the trained model through the test set.
In another embodiment, the present application further provides an electronic device, including:
a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein,
the processor, when executing the program, implements a method as described in any of the preceding.
In another embodiment, the present application also provides a computer storage medium storing computer-executable instructions for performing a method as described in any one of the preceding claims.
While the application has been described in detail in the foregoing general description and specific examples, it will be apparent to those skilled in the art that modifications and improvements can be made thereto. Accordingly, such modifications or improvements may be made without departing from the spirit of the application and are intended to be within the scope of the application as claimed.
Claims (7)
1. The method for detecting the target in the inspection area of the natural gas long-distance pipeline is characterized by comprising the following steps of:
s100: acquiring an input image of an area where a natural gas long-distance pipeline is located;
s200: preprocessing the acquired input image;
s300: constructing a long-distance pipeline detection model and training;
the model comprises a backbone network, and depth separable convolution DWC3 is introduced into the backbone network, so that the model can ensure the accuracy and simultaneously improve the feature extraction speed; the model also comprises a feature fusion network, and the feature fusion network enables the model to automatically learn and pay attention to the most important features in the image by introducing a attention module CBAM of a convolution layer;
s400: and inputting the preprocessed input image into a trained long-distance pipeline detection model to identify whether a natural gas long-distance pipeline exists in the input image.
2. The method according to claim 1, wherein in step S200, the preprocessing of the acquired input image comprises the steps of: and carrying out convolution operation on the input image based on the Gaussian kernel function to obtain a binarized real density map.
3. The method according to claim 1, wherein in step S300, the long-distance pipeline detection model is trained by:
s301: constructing a data set, and dividing the data set into a training set and a verification set;
s302: initializing the weight of the model, training the model through a training set, calculating a loss function loss in the training process, carrying out back propagation to optimize the network weight, and completing the model training when the loss function loss is converged;
s303: testing the trained model through a test set, evaluating the model through average detection precision in the test process, and passing the model test when the average detection precision reaches 0.95; otherwise, the training parameters are adjusted to train the model again.
4. The utility model provides a regional target detection device is patrolled and examined to natural gas long distance pipeline, its characterized in that, the device includes:
the acquisition module is used for acquiring an input image of the region where the natural gas long-distance pipeline is located;
the preprocessing module is used for preprocessing the acquired input image;
the model construction and training module is used for constructing a long-distance pipeline detection model and training; the model comprises a backbone network, and depth separable convolution DWC3 is introduced into the backbone network, so that the model can ensure the accuracy and simultaneously improve the feature extraction speed; the model also comprises a feature fusion network, and the feature fusion network enables the model to automatically learn and pay attention to the most important features in the image by introducing a attention module CBAM of a convolution layer;
the detection module is used for inputting the preprocessed input image into a trained long-distance pipeline detection model so as to identify whether the input image contains a natural gas long-distance pipeline or not.
5. The apparatus of claim 4, wherein the model building and training module comprises:
the data set constructing sub-module is used for constructing a data set and dividing the data set into a training set and a verification set;
the training sub-module is used for training the model through the training set;
and the test sub-module is used for testing the trained model through the test set.
6. An electronic device, comprising:
a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein,
the processor, when executing the program, implements the method of any one of claims 1 to 3.
7. A computer storage medium having stored thereon computer executable instructions for performing the method of any of claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311474290.9A CN117197415B (en) | 2023-11-08 | 2023-11-08 | Method, device and storage medium for detecting target in inspection area of natural gas long-distance pipeline |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311474290.9A CN117197415B (en) | 2023-11-08 | 2023-11-08 | Method, device and storage medium for detecting target in inspection area of natural gas long-distance pipeline |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117197415A true CN117197415A (en) | 2023-12-08 |
| CN117197415B CN117197415B (en) | 2024-01-30 |
Family
ID=88987298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311474290.9A Active CN117197415B (en) | 2023-11-08 | 2023-11-08 | Method, device and storage medium for detecting target in inspection area of natural gas long-distance pipeline |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117197415B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117451671A (en) * | 2023-12-22 | 2024-01-26 | 四川泓宝润业工程技术有限公司 | Method and device for inverting gas concentration, storage medium and electronic equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111461291A (en) * | 2020-03-13 | 2020-07-28 | 西安科技大学 | Long-distance pipeline inspection method based on YO L Ov3 pruning network and deep learning defogging model |
| WO2021016873A1 (en) * | 2019-07-30 | 2021-02-04 | 珠海全志科技股份有限公司 | Cascaded neural network-based attention detection method, computer device, and computer-readable storage medium |
| CN112766195A (en) * | 2021-01-26 | 2021-05-07 | 西南交通大学 | Electrified railway bow net arcing visual detection method |
| CN114120019A (en) * | 2021-11-08 | 2022-03-01 | 贵州大学 | Lightweight target detection method |
| CN115170550A (en) * | 2022-07-29 | 2022-10-11 | 江苏大学 | Deep learning-based battery defect detection method and system |
| CN116051977A (en) * | 2022-12-01 | 2023-05-02 | 长春工业大学 | Multi-branch fusion-based lightweight foggy weather street view semantic segmentation algorithm |
| CN116824542A (en) * | 2023-06-13 | 2023-09-29 | 重庆市荣冠科技有限公司 | Light-weight foggy-day vehicle detection method based on deep learning |
| CN116958907A (en) * | 2023-09-18 | 2023-10-27 | 四川泓宝润业工程技术有限公司 | Method and system for inspecting surrounding hidden danger targets of gas pipeline |
-
2023
- 2023-11-08 CN CN202311474290.9A patent/CN117197415B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021016873A1 (en) * | 2019-07-30 | 2021-02-04 | 珠海全志科技股份有限公司 | Cascaded neural network-based attention detection method, computer device, and computer-readable storage medium |
| CN111461291A (en) * | 2020-03-13 | 2020-07-28 | 西安科技大学 | Long-distance pipeline inspection method based on YO L Ov3 pruning network and deep learning defogging model |
| CN112766195A (en) * | 2021-01-26 | 2021-05-07 | 西南交通大学 | Electrified railway bow net arcing visual detection method |
| CN114120019A (en) * | 2021-11-08 | 2022-03-01 | 贵州大学 | Lightweight target detection method |
| CN115170550A (en) * | 2022-07-29 | 2022-10-11 | 江苏大学 | Deep learning-based battery defect detection method and system |
| CN116051977A (en) * | 2022-12-01 | 2023-05-02 | 长春工业大学 | Multi-branch fusion-based lightweight foggy weather street view semantic segmentation algorithm |
| CN116824542A (en) * | 2023-06-13 | 2023-09-29 | 重庆市荣冠科技有限公司 | Light-weight foggy-day vehicle detection method based on deep learning |
| CN116958907A (en) * | 2023-09-18 | 2023-10-27 | 四川泓宝润业工程技术有限公司 | Method and system for inspecting surrounding hidden danger targets of gas pipeline |
Non-Patent Citations (4)
| Title |
|---|
| GANG LI 等: "YOLO-RFF: An Industrial Defect Detection Method Based on Expanded Field of Feeling and Feature Fusion", 《FRONTIERS IN PLANT SCIENCE》, pages 1 - 16 * |
| 周登文 等: "基于区域互补注意力和多维注意力的轻量级图像超分辨率网络", 《模式识别与人工智能》, vol. 35, no. 7, pages 625 - 636 * |
| 孙洁娣 等: "改进稠密块轻量化神经网络的管道泄漏孔径识别", 《仪器仪表学报》, vol. 43, no. 3, pages 98 - 108 * |
| 张建恒: "融合注意力机制的轻量化道路交通标志检测方法", 《电子测量技术》, pages 1 - 9 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117451671A (en) * | 2023-12-22 | 2024-01-26 | 四川泓宝润业工程技术有限公司 | Method and device for inverting gas concentration, storage medium and electronic equipment |
| CN117451671B (en) * | 2023-12-22 | 2024-05-14 | 四川泓宝润业工程技术有限公司 | Method and device for inverting gas concentration, storage medium and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117197415B (en) | 2024-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111259930B (en) | General target detection method of self-adaptive attention guidance mechanism | |
| CN111680706B (en) | Dual-channel output contour detection method based on coding and decoding structure | |
| CN112766087A (en) | Optical remote sensing image ship detection method based on knowledge distillation | |
| CN117197415B (en) | Method, device and storage medium for detecting target in inspection area of natural gas long-distance pipeline | |
| CN106980858A (en) | The language text detection of a kind of language text detection with alignment system and the application system and localization method | |
| CN107844743A (en) | A kind of image multi-subtitle automatic generation method based on multiple dimensioned layering residual error network | |
| CN111914924A (en) | Rapid ship target detection method, storage medium and computing device | |
| CN106198551A (en) | The detection method of a kind of transmission line of electricity defect and device | |
| CN113487600B (en) | Feature enhancement scale self-adaptive perception ship detection method | |
| CN111476307A (en) | Lithium battery surface defect detection method based on depth field adaptation | |
| CN115496891A (en) | Wheat lodging degree grading method and device | |
| CN116958907A (en) | Method and system for inspecting surrounding hidden danger targets of gas pipeline | |
| Peng et al. | Research on oil leakage detection in power plant oil depot pipeline based on improved YOLO v5 | |
| CN116883393B (en) | Metal surface defect detection method based on anchor frame-free target detection algorithm | |
| CN114078209A (en) | Lightweight target detection method for improving small target detection precision | |
| CN116863271A (en) | Lightweight infrared flame detection method based on improved YOLO V5 | |
| CN117058476A (en) | Target detection method based on random uncertainty | |
| CN115239733A (en) | Crack detection method, crack detection device, terminal equipment and storage medium | |
| CN116385876A (en) | Optical remote sensing image ground object detection method based on YOLOX | |
| CN102938068B (en) | Bridge structure multi-system damage identification method | |
| CN115410102A (en) | SAR image airplane target detection method based on combined attention mechanism | |
| CN113780462A (en) | Vehicle detection network establishment method based on unmanned aerial vehicle aerial image and application thereof | |
| Yin et al. | Remote sensing object detection based on dilated convolution and multi-scale feature fusion | |
| CN113989179B (en) | Train wheel set tread defect detection method and system based on target detection algorithm | |
| Dang et al. | Ground target detection method on moving base based on improved YoloV5 |
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 |