CN113392705A - Method for identifying pipeline leakage target in desert area based on convolutional neural network - Google Patents

Method for identifying pipeline leakage target in desert area based on convolutional neural network Download PDF

Info

Publication number
CN113392705A
CN113392705A CN202110522573.0A CN202110522573A CN113392705A CN 113392705 A CN113392705 A CN 113392705A CN 202110522573 A CN202110522573 A CN 202110522573A CN 113392705 A CN113392705 A CN 113392705A
Authority
CN
China
Prior art keywords
data
neural network
convolutional neural
pipeline
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.)
Pending
Application number
CN202110522573.0A
Other languages
Chinese (zh)
Inventor
赵青
兰馨雨
刘爽
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN202110522573.0A priority Critical patent/CN113392705A/en
Publication of CN113392705A publication Critical patent/CN113392705A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • G06N3/047—Probabilistic or stochastic 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/14—Pipes

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Probability & Statistics with Applications (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a desert area pipeline leakage target identification method based on a convolutional neural network. The invention provides an automatic target identification method based on a convolutional neural network, which can automatically extract the characteristics of pipelines and leakage from input data, only needs to manually input part of labels during training and does not need excessive intervention. The trained network can be used for efficiently identifying the underground image of the pipeline in the desert area in real time, and can be widely applied to leakage identification of the underground pipeline in the desert area.

Description

Method for identifying pipeline leakage target in desert area based on convolutional neural network
Technical Field
The invention relates to the technical field of target identification, in particular to a method for identifying a pipeline leakage target in a desert area based on a convolutional neural network.
Background
The pipelines of the Xinjiang Tahe are numerous, the pipelines with complex pipeline conditions are positioned in sensitive areas such as populus diversifolia, salix integra and farmlands, and once the pipelines leak, the risk of environmental pollution is high.
The ground penetrating radar is an effective means which can accurately detect the ground bottom target without damage at present and has higher practicability. The main work of the device is completed by matching a transmitting antenna and a receiving antenna, firstly, the transmitting antenna is utilized to transmit high-frequency radar pulse waves penetrating through the stratum downwards, then, the receiving antenna is utilized to receive reflected waves caused by the change of the dielectric constant of an underground object to detect an underground target, and the change of the characteristics of the underground structure and related materials is detected on the basis of the reflected waves.
Because the underground situation is relatively complex in actual detection, radar waves are particularly easily interfered by various clutter when being propagated in the underground situation, and the signal-to-noise ratio of a B-scan data image is reduced. Therefore, the premise of accurately representing the underground target is to be capable of distinguishing clutter and noise and extracting useful information really representing the target object from radar data. At present, a plurality of methods for processing ground penetrating radar data exist, and the average value of the measurement of the adjacent position can be taken for a plurality of times on the detection means to reduce the non-target echo; in the data processing method, low-frequency, high-pass, band-pass and other frequency domain filtering is commonly used to eliminate non-target frequency bands, remove background through filtering, and perform multiple superposition, offset, deconvolution, complex signal processing and the like. For the interpretation of ground penetrating radar data, firstly, the data is preprocessed according to the characteristics of the radar working environment, the processed related data is analyzed through target characteristics, and manual interpretation is carried out from the data.
Therefore, a method for analyzing the target of the underground pipeline in the desert area in real time and accurately and widely applied to the leakage identification of the underground pipeline in the desert area is needed.
Disclosure of Invention
In order to solve the problems in the prior art, the present invention provides a solution to the problems mentioned in the background above.
In order to achieve the purpose, the invention provides the following technical scheme: a desert area pipeline leakage target identification method based on a convolutional neural network comprises the following steps:
s1, collecting target data by using a ground penetrating radar;
s2, preprocessing the data;
s3, constructing a convolutional neural network;
and S4, training and testing the convolutional neural network.
Preferably, the collecting target data by using a ground penetrating radar in step S1 specifically includes:
s11, defining a measuring area: dividing the detection area according to the map and the pipeline walking diagram, and respectively selecting a pipeline direction detection line and a vertical pipeline direction measuring line for detection and collection;
s12, detecting the measuring line: detecting and collecting according to the surveying and designed line, adopting a continuous detection mode to move forward stably along a certain direction of a measuring line, reading the abnormal section of the radar in real time in the detection process, recording the horizontal coordinate of the abnormal part, taking the horizontal coordinate as the horizontal coordinate of a target object, marking and recording the horizontal coordinate in a class newspaper;
s13, simulating leakage detection: for the simulation of the leakage pipeline, a pit is dug above the pipeline, water is poured into the pit to enable the pit to naturally permeate, the area is buried after the water is completely permeated, and then the area is detected and collected according to the detection step in the step S12.
Preferably, the preprocessing the data in step S2 specifically includes:
s21, performing transverse data interpolation and time sub-sampling on the data: the data in the time direction is time sub-sampled and the lateral data in the lateral group is interpolated to place the amount of data in both dimensions in one order of magnitude.
S22, performing sliding filtering on the data to remove the background;
s23, labeling the whole data with a leakage label or a non-leakage label according to the abscissa position of the target object marked in the step S12, performing data segmentation on the data, and dividing the whole data into a plurality of modules according to the equal abscissa length;
and S24, corresponding the segmented data to the label as the input of the convolutional neural network.
Preferably, the time sub-sampling mathematical expression in step S21 is as follows:
A′i[m，n]＝Ai[m，tsamplen]；
the mathematical expression for the horizontal axis data interpolation is:
Figure BDA0003064615710000031
preferably, the label of the data in step S23 is determined by whether the target horizontal position, i.e. the leaking position or the non-leaking position of the pipeline is located at the horizontal position of the transceiving antenna, and the mathematical expression is as follows:
Figure BDA0003064615710000032
preferably, the convolutional neural network in step S3 includes 4 convolutional layers, 2 pooling layers, 3 Dropout layers, 1 Flatten layer and 3 fully-connected layers; the Flatten layer can level the data set into one bit data input into the fully connected layer.
Preferably, the training step in step S4 specifically includes:
s41, setting training step number n, learning rate lr and training sample size batch _ size;
s42, inputting the data and the labels into a convolutional neural network for training;
s43, adjusting the training parameters to enable the loss rate of the cost function to be close to 0 and the accuracy rate to be close to 1;
s44, storing convolutional layer parameters, obtaining a convolutional neural network weight file suitable for desert pipeline discrimination after training is completed, and storing the convolutional neural network weight file;
the testing steps are specifically as follows: and inputting the test data into a convolutional neural network model obtained after training is finished, and reading the stored weight file through a network to obtain a prediction result.
The invention has the beneficial effects that: the method for identifying the pipeline leakage target in the desert area is an automatic identification method based on the convolutional neural network, can automatically extract the characteristics of the pipeline and the leakage from input data, only needs to manually input part of labels in training, and does not need excessive intervention. The trained network can be used for efficiently identifying the underground image of the pipeline in the desert area in real time, and can be widely applied to leakage identification of the underground pipeline in the desert area.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a schematic diagram of a neural network according to the present invention;
FIG. 3 is a schematic diagram of probe data according to an embodiment of the present invention;
fig. 4 is a schematic diagram illustrating a background removing effect according to an embodiment of the present invention, where fig. 4(a) is an original image, and fig. 4(b) is an image after background removing processing;
FIG. 5 is a schematic diagram of data slicing according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a target recognition probability curve according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-6, the present invention provides a technical solution: a specific flow of a method for identifying a pipeline leakage target in a desert area based on a convolutional neural network is shown in figure 1, in the embodiment, a pulse EKKO PRO geological radar is adopted as an instrument, the antenna frequency is 250mHz, and an experimental area is a pipeline area of a Xinjiang Tahe. A total of 10km of data was acquired via probing, which was marked as shown in fig. 3, where the marked areas are between the targets of the leak.
The performance of the convolutional neural network in computer vision almost laminates all other model algorithms, and the behavior of the convolutional neural network in the field of computer vision is dominated. Compared with a fully-connected network, the method can greatly improve the operation speed, and the applicability of the method to two-dimensional images also makes the method become the most popular neural network structure at present. The B-scan data of the ground penetrating radar is formed by splicing a plurality of rows of A-scan data formed by real numbers, so that the B-scan data can be approximately regarded as an image, the B-scan data can be subjected to feature extraction by means of a convolutional neural network, the radar data is identified by the neural network, feature identification is easier, and the identification accuracy can be improved.
The method comprises three parts of target data acquisition and data preprocessing, convolutional neural network building learning and network prediction. The target data acquisition part is used for acquiring real data of the desert area and providing data support for the convolutional neural network. The network building learning part needs to modify and build the convolutional neural network according to the format and the data characteristics of the data and input the data into the network. The network prediction part predicts the image of the target un-calibrated point by using the training result and obtains a prediction result on the basis of the previous step, which is as follows:
step 1: data acquisition for desert area pipeline by using ground penetrating radar
(1) Delimiting a measuring area
And dividing the detection area according to the map and the pipeline walking diagram, and respectively selecting the detection along the pipeline direction and the measuring line perpendicular to the pipeline direction for detection and collection.
(2) Detecting the measuring line
And detecting and collecting according to the line subjected to the reconnaissance and design, stably moving forward along a certain direction of the measuring line in a continuous detection mode, reading the abnormal section of the radar in real time in the detection process, recording the abscissa of the abnormal part, using the abscissa as the abscissa of the target object, marking and recording in a class newspaper.
(3) Analog leak detection
For the simulation of the leakage pipeline, a pit of 50cm × 50cm × 50cm is dug above the pipeline, water is poured to enable the pit to naturally penetrate, the area is buried after the water completely penetrates, and then the area is detected and collected according to the detection steps in the step 1 and the step 2.
Step 2: data preprocessing of training and testing samples
(1) The data is first interpolated laterally and sub-sampled in time. Because the magnitude of data obtained by simulation is often far higher than that of a horizontal sampling group in the time direction, if data such as slice labels are directly carried out, the difference of horizontal and vertical data amounts of the data groups is too large, and a target characteristic is difficult to distinguish by a convolution kernel, sampling is carried out in the time direction under the condition that the morphological characteristics of an image are not influenced, interpolation is carried out in the horizontal group, and the data amounts of two dimensions of the data are approximately in one magnitude.
The time sub-sampling of the data is mathematically expressed as:
A′i[m，n]＝Ai[m，tsamplen]
wherein A isiRepresenting original array, A'iRepresenting the array after change, m representing the array width, n representing the original array length, tsampleRepresents a time sub-sampling multiple;
the mathematical expression for the horizontal axis data interpolation is:
Figure BDA0003064615710000061
wherein XiRepresenting the interpolated value, XdstRepresenting the original data value, widthdstRepresents XdstAnd XdstA difference value of +1, widthsrcIndicating the distance of the interpolated position from the last reference value.
(2) Because the reflection fluctuation of the direct wave on the earth surface is large due to the fluctuation of the earth surface, and the underground dielectric constant is uneven, the condition that the data representing the underground target part is not outstanding enough can be caused, and the final result can be influenced in the learning process, so that the aim of background removal processing needs to be fulfilled by performing sliding filtering on the data. As shown in fig. 4, fig. 4(a) shows an original image, and fig. 4(b) shows an image after background removal processing.
(3) For the data obtained in step 1, according to the abscissa position of the target object marked in step 1(2), "leakage" or "non-leakage" labels are marked on the whole data, data segmentation is performed on the data, the whole data is divided into a plurality of modules according to equal abscissa lengths, and the label of the data is determined by whether the target horizontal position (i.e. the position where the pipeline leaks or the position where the pipeline does not leak) is located between the horizontal positions of the transmitting and receiving antennas, and is expressed by a formula:
Figure BDA0003064615710000071
wherein onehot (m)i) The function shows how to convert the values of different text labels representing categories into third-order vector form for input into the network, i.e., [100 ] onehot (0) ]]Indicating no object, onehot (1) ═ 010]Indicating no leakage in the pipe, onehot (2) ═ 001]Indicating a pipe leak.
FIG. 5 is a schematic diagram of data slicing, in which AiRepresenting a B-scan data array, X [ k ]]Representing the array after slicing, Y [ i ]]Representing an array of tags. A in the figureiI.e. the B-scan image corresponds to the image of a leakage target of a certain pipe. So the label corresponding to the sub-picture section representing the pipe is [010 ]]The section indicating the existence of leakage is marked as [001 ]]And the label corresponding to the sub-image of the leakage target away from the pipeline is set to [100 ]]Indicating no target.
The data after segmentation can correspond to the label and be input into the neural network together, and then the association between the data and the label is obtained through the step-by-step learning of the multi-layer network, so that the task of identification is completed.
And step 3: constructing a convolutional neural network to match test data
A convolutional neural network module cascaded by convolutional layers and pooling layers is constructed, as shown in fig. 2, the convolutional neural network comprises 4 convolutional layers, 2 pooling layers, 3 Dropout layers, 1 Flatten layer and 3 fully-connected layers. The 4-layer convolutional layer can modularize data characteristics with great efficiency, and calculation efficiency is improved. The pooling layer is added after the first two convolutional network layers because the data input into the network is large and information needs to be gathered through the pooling layer. And adding a Flatten layer after the convolution layer is finished, wherein the Flatten layer levels the data set into one-bit data and inputs the one-bit data into a full connection layer, namely the layer arranges the two-dimensional data in columns, each column is added into the previous column, and finally the two-dimensional data is changed into one-bit data. Three full-connection layers are added later to play a role in classifying the network, the last full-connection layer outputs three data, the activation function selects a softmax function suitable for classification, the network can express the attributive probability of each type through learning, and then classification and identification are carried out on the data.
For an image I of size r x clargeFirst, a size of a × b (a) is extracted from the data<r，b<c) Module sample I ofsmallAfter training the sparse autoencoder, j different feature results and corresponding activation values f (W) can be obtained(1)Ismall+b(1)) Wherein W is(1)Heel b(1)Is a value obtained through training; next, the original image I is processedlargeEach of samples I of size a x bmRespectively calculate their corresponding activation values fm(W(1)Ismall+b(1)) Then, the activation value f of the sample obtained in the previous step is divided into activation values f after image blockingmConvolution operation is carried out, and then j (r-a +1) x (c-b +1) mappings containing weight values are obtained. Adding pooling layers to convolutional networks further translates activation stepsThe output of the step by compressing the output of the neuron region into a single output.
And 4, step 4: training convolutional neural networks
(1) Setting training step number n, learning rate lr and training sample size batch _ size;
(2) inputting data and a label into a neural network for training;
(3) adjusting the training parameters to enable the cost function to have a loss rate close to 0 and an accuracy rate close to 1;
(4) and storing the convolutional layer parameters to obtain and store a trained convolutional neural network weight file suitable for desert pipeline discrimination.
Inputting the test data into the convolutional neural network model obtained in step 4, and reading the stored weight file through the network to obtain a prediction result, as shown in fig. 6, a probability curve shown on the graph can clearly represent the target position.
The invention provides an automatic target identification method based on a convolutional neural network, which can automatically extract the characteristics of pipelines and leakage from input data, only needs to manually input part of labels in training and does not need excessive intervention. The trained network can be used for efficiently identifying the underground image of the pipeline in the desert area in real time, and can be widely applied to leakage identification of the underground pipeline in the desert area.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that various changes in the embodiments and/or modifications of the invention can be made, and equivalents and modifications of some features of the invention can be made without departing from the spirit and scope of the invention.

Claims (7)

1. A desert area pipeline leakage target identification method based on a convolutional neural network is characterized by comprising the following steps:
s1, collecting target data by using a ground penetrating radar;
s2, preprocessing the data;
s3, constructing a convolutional neural network;
and S4, training and testing the convolutional neural network.
2. The method for identifying the desert area pipeline leakage target based on the convolutional neural network as claimed in claim 1, wherein: the collecting of the target data by using the ground penetrating radar in the step S1 specifically includes:
s11, defining a measuring area: dividing the detection area according to the map and the pipeline walking diagram, and respectively selecting a pipeline direction detection line and a vertical pipeline direction measuring line for detection and collection;
s12, detecting the measuring line: detecting and collecting according to the surveying and designed line, adopting a continuous detection mode to move forward stably along a certain direction of a measuring line, reading the abnormal section of the radar in real time in the detection process, recording the horizontal coordinate of the abnormal part, taking the horizontal coordinate as the horizontal coordinate of a target object, marking and recording the horizontal coordinate in a class newspaper;
s13, simulating leakage detection: for the simulation of the leakage pipeline, a pit is dug above the pipeline, water is poured into the pit to enable the pit to naturally permeate, the area is buried after the water is completely permeated, and then the area is detected and collected according to the detection step in the step S12.
3. The method for identifying the desert area pipeline leakage target based on the convolutional neural network as claimed in claim 1, wherein: the preprocessing of the data in the step S2 specifically includes:
s21, performing transverse data interpolation and time sub-sampling on the data: the data in the time direction is time sub-sampled and the lateral data in the lateral group is interpolated to place the amount of data in both dimensions in one order of magnitude.
S22, performing sliding filtering on the data to remove the background;
s23, labeling the whole data with a leakage label or a non-leakage label according to the abscissa position of the target object marked in the step S12, performing data segmentation on the data, and dividing the whole data into a plurality of modules according to the equal abscissa length;
and S24, corresponding the segmented data to the label as the input of the convolutional neural network.
4. The method for identifying the desert area pipeline leakage target based on the convolutional neural network as claimed in claim 3, wherein: the time sub-sampling mathematical expression in step S21 is:
A′i[m，n]＝Ai[m，tsamplen]；
the mathematical expression for the horizontal axis data interpolation is:
Figure FDA0003064615700000021
5. the method for identifying the desert area pipeline leakage target based on the convolutional neural network as claimed in claim 3, wherein: the label of the data in step S23 is determined by whether the target horizontal position, i.e. the leaking position of the pipeline or the non-leaking position of the pipeline, is located at the horizontal position of the transceiving antenna, and the mathematical expression is as follows:
Figure FDA0003064615700000022
6. the method for identifying the desert area pipeline leakage target based on the convolutional neural network as claimed in claim 1, wherein: the convolutional neural network in the step S3 includes 4 convolutional layers, 2 pooling layers, 3 Dropout layers, 1 Flatten layer, and 3 fully-connected layers; the Flatten layer can level the data set into one bit data input into the fully connected layer.
7. The method for identifying the desert area pipeline leakage target based on the convolutional neural network as claimed in claim 1, wherein: the training step in step S4 specifically includes:
s41, setting training step number n, learning rate lr and training sample size batch _ size;
s42, inputting the data and the labels into a convolutional neural network for training;
s43, adjusting the training parameters to enable the loss rate of the cost function to be close to 0 and the accuracy rate to be close to 1;
s44, storing convolutional layer parameters, obtaining a convolutional neural network weight file suitable for desert pipeline discrimination after training is completed, and storing the convolutional neural network weight file;
the testing steps are specifically as follows: and inputting the test data into a convolutional neural network model obtained after training is finished, and reading the stored weight file through a network to obtain a prediction result.
CN202110522573.0A 2021-05-13 2021-05-13 Method for identifying pipeline leakage target in desert area based on convolutional neural network Pending CN113392705A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110522573.0A CN113392705A (en) 2021-05-13 2021-05-13 Method for identifying pipeline leakage target in desert area based on convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110522573.0A CN113392705A (en) 2021-05-13 2021-05-13 Method for identifying pipeline leakage target in desert area based on convolutional neural network

Publications (1)

Publication Number Publication Date
CN113392705A true CN113392705A (en) 2021-09-14

Family

ID=77617142

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110522573.0A Pending CN113392705A (en) 2021-05-13 2021-05-13 Method for identifying pipeline leakage target in desert area based on convolutional neural network

Country Status (1)

Country Link
CN (1) CN113392705A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114694051A (en) * 2022-03-16 2022-07-01 哈尔滨工业大学水资源国家工程研究中心有限公司 Electromagnetic method pipeline disease identification and positioning method based on improved CapsNet network

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108108677A (en) * 2017-12-12 2018-06-01 重庆邮电大学 One kind is based on improved CNN facial expression recognizing methods
CN108182450A (en) * 2017-12-25 2018-06-19 电子科技大学 A kind of airborne Ground Penetrating Radar target identification method based on depth convolutional network
CN109685011A (en) * 2018-12-25 2019-04-26 北京华航无线电测量研究所 A kind of underground utilities detection recognition method based on deep learning
CN109838696A (en) * 2019-01-09 2019-06-04 常州大学 Pipeline fault diagnostic method based on convolutional neural networks
CN109931506A (en) * 2019-03-14 2019-06-25 三川智慧科技股份有限公司 Pipeline leakage detection method and device
US20190383904A1 (en) * 2018-06-13 2019-12-19 Metawave Corporation Autoencoder assisted radar for target identification
CN110598767A (en) * 2019-08-29 2019-12-20 河南省收费还贷高速公路管理有限公司航空港分公司 SSD convolutional neural network-based underground drainage pipeline defect identification method
CN112762362A (en) * 2021-01-15 2021-05-07 中国海洋石油集团有限公司 Underwater pipeline leakage acoustic emission detection method based on convolutional neural network

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108108677A (en) * 2017-12-12 2018-06-01 重庆邮电大学 One kind is based on improved CNN facial expression recognizing methods
CN108182450A (en) * 2017-12-25 2018-06-19 电子科技大学 A kind of airborne Ground Penetrating Radar target identification method based on depth convolutional network
US20190383904A1 (en) * 2018-06-13 2019-12-19 Metawave Corporation Autoencoder assisted radar for target identification
CN109685011A (en) * 2018-12-25 2019-04-26 北京华航无线电测量研究所 A kind of underground utilities detection recognition method based on deep learning
CN109838696A (en) * 2019-01-09 2019-06-04 常州大学 Pipeline fault diagnostic method based on convolutional neural networks
CN109931506A (en) * 2019-03-14 2019-06-25 三川智慧科技股份有限公司 Pipeline leakage detection method and device
CN110598767A (en) * 2019-08-29 2019-12-20 河南省收费还贷高速公路管理有限公司航空港分公司 SSD convolutional neural network-based underground drainage pipeline defect identification method
CN112762362A (en) * 2021-01-15 2021-05-07 中国海洋石油集团有限公司 Underwater pipeline leakage acoustic emission detection method based on convolutional neural network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
廖彬彬: "深度学习在探地雷达目标识别中的应用研究", 《中国优秀硕士学位论文全文数据库 基础科学辑》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114694051A (en) * 2022-03-16 2022-07-01 哈尔滨工业大学水资源国家工程研究中心有限公司 Electromagnetic method pipeline disease identification and positioning method based on improved CapsNet network

Similar Documents

Publication Publication Date Title
Liu et al. Detection and localization of rebar in concrete by deep learning using ground penetrating radar
Wang et al. Successful leveraging of image processing and machine learning in seismic structural interpretation: A review
CN112700429B (en) Airport pavement underground structure disease automatic detection method based on deep learning
Hou et al. Improved Mask R-CNN with distance guided intersection over union for GPR signature detection and segmentation
Brown et al. Multiple methods, maps, and management applications: Purpose made seafloor maps in support of ocean management
CN104376330B (en) Polarimetric SAR Image Ship Target Detection method based on super-pixel scattering mechanism
CN109100741A (en) A kind of object detection method based on 3D laser radar and image data
CN109685011A (en) A kind of underground utilities detection recognition method based on deep learning
CN114296075B (en) Ground penetrating radar image artificial intelligence identification method and device
CN103236063B (en) Based on the SAR image oil spilling detection method of multiple dimensioned spectral clustering and decision level fusion
CN105809194B (en) A kind of method that SAR image is translated as optical image
CN110009010A (en) Wide area optical remote sensing target detection method based on the re-detection of interest region
CN111624606B (en) Radar image rainfall identification method
CN111476159A (en) Method and device for training and detecting detection model based on double-angle regression
CN106446919A (en) Quick detection method for ground penetrating radar hyperbolic curve target
Pasolli et al. Automatic detection and classification of buried objects in GPR images using genetic algorithms and support vector machines
CN113610070A (en) Landslide disaster identification method based on multi-source data fusion
CN114549589B (en) Rotator vibration displacement measurement method and system based on lightweight neural network
CN112734695B (en) SAR image change detection method based on regional enhancement convolutional neural network
CN111381275A (en) First arrival picking method and device for seismic data
Yang et al. Evaluation and mitigation of rain effect on wave direction and period estimation from X-band marine radar images
CN115308803A (en) Coal seam thickness prediction method, device, equipment and medium
CN113392705A (en) Method for identifying pipeline leakage target in desert area based on convolutional neural network
Zhao et al. A Comprehensive Horizon‐Picking Method on Subbottom Profiles by Combining Envelope, Phase Attributes, and Texture Analysis
Feng et al. Subsurface Object 3D Modeling Based on Ground Penetration Radar Using Deep Neural Network

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20210914