CN110826588A - Drainage pipeline defect detection method based on attention mechanism - Google Patents

Drainage pipeline defect detection method based on attention mechanism Download PDF

Info

Publication number
CN110826588A
CN110826588A CN201910810653.9A CN201910810653A CN110826588A CN 110826588 A CN110826588 A CN 110826588A CN 201910810653 A CN201910810653 A CN 201910810653A CN 110826588 A CN110826588 A CN 110826588A
Authority
CN
China
Prior art keywords
module
stride
layer
filters
total
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
CN201910810653.9A
Other languages
Chinese (zh)
Inventor
潘刚
***·乌马尔·泽山
郑耀先
孙迪
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.)
Tianjin University
Original Assignee
Tianjin University
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 Tianjin University filed Critical Tianjin University
Priority to CN201910810653.9A priority Critical patent/CN110826588A/en
Publication of CN110826588A publication Critical patent/CN110826588A/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
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2431—Multiple classes
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Tourism & Hospitality (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Educational Administration (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Quality & Reliability (AREA)
  • Development Economics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to a drainage pipeline defect detection method based on an attention mechanism, which comprises the following steps: extracting a defect picture from the pipeline defect detection report, classifying the defect picture according to the defect type corresponding to the defect picture, and totally dividing into: deformation, corrosion, scaling, stagger, deposition, leakage, and cracking; constructing an attention mechanism module, wherein the attention mechanism module is a CBAM module; building a neural network model based on an attention mechanism, wherein the last 3 convolutional layers of the VGG16 network are removed from the convolutional network, and a plurality of CBAM modules are added on the basis; training the neural network by using a back propagation algorithm, verifying in the training process, and storing an optimal model for network training when the verification accuracy is highest; and testing the test set by using the stored optimal model to obtain a classification result of the pipeline defect picture.

Description

Drainage pipeline defect detection method based on attention mechanism
Technical Field
The invention relates to the computer field of computer vision, computer image processing, deep learning and the like and the drainage pipeline abnormity detection field. The invention more focuses on the application of the deep learning technology to the drainage pipeline defect detection.
Background
Drainage pipe systems are one of the largest infrastructures in cities intended to collect and transport wastewater and rain water. The normal use of the system is very important for the safety of urban drainage. With the rapid development of cities in recent years, China highlights the problems of insufficient underground pipeline construction scale, low management level and the like. Some cities have rainstorm, road collapse and other accidents in succession, and the life and urban operation order of people are seriously influenced. Regular inspection and repair of drainage pipelines is therefore an indispensable measure in urban construction.
Pipeline Closed Circuit Television (CCTV) is the most widely used technology in the pipeline Inspection field to date. The system consists of a pipeline detection robot and a CCTV camera arranged on the robot, and the equipment acquires video data (1) containing the internal information of the pipeline under the remote control operation of staff or the automatic control of a computer. And the inspector performs manual defect identification according to the captured video and then writes a pipeline defect detection report. This approach is often overly dependent on the experience of the test person, is highly subjective, and requires a great deal of time and effort.
In recent years, deep learning, particularly Convolutional Neural Networks (CNNs), has made tremendous progress in the field of computer vision. A convolutional neural network-based pipeline defect detection method is then generated, for example: kumar et al, 2018, proposed a simple classification convolutional neural network for classifying three defects, root invasion, deposition and fissures [2 ]; li et al use a deep convolutional neural network with hierarchical classification to detect classification of sewer defect types from unbalanced CCTV inspection data detection [3 ]. However, these methods do not utilize the unique features of the pipeline data, and the effect is not ideal.
Reference to the literature
[1]Costello，S.，Chapman，D.，Rogers，C.，Metje，N.：Underground assetlocation and condition assessment technologies.Tunnelling and UndergroundSpace Technology 22(5-6)，524-542(2007).
[2]Kumar，Srinath S.，et al.″Automated defect classification in sewerclosed circuit television inspections using deep convolutional neuralnetworks.″Automation in Construction 91(2018)：273-283.
[3]Li，Duanshun，Anran Cong，and Shuai Guo.″Sewer damage detection fromimbalanced CCTV inspection data using deep convolutional neural networks withhierarchica1 classification.″Automation in Construction 101(2019)：199-208.
Disclosure of Invention
The invention uses a structure based on a VGG-16 convolutional neural network as a main model, provides an automatic detection method for the pipeline defects based on an attention mechanism to solve the problems, and utilizes a large number of labeled pipeline abnormal samples to train and test. The technical scheme of the invention is as follows:
a drainage pipeline defect detection method based on an attention mechanism comprises the following steps of:
step 1: extracting a defect picture from the pipeline defect detection report, classifying the defect picture according to the defect type corresponding to the defect picture, and totally dividing into: deformation, corrosion, scaling, stagger, deposition, leakage, and cracking.
Step 2: the data set is divided into a training set test set and a validation set.
And step 3: and constructing an Attention mechanism Module, wherein the adopted Attention mechanism Module is a CBAM (ConvolationBlock Attention Module), and the Module can be directly added into a convolutional neural network model and used as a network layer. The input of the module is the output O e R of the convolution layerC×H×WWherein C is the channel number of the Feature map, H and W are the height and width of the Feature map respectively, and O passes through the channel attention module A successivelyC∈RC×1×1And spatial attention Module AS∈R1×H×WFinally obtaining output Oa∈RC×H×W.
Figure BDA0002184955120000021
Figure BDA0002184955120000022
Wherein
Figure BDA0002184955120000023
Representing element-by-element multiplication, O1Intermediate results are shown. O obtained via CBAM ModuleaCan be directly re-fed into other parts of the convolutional neural network.
And 4, step 4: building a neural network model based on an attention mechanism, wherein the last 3 convolutional layers of the VGG16 network are removed from the convolutional network, and a plurality of CBAM modules are added on the basis, and the general structure of the convolutional neural network is as follows:
the first convolution layer has a convolution kernel size of 3 × 3, 64 filters in total, and a stride of 1.
The first attention module, CBAM module, scaling factor r 8.
The second convolution layer has a convolution kernel size of 3 × 3, 64 filters in total, and a stride of 1.
The first largest pooling layer, pooling size 2 x 2, stride 2.
In the third convolution layer, the convolution kernel size is 3 × 3, 128 filters in total, and stride is 1.
The second attention module, CBAM module, scales by a factor r of 8.
The fourth convolution layer has a convolution kernel size of 3 × 3, 128 filters in total, and a stride of 1.
The second largest pooling layer, pooling size 2 x 2, stride 2.
In the fifth convolution layer, the convolution kernel size is 3 × 3, 256 filters in total, and stride is 1.
The third attention module, CBAM module, scales by a factor r of 8.
In the sixth convolution layer, the convolution kernel size is 3 × 3, 256 filters in total, and stride 1.
The fourth attention module, CBAM module, scales by a factor r of 8.
The seventh convolution layer has a convolution kernel size of 3 × 3, a total of 256 filters, and a stride of 1.
The third largest pooling layer, pooling size 2 x 2, stride 2.
In the eighth convolution layer, the convolution kernel size is 3 × 3, and there are 512 filters in total, and stride is 1.
The fifth attention module, CBAM module, scales by a factor r of 8.
In the ninth convolution layer, the convolution kernel size is 3 × 3, 512 filters in total, and stride is 1.
The sixth attention module, CBAM module, scales by a factor r of 8.
In the tenth convolution layer, the convolution kernel size is 3 × 3, a total of 512 filters, and the stride is 1.
The fourth largest pooling layer, pooling size 2 x 2, stride 2.
In the first full link layer, the number of nodes is 4096.
In the second full connection layer, the number of nodes is 4096.
And finally, outputting the probability that the image corresponds to 7 abnormal categories by a Softmax layer.
And 5: and training the neural network by using a back propagation algorithm, verifying in the training process, and storing the optimal model of the network training when the verification accuracy is highest.
Step 6: and testing the test set by using the stored optimal model to obtain a classification result of the pipeline defect picture.
Drawings
FIG. 1 is a flow chart of the present invention
FIG. 2 is a network architecture diagram of the present invention
FIG. 3 is a structural diagram of CBAM
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail below with reference to the accompanying drawings by way of examples. It is clear that the described implementations are only a part of the embodiments of the invention, and not an exhaustive list of all embodiments. And features of the implementations and embodiments of the present description may be combined with each other without conflict.
The processing steps of the invention comprise: preparing and processing data, establishing a training set verification set test set, training a convolutional neural network, and automatically classifying defects by adopting a trained network model.
Step 1: and (4) preparing and processing data. And extracting a defect picture inside the pipeline from the obtained drainage pipeline defect detection report by adopting a python program, and classifying the picture according to the report content and the defect type. Selecting seven types of pictures with good picture quality, such as deformation, corrosion, scaling, dislocation, deposition, leakage, cracking and the like.
Step 2: and dividing seven types of selected deformation, corrosion, scaling, stagger, deposition, leakage and other defect types into a training set, a verification set and a test set respectively, wherein the division ratio is 3: 1.
And step 3: and (4) building an attention mechanism module, wherein the adopted attention mechanism structure is a CBAM module. The input of the module is the output O e R of the convolution layerC×H×WWhere C is the number of channels of the Feature map, and H and W are the height and width of the Feature map, respectively. The attention mechanism module can be divided into 2 parts, wherein the first part is a channel attention module, and the second part is a space attention module.
The first part passes the input O through the channel attention module aC(O)∈RC×1×1To obtain an intermediate output O1∈RC×H×W。
The part can be divided into the following 3 steps:
(1) firstly, respectively carrying out global average pooling GloAvgPool and global maximum pooling GloMaxPool on input, and then respectively passing the result through two full-connection layers Dense1And Dense2To obtain two outputs d with the same dimension1∈RC×1×1And d2∈RC×1×1Namely:
d1＝Dense2(Dense1(GloAvgPool(O)))
d2＝Dense2(Dense1(GloMaxPool(O)))
wherein Dense1The number of the nodes is
Figure BDA0002184955120000041
r represents a scaling factor, and in this experiment r is 8. Dense2The number of the nodes is C.
(2) Will d1And d2Adding, sending the result into sigmoid function to obtain channel attention weight AC(O)∈RC×1×1. The calculation formula is as follows:
AC(O)＝Sigmoid(d1+d2)
(3) input O and channel attention weight AC(O)∈RC×1×1Performing element-by-element multiplication to obtain the output O of the first part1∈RC×H×W。
Figure BDA0002184955120000042
Representing element-by-element multiplication, the calculation formula is as follows:
Figure BDA0002184955120000043
intermediate result O to be obtained by the second part1Through space attention Module AS(O1)∈R1×H×WFinally obtaining output Oa∈RC×H×W. The part can be divided into the following 4 steps:
(1) the inputs were averaged in channel dimensions ChanAvgPool and maximal pooled ChanMaxPool, respectively, and the results were then concatenated. The calculation formula is as follows:
dc＝[ChanAvgPool(O1)；ChanMaxPool(O1)]
(2) the result d obtainedcPassing through a convolutional layer having a filter with a size of 7 × 7, and then feeding the output of the convolutional layer into a sigmoid function to obtain a spatial attention weight AS(O1)∈R1×H×W。
AS(O1)＝Sigmoid(Con2D7×7(dc))
(3) Intermediate result O1And spatial attention weight AS(O1)∈R1×H×WPerforming element-by-element multiplication to obtain output Oa∈RC×H×W. The calculation formula is as follows:
Figure BDA0002184955120000044
and 4, step 4: and (3) building a convolutional neural network model based on an attention mechanism, wherein the last 3 convolutional layers of the VGG16 network are removed from the convolutional network, and a plurality of CBAM modules are added on the basis. The model overall network structure is as follows:
the first convolution layer has a convolution kernel size of 3 × 3, 64 filters in total, and a stride of 1.
The first attention module, CBAM module, scaling factor r 8.
The second convolution layer has a convolution kernel size of 3 × 3, 64 filters in total, and a stride of 1.
The first largest pooling layer, pooling size 2 x 2, stride 2.
In the third convolution layer, the convolution kernel size is 3 × 3, 128 filters in total, and stride is 1.
The second attention module, CBAM module, scales by a factor r of 8.
The fourth convolution layer has a convolution kernel size of 3 × 3, 128 filters in total, and a stride of 1.
The second largest pooling layer, pooling size 2 x 2, stride 2.
In the fifth convolution layer, the convolution kernel size is 3 × 3, 256 filters in total, and stride is 1.
The third attention module, CBAM module, scales by a factor r of 8.
In the sixth convolution layer, the convolution kernel size is 3 × 3, 256 filters in total, and stride 1.
The fourth attention module, CBAM module, scales by a factor r of 8.
The seventh convolution layer has a convolution kernel size of 3 × 3, a total of 256 filters, and a stride of 1.
The third largest pooling layer, pooling size 2 x 2, stride 2.
In the eighth convolution layer, the convolution kernel size is 3 × 3, and there are 512 filters in total, and stride is 1.
The fifth attention module, CBAM module, scales by a factor r of 8.
In the ninth convolution layer, the convolution kernel size is 3 × 3, 512 filters in total, and stride is 1.
The sixth attention module, CBAM module, scales by a factor r of 8.
In the tenth convolution layer, the convolution kernel size is 3 × 3, a total of 512 filters, and the stride is 1.
The fourth largest pooling layer, pooling size 2 x 2, stride 2.
In the first full link layer, the number of nodes is 4096.
In the second full connection layer, the number of nodes is 4096.
And finally, outputting the probability that the image corresponds to 7 abnormal categories by a Softmax layer.
And 5: and training the convolutional neural network model, and storing the optimal model weight.
Step 6: and automatically classifying the pipeline defect pictures by adopting the stored optimal model weight.
The network structure is built by taking TensorFlow as a background and using a Keras deep learning library. The language used was Python. In this embodiment, a convolutional neural network structure and internal operations such as a convolutional layer, a max pooling layer, a full link layer, a Softmax layer, an Add function, and a Multiply function are constructed by writing convolutional network layers in the library through a function model in the Keras deep learning library as an overall structure.
According to the pipeline abnormal type detection method provided by the embodiment of the invention, after the image to be recognized is obtained, the classification can be carried out without manually defining the characteristics by a user, and the type of the image to be recognized can be judged by directly utilizing the deep learning network obtained by pre-training. And because the attention mechanism module is added in the network, the network can gradually extract the unique characteristics of different defect pictures in the training process, so that the recognition capability of the network on the pipeline defect pictures is greatly improved, the automatic detection and classification of the pipeline defects can be accurately realized, and the network has good application value.

Claims (3)

1. A drainage pipeline defect detection method based on an attention mechanism comprises the following steps:
step 1: extracting a defect picture from the pipeline defect detection report, classifying the defect picture according to the defect type corresponding to the defect picture, and totally dividing into: deformation, corrosion, scaling, stagger, deposition, leakage, and cracking.
Step 2: dividing a data set into a training set test set and a verification set;
and step 3: constructing an Attention mechanism Module, wherein the adopted Attention mechanism Module is a CBAM (ConvolationBlock Attention Module), and the Module can be directly added into a convolutional neural network model and used as a network layer; the input of the module is the output O e R of the convolution layerC×H×WWherein C is the channel number of the Feature map, H and W are the height and width of the Feature map respectively, and O passes through the channel attention module A successivelyC∈RC×1×1And spatial attention Module AS∈R1×H×WFinally obtaining output Oa∈RC×H×W.
Figure FDA0002184955110000011
Figure FDA0002184955110000012
Wherein
Figure FDA0002184955110000013
Representing element-by-element multiplication, O1Representing an intermediate result; o obtained via CBAM ModuleaCan be directly re-fed into other parts of the convolutional neural network;
and 4, step 4: building a neural network model based on an attention mechanism, wherein the last 3 convolutional layers of the VGG16 network are removed from the convolutional network, and a plurality of CBAM modules are added on the basis, and the general structure of the convolutional neural network is as follows:
a first convolution layer with convolution kernel size of 3 × 3, 64 filters in total, and stride of 1;
a first attention module, CBAM module, scaling factor r 8;
a second convolution layer with a convolution kernel size of 3 × 3, 64 filters in total, and a stride of 1;
a first maximum pooling layer with a pooling size of 2 x 2 and a stride of 2;
a third convolution layer, the convolution kernel size is 3 × 3, 128 filters in total, and stride is 1;
a second attention module, CBAM module, scaling factor r 8;
a fourth convolution layer, with a convolution kernel size of 3 × 3, 128 filters in total, and a stride of 1;
a second maximum pooling layer with a pooling size of 2 x 2 and a stride of 2;
a fifth convolution layer with a convolution kernel size of 3 × 3, a total of 256 filters, and a stride of 1;
a third attention module, CBAM module, scaling factor r 8;
a sixth convolution layer with a convolution kernel size of 3 × 3, a total of 256 filters, and a stride of 1;
a fourth attention module, CBAM module, scaling factor r 8;
a seventh convolution layer, with a convolution kernel size of 3 × 3, a total of 256 filters, and a stride of 1;
a third maximum pooling layer with a pooling size of 2 x 2 and a stride of 2;
an eighth convolution layer, the convolution kernel size is 3 × 3, 512 filters in total, and stride is 1;
a fifth attention module, CBAM module, scaling factor r 8;
a ninth convolution layer, with a convolution kernel size of 3 × 3, a total of 512 filters, and a stride of 1;
a sixth attention module, CBAM module, scaling factor r 8;
a tenth convolution layer with a convolution kernel size of 3 × 3, a total of 512 filters, and a stride of 1;
a fourth maximum pooling layer, the pooling size being 2 x 2, and the stride being 2;
4096 nodes in the first full-connection layer;
in the second full-connection layer, the number of nodes is 4096;
a Softmax layer, which finally outputs the probability that the image corresponds to 7 abnormal categories;
and 5: training the neural network by using a back propagation algorithm, verifying in the training process, and storing an optimal model for network training when the verification accuracy is highest;
step 6: and testing the test set by using the stored optimal model to obtain a classification result of the pipeline defect picture.
2. The method according to claim 1, wherein in step 2), the data enhancement method is selected from gaussian noise, image inversion and color dithering.
3. The method of claim 1, wherein in step 5), network training is performed using an Nvidia GPU.
CN201910810653.9A 2019-08-29 2019-08-29 Drainage pipeline defect detection method based on attention mechanism Pending CN110826588A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910810653.9A CN110826588A (en) 2019-08-29 2019-08-29 Drainage pipeline defect detection method based on attention mechanism

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910810653.9A CN110826588A (en) 2019-08-29 2019-08-29 Drainage pipeline defect detection method based on attention mechanism

Publications (1)

Publication Number Publication Date
CN110826588A true CN110826588A (en) 2020-02-21

Family

ID=69547856

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910810653.9A Pending CN110826588A (en) 2019-08-29 2019-08-29 Drainage pipeline defect detection method based on attention mechanism

Country Status (1)

Country Link
CN (1) CN110826588A (en)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111458148A (en) * 2020-04-26 2020-07-28 上海电机学院 CBAM-based convolutional neural network rolling bearing fault diagnosis method
CN111784645A (en) * 2020-06-15 2020-10-16 北京科技大学 Crack detection method for filling pipeline
CN112241681A (en) * 2020-09-15 2021-01-19 天津大学 Liquid leakage detection method based on three-frame difference method and VGG-16
CN112364783A (en) * 2020-11-13 2021-02-12 诸暨思看科技有限公司 Part detection method and device and computer readable storage medium
CN112884709A (en) * 2021-01-18 2021-06-01 燕山大学 Yoov 3 strip steel surface defect detection and classification method introducing attention mechanism
CN112990391A (en) * 2021-05-20 2021-06-18 四川大学 Feature fusion based defect classification and identification system of convolutional neural network
CN113111828A (en) * 2021-04-23 2021-07-13 中国科学院宁波材料技术与工程研究所 Three-dimensional defect detection method and system for bearing
CN113362320A (en) * 2021-07-07 2021-09-07 北京工业大学 Wafer surface defect mode detection method based on deep attention network
CN113421252A (en) * 2021-07-07 2021-09-21 南京思飞捷软件科技有限公司 Actual detection method for vehicle body welding defects based on improved convolutional neural network
CN116363441A (en) * 2023-05-31 2023-06-30 克拉玛依市百事达技术开发有限公司 Pipeline corrosion detection system with marking function

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007024746A (en) * 2005-07-20 2007-02-01 Kitakyushu Foundation For The Advancement Of Industry Science & Technology Method and device for defect detection/discrimination inside conduit
US20160163035A1 (en) * 2014-12-03 2016-06-09 Kla-Tencor Corporation Automatic Defect Classification Without Sampling and Feature Selection
CN108038850A (en) * 2017-12-08 2018-05-15 天津大学 A kind of drainage pipeline Exception Type automatic testing method based on deep learning
WO2019085793A1 (en) * 2017-11-01 2019-05-09 腾讯科技（深圳）有限公司 Image classification method, computer device and computer readable storage medium
CN109800770A (en) * 2018-12-28 2019-05-24 广州海昇计算机科技有限公司 A kind of method, system and device of real-time target detection
CN110084210A (en) * 2019-04-30 2019-08-02 电子科技大学 The multiple dimensioned Ship Detection of SAR image based on attention pyramid network

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007024746A (en) * 2005-07-20 2007-02-01 Kitakyushu Foundation For The Advancement Of Industry Science & Technology Method and device for defect detection/discrimination inside conduit
US20160163035A1 (en) * 2014-12-03 2016-06-09 Kla-Tencor Corporation Automatic Defect Classification Without Sampling and Feature Selection
WO2019085793A1 (en) * 2017-11-01 2019-05-09 腾讯科技（深圳）有限公司 Image classification method, computer device and computer readable storage medium
CN108038850A (en) * 2017-12-08 2018-05-15 天津大学 A kind of drainage pipeline Exception Type automatic testing method based on deep learning
CN109800770A (en) * 2018-12-28 2019-05-24 广州海昇计算机科技有限公司 A kind of method, system and device of real-time target detection
CN110084210A (en) * 2019-04-30 2019-08-02 电子科技大学 The multiple dimensioned Ship Detection of SAR image based on attention pyramid network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
SANGHYUN WOO ET AL.: "CBAM: Convolutional Block Attention Module", pages 1 - 17 *
蒋柱武 等: "《给排水管道工程》", vol. 1, 上海：同济大学出版社, pages: 298 - 300 *

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111458148A (en) * 2020-04-26 2020-07-28 上海电机学院 CBAM-based convolutional neural network rolling bearing fault diagnosis method
CN111784645A (en) * 2020-06-15 2020-10-16 北京科技大学 Crack detection method for filling pipeline
CN111784645B (en) * 2020-06-15 2023-08-22 北京科技大学 Filling pipeline crack detection method
CN112241681A (en) * 2020-09-15 2021-01-19 天津大学 Liquid leakage detection method based on three-frame difference method and VGG-16
CN112364783A (en) * 2020-11-13 2021-02-12 诸暨思看科技有限公司 Part detection method and device and computer readable storage medium
CN112884709A (en) * 2021-01-18 2021-06-01 燕山大学 Yoov 3 strip steel surface defect detection and classification method introducing attention mechanism
CN113111828A (en) * 2021-04-23 2021-07-13 中国科学院宁波材料技术与工程研究所 Three-dimensional defect detection method and system for bearing
CN112990391A (en) * 2021-05-20 2021-06-18 四川大学 Feature fusion based defect classification and identification system of convolutional neural network
CN113362320A (en) * 2021-07-07 2021-09-07 北京工业大学 Wafer surface defect mode detection method based on deep attention network
CN113421252A (en) * 2021-07-07 2021-09-21 南京思飞捷软件科技有限公司 Actual detection method for vehicle body welding defects based on improved convolutional neural network
CN113421252B (en) * 2021-07-07 2024-04-19 南京思飞捷软件科技有限公司 Improved convolutional neural network-based vehicle body welding defect detection method
CN113362320B (en) * 2021-07-07 2024-05-28 北京工业大学 Wafer surface defect mode detection method based on deep attention network
CN116363441A (en) * 2023-05-31 2023-06-30 克拉玛依市百事达技术开发有限公司 Pipeline corrosion detection system with marking function
CN116363441B (en) * 2023-05-31 2023-08-08 克拉玛依市百事达技术开发有限公司 Pipeline corrosion detection system with marking function

Similar Documents

Publication Publication Date Title
CN110826588A (en) Drainage pipeline defect detection method based on attention mechanism
US20210319265A1 (en) Method for segmentation of underground drainage pipeline defects based on full convolutional neural network
Tan et al. Automatic detection of sewer defects based on improved you only look once algorithm
KR102166458B1 (en) Defect inspection method and apparatus using image segmentation based on artificial neural network
CN109671071B (en) Underground pipeline defect positioning and grade judging method based on deep learning
CN110555831B (en) Deep learning-based drainage pipeline defect segmentation method
US11715190B2 (en) Inspection system, image discrimination system, discrimination system, discriminator generation system, and learning data generation device
CN112070134A (en) Power equipment image classification method and device, power equipment and storage medium
CN111161224A (en) Casting internal defect grading evaluation system and method based on deep learning
CN111815576B (en) Method, device, equipment and storage medium for detecting corrosion condition of metal part
CN115497015A (en) River floating pollutant identification method based on convolutional neural network
CN112288700A (en) Rail defect detection method
CN115823508A (en) Air-ground integrated pipe network safety monitoring method and system
CN116703885A (en) Swin transducer-based surface defect detection method and system
CN114863361A (en) Distribution line bird nest detection method based on USRNet and YOLOv3
CN113516652A (en) Battery surface defect and adhesive detection method, device, medium and electronic equipment
US11663815B2 (en) System and method for inspection of heat recovery steam generator
CN113780111A (en) Pipeline connector based on optimized YOLOv3 algorithm and defect accurate identification method
CN113111923A (en) Water supply network leakage detection and positioning method based on one-dimensional migration learning convolutional neural network integrated model
Chen et al. Deep Learning Based Underground Sewer Defect Classification Using a Modified RegNet
CN115239034B (en) Method and system for predicting early defects of wind driven generator blade
CN115082650A (en) Implementation method of automatic pipeline defect labeling tool based on convolutional neural network
CN116188391A (en) Method and device for detecting broken gate defect, electronic equipment and storage medium
CN115330703A (en) Remote sensing image cloud and cloud shadow detection method based on context information fusion
Wang et al. Automatic Damage Segmentation Framework for Buried Sewer Pipes Based on Machine Vision: Case Study of Sewer Pipes in Zhengzhou, China

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