"一个人必须不停地写作，才能不被茫茫人海淹没"

"Everything is going to be allright" "一切都会好起来的"

参考文献

内容：按照缺陷识别、量化两个方面进行文献整理，并归纳引用文章的主要贡献.

异常检测（Anomaly Detection）：结合管道内检测的实际需求以及数据类型，可认为管道内的缺陷检测是异常检测在时序序列上的应用

目标检测（Object Detection）

1. 刘金海, 吴振宁, 王增国, 汪刚, and 殷宇殿. "漏磁内检测数据中的管壁缺陷特征提取方法." 北京工业大学学报 40, no. 7 (2014): 1041-1047.

2. 赵重阳. 管道漏磁检测数据特征提取及特征分析方法研究[D].东北大学,2014.

3. 苏铭. 基于图像信息的金属板缺陷特征提取方法研究[D].东北大学,2013.

本文的创新点是将主元分析方法、连通域和链码等算法和概念应用到金属板无损探伤的特征提取中来，通过对漏磁检测得到的金属板图像进行TPCA操作，就可以确定金属板缺陷区域的边缘，通过连通的概念能快速的遍历金属板缺陷边缘，并存储缺陷边缘的坐标，并同时实现对缺陷的编号，再根据链码信息计算出金属板缺階区域的几何特征参量，最终达到了金属板无损探伤特征提取的目的.

4. 王婷婷. 金属表面缺陷特征智能提取及特征分析的方法研究[D].东北大学,2017.

提出缺陷自适应阈值检测（结合差分阈值：相邻点插值，与幅值阈值：最大值）方法对漏磁数据进行缺陷检测，实现缺陷异常数据提取。

5. 马天航,胡家铖,郑莉,刚蓓,刘思娇.一种基于人工提取缺陷块的边界搜索方法[J].无损检测,2020,42(08):1-7.

1. Liu, Jinhai, Xiangkai Shen, Jianfeng Wang, Lin Jiang, and Huaguang Zhang. "An Intelligent Defect Detection Approach Based on Cascade Attention Network Under Complex Magnetic Flux Leakage Signals." IEEE Transactions on Industrial Electronics (2022).

2. 梁海波,王怡.基于深度学习的天然气钢制管道缺陷检测方法研究[J/OL].电子测量与仪器学报:1-11[2022-10-22]. 西南石油大学机电工程学院

本文针对天然气钢制管道腐蚀缺陷信号，提出了一种基于 VMD-1D-CNN-RF 的缺陷识别方法，对腐蚀缺陷进行特征提取和特征分类。特征提取的对象是超声回波信号，通过 VMD 对缺陷信号分解和重构，用 1D-CNN 对重构后的信号进行特征提取，最后用随机森林对特征进行分类识别。(小波基函数有 Haar 小波，Daubenchies 小波，Biorthogonal 小波，Symlets 小波，Mexican Hat 小波等)

3. Yan, Y., D. Liu, B. Gao, G. Y. Tian, and Z. C. Cai. "A deep learning-based ultrasonic pattern recognition method for inspecting girth weld cracking of gas pipeline." IEEE Sensors Journal 20, no. 14 (2020): 7997-8006. Q3

The proposed method utilizes a deep Convolution Neural Network (CNN) integrated with a pre-trained Support Vector Machine (SVM) classifier to extract the high-level features from the time-frequency representation of A-scan signals measured by bulk-wave EMAT and classify these signals into defective or non-defective groups.

They compared the performance of the proposed CNN with four widely used feature extraction algorithms, which include DWT, WPT, Shannon Entropy, and statistical features. Feature vectors extracted by different methods are input to the same RBF-kernel based SVM classifier

4. 王之桢. 基于漏磁数据成像的管道缺陷识别方法研究[D].沈阳工业大学,2021.

研究了管道漏磁数据预处理方法（基于阈值分割的漏磁异常数据检测算法）；研究了图像改进 Canny 算子的边缘提取和基于改进霍夫变换的直线检测算法识别管道环焊缝位置；研究了基于快速傅里叶变换的互相关算法；研究了各种图像处理算法。以管道漏磁数据作为实验数据，经实验证明，该方法效率较高，且可以有效识别出所有异常环焊缝位置，及异常数据所在的通道编号，还可以有效识别出管体缺陷所在的里程点位置。

5. Zhang, Min, Yanbao Guo, Qiuju Xie, Yuansheng Zhang, Deguo Wang, and Jinzhong Chen. "Defect identification for oil and gas pipeline safety based on autonomous deep learning network." Computer Communications 195 (2022): 14-26.

Autonomous deep learning algorithm (AUTDL) network was proposed, it includes three modules: MFL signal input module, data transformation module, and failure defect identification module. 

In the experiment, five kinds of MFL defect data of the pipeline are collected, includes normal condition, defects at the girth welding position, corrosion at the spiral welding position, transverse mechanical scratches, and dent. A total of 770 labels are made for all the collected pipeline defect data.

6. Liu, Shucong, Hongjun Wang, and Rui Li. "Attention Module Magnetic Flux Leakage Linked Deep Residual Network for Pipeline In-Line Inspection." Sensors 22, no. 6 (2022): 2230.

Data: Pipeline MFL pseudo-color image is used to directly map the converted MFL signal data to the RGB space, encode the three components, and finally fuse the three components into a pseudo-color image according to different proportions to obtain a complete MFL pseudo-color image, with size of (9000, 112, 112, 3).

The method introduces the attention mechanism and designs the spatial attention module (Spatial_AM) and the channel attention module (Channel_AM).

7. Bastian, Blossom Treesa, N. Jaspreeth, S. Kumar Ranjith, and C. V. Jiji. "Visual inspection and characterization of external corrosion in pipelines using deep neural network." NDT & E International 107 (2019): 102134. Q3

Data: 140,000 optical images with (256, 256, 3)

A vision based method for classification(no/low/medium/high corrosion) and localisation of corrosion in pipelines was proposed.

First, at the feature level, a recursive residual subnet is proposed to inject the mechanism features into the network, so that the network performance can be enhanced. Second, an experience-aided subnet is proposed to incorporate expert experience at the decision level, which supervises the network training with labels, so that the network stability can be improved. Third, at the modeling level, a cascade expression subnet based on two-point representation is first proposed in the defect size quantification area, where both the value and distribution of labels are considered to boost the network precision.   

Data:  The experiments are carried out with the defects from a pipeline network in northern China, where the pipeline is 800 meters long, the wall thickness is 12.7 mm, the outer diameter is 426 mm, and the material of the pipeline is X65 carbon steel. The driving speed and operation pressure of PIR are 0.5 m/s and 3 MPa. 479 defect samples are considered, with their length and width ranging from 8 mm to 60 mm, depth ranging from 0.2 mm to 10 mm. The size of defect signal is unified as 88×50.

2. Long, Yue, Jinghua Zhang, Songling Huang, Lisha Peng, Wenzhi Wang, Shen Wang, and Wei Zhao. "A Novel Crack Quantification Method for Ultra-High-Definition Magnetic Flux Leakage Detection in Pipeline Inspection." IEEE Sensors Journal 22, no. 16 (2022): 16402-16413.

This paper presents a detection topology in which the magnetic field signal is integrated first and then sampled, and it can effectively sample the leakage magnetic field of cracks.

3. 崔国宁. 基于深度学习的管道缺陷漏磁数据识别方法研究[D].沈阳工业大学,2022.DOI:10.27322/d.cnki.gsgyu.2022.001193.

在实际应用的管道检测中，包括管道缺陷的形状尺寸、管道中特殊组件产生的漏磁场、检测探头的提离值、管道所受的应力和磁化强度以及检测器的移动速度等因素，均会对漏磁信号的检测结果产生重要的影响，其中对漏磁信号影响最大的是缺陷的形状尺寸，不同大小的缺陷会对磁场磁通量的变化趋势产生相应的影响。

数据：轴向、径向和周向三个方向的分量均提取 20×100 大小的二维数组，组合为一个 60×100 的数据输入到网络中。输入层对输入数据的结构要求为 60×100×1 的数据。所有缺陷的样本数据共有 157个，其中有 122 个缺陷的数据作为训练集用于模型训练，有 20 个缺陷数据作为模型的验证集，剩余 15 个缺陷数据不参与网络模型的训练过程.