基于快速傅里叶卷积的电成像测井图像修复

苏乾潇; 乔德新; 任义丽; 冯周; 林盛斓; 黄睿琦

doi:10.13700/j.bh.1001-5965.2023.0754

基于快速傅里叶卷积的电成像测井图像修复

doi: 10.13700/j.bh.1001-5965.2023.0754

中国石油勘探开发研究院，北京 100083

基金项目:

国家自然科学基金(42372175)；国家科技重大专项(2025ZD1401501)

详细信息

通讯作者:
E-mail：qdx@petrochina.com.cn

中图分类号: P631.84
计量
- 文章访问数: 702
- HTML全文浏览量: 230
- PDF下载量: 51
- 被引次数: 0
出版历程
- 收稿日期: 2023-11-20
- 录用日期: 2024-01-05
- 网络出版日期: 2024-02-04
- 整期出版日期: 2026-01-15

Inpainting of electrical imaging logging images based on fast Fourier convolution

Research Institute of Petroleum Exploration and Development，Beijing 100083，China

Funds:

National Natural Science Foundation of China (42372175); National Science and Technology Major Project (2025ZD1401501)

More Information

Corresponding author: E-mail：qdx@petrochina.com.cn

摘要

摘要:
成像测井是复杂储层测井评价中的重要技术手段。通过成像测井，可以获得井周的电阻率分布二维图像，用于评价井壁缝洞发育和地层沉积构造等。但由于电阻率成像测井仪器的特点，电阻率测井图像上会出现空白条带，这增加了计算机对电成像资料处理的难度。目前的图像修复方法和现有的神经网络图像修复方法在待填充部分占比较大时效果都不好。因此，迫切需要基于深度学习的智能修复方法。以一种基于快速傅里叶卷积的成像测井图像空白条带填充网络为基础，将西南油气田的电成像测井图像构建为数据集，训练得到一种基于快速傅里叶卷积的成像测井图像空白条带智能填充深度学习算法。对比各种算法的时间，结果表明：所提算法在条带宽度大的成像测井图像中修复效果较优，同时修复效率提升明显。通过所提方法实现了成像测井空白条带的快速、准确和智能化修复，实现了全井眼图像的快速生成，并解决了全井眼图像获取的困难。
- 电成像测井 /
- 空白条带填充 /
- 神经网络 /
- 快速傅里叶卷积神经网络 /
- 图像修复
Abstract:
Imaging logging is an important technique for complex reservoir evaluation. It provides a two-dimensional image of the wellbore, showing the wellbore’s structural features and playing an important role in evaluating seam holes and sedimentary structures. However, the lack of a resistivity imaging logging device causes blank streaks to show up on the logging images, which makes computer data processing more challenging and affects visual continuity in manual identification. Currently, traditional image inpainting methods and neural networks do not perform well on filling the well logging images. Therefore, there is an urgent need to research a deep learning-based image inpainting method for blank streaks in imaging logging images. A dataset was constructed using electrical imaging logging images from the LG area of the Southwest Oil and Gas Field. This dataset was used to train a new deep learning algorithm for intelligent restoration of blank streaks based on fast Fourier convolution, based on a fast Fourier convolution neural network for filling blank streaks in imaging logging images. This technique solves the challenge of acquiring entire wellbore photos and makes it possible to quickly, accurately, and intelligently inpaint blank streaks in well logging images.
- electrical imaging logging /
- blank streaks filling /
- neural network /
- fast Fourier convolutions neural network /
- image inpainting

HTML全文

图 1 基于快速傅里叶卷积的电成像测井图像修复神经网络

Figure 1. Inpainting neural network of electrical imaging logging images based on fast Fourier convolution

下载: 全尺寸图片幻灯片

图 2 傅里叶卷积模块结构

Figure 2. Structural diagram of Fourier convolution module

下载: 全尺寸图片幻灯片

图 3 频谱转换模块结构示意图

Figure 3. Structural diagram of spectral transform module

下载: 全尺寸图片幻灯片

图 4 截取单极板图像

Figure 4. Capture single pad images

下载: 全尺寸图片幻灯片

图 5 Log-LaMa网络与Criminisi算法、Filtersim算法、FcF网络及MAT网络填充测井图像对比（井眼覆盖率为80%）

Figure 5. Comparison of the filling performance on logging images by Log-LaMa network with Criminisi algorithm, Filtersim algorithm, FcF network and MAT network (borehole coverage rate is 80%)

下载: 全尺寸图片幻灯片

图 6 Log-LaMa网络与Criminisi算法、Filtersim算法、FcF网络及MAT网络填充测井图像对比（井眼覆盖率为65%）

Figure 6. Comparison of the filling performance on logging images by Log-LaMa network with Criminisi algorithm, Filtersim algorithm, FcF network and MAT network (borehole coverage rate is 65%)

下载: 全尺寸图片幻灯片

图 7 Log-LaMa网络与Criminisi算法、Filtersim算法、FcF网络及MAT网络填充测井图像对比（井眼覆盖率为90%）

Figure 7. Comparison of the filling performance on logging images by Log-LaMa network with Criminisi algorithm, Filtersim algorithm, FcF network and MAT network (borehole coverage rate is 90%)

下载: 全尺寸图片幻灯片

表 1 下采样卷积层所采用卷积核个数、尺寸及通道数

Table 1. The number, size, and channels of the convolutional kernels used in the downsampling convolutional layer

层数	卷积核个数	卷积核尺寸	卷积核通道数
1	64	7×7	4
2	128	3×3	64
3	256	3×3	128

下载: 导出CSV

表 2 竖条状掩码示例

Table 2. Examples of vertical striped masks

掩码占比/%	掩码
10
20
30
40

下载: 导出CSV

表 3 随机形状掩码示例

Table 3. Examples of random shaped masks

掩码类型	掩码
细
中
粗

下载: 导出CSV

表 4 不同算法模型填充速度对比

Table 4. Comparison of filling speeds for different algorithm models

算法	填充每10 m电成像测井图像所耗时间/s
Log-LaMa	0.375
Criminisi	26 640
Filtersim	13
FcF	0.6
MAT	0.25

下载: 导出CSV

表 5 不同宽度条带下Log-LaMa网络与Filtersim算法填充FID对比

Table 5. Comparison of FID for the filling performance of Log-LaMa network and Filtersim algorithm under different stripe widths

条带占比/%	FID
条带占比/%	Log-LaMa	Criminisi	Filtersim	FcF	MAT
10	0.150	2.391	0.203	0.156	0.151
20	0.422	25.680	0.650	0.593	0.617
30	0.890	30.010	1.124	0.933	1.015
40	1.511	34.200	1.941	1.477	1.520

下载: 导出CSV

参考文献(26)

[1]	侯振学, 于雪娴, 李东旭, 等. 电成像测井处理新技术在储层评价方面的应用[J]. 地球物理学进展, 2020, 35(2): 573-578. HOU Z X, YU X X, LI D X, et al. Application of new processing technology of electrical imaging logging in reservoir evaluation[J]. Progress in Geophysics, 2020, 35(2): 573-578(in Chinese).
[2]	程超, 张亮, 李培彦, 等. 碎屑岩泥质分布形式的测井评价研究进展[J]. 地球物理学进展, 2021, 36(3): 1052-1061. CHENG C, ZHANG L, LI P Y, et al. Advances in logging evaluation of shale distribution in clastic rocks[J]. Progress in Geophysics, 2021, 36(3): 1052-1061(in Chinese).
[3]	ASSOUS S, WHETTON J A, ELKINGTON P A. Microresistivity image inpainting and visualization[C]//Proceedings of the SPE Annual Technical Conference and Exhibition. Pittsburgh: SPE, 2013: D021S039R008.
[4]	李雪涛, 王耀雄, 高放. 图像修复方法综述[J]. 激光与光电子学进展, 2023, 60(2): 29-44. LI X T, WANG Y X, GAO F. Review of image inpainting methods[J]. Laser & Optoelectronics Progress, 2023, 60(2): 29-44(in Chinese).
[5]	JAIN J, ZHOU Y Q, YU N, et al. Keys to better image inpainting: structure and texture go hand in hand[C]//Proceedings of the 2023 IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE Press, 2023: 208-217.
[6]	YU J H, LIN Z, YANG J M, et al. Generative image inpainting with contextual attention[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2018: 5505-5514.
[7]	LAHIRI A, JAIN A K, AGRAWAL S, et al. Prior guided GAN based semantic inpainting[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2020: 13693-13702.
[8]	YI Z L, TANG Q, AZIZI S, et al. Contextual residual aggregation for ultra high-resolution image inpainting[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2020: 7505-7514.
[9]	ZHAO L, MO Q H, LIN S H, et al. UCTGAN: diverse image inpainting based on unsupervised cross-space translation[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2020: 5740-5749.
[10]	ZHOU T, DING C X, LIN S W, et al. Learning oracle attention for high-fidelity face completion[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2020: 7677-7686.
[11]	SUVOROV R, LOGACHEVA E, MASHIKHIN A, et al. Resolution-robust large mask inpainting with Fourier convolutions[C]//Proceedings of the 2022 IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE Press, 2022: 3172-3182.
[12]	LI W B, LIN Z, ZHOU K, et al. MAT: mask-aware transformer for large hole image inpainting[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2022: 10748-10758.
[13]	陈春. 中国石油新一代测井软件CIFLog[J]. 石油勘探与开发, 2011, 38(3): 281. CHEN C. China Petroleum’s next-generation well logging software CIFLog[J]. Petroleum Exploration and Development, 2011, 38(3): 281(in Chinese).
[14]	斯伦贝谢. Techlog 2016九大测井解释新技术[J]. 国外测井技术, 2018, 39(1): 71-73. Schlumberger. Techlog 2016 nine new logging interpretation technologies[J]. World Well Logging Technology, 2018, 39(1): 71-73(in Chinese).
[15]	ZHANG T F, SWITZER P, JOURNEL A. Filter-based classification of training image patterns for spatial simulation[J]. Mathematical Geology, 2006, 38(1): 63-80.
[16]	孙建孟, 赵建鹏, 赖富强, 等. 电测井图像空白条带填充方法[J]. 测井技术, 2011, 35(6): 532-537. SUN J M, ZHAO J P, LAI F Q, et al. Methods to fill in the gaps between pads of electrical logging images[J]. Well Logging Technology, 2011, 35(6): 532-537(in Chinese).
[17]	张翔, 张猛, 肖小玲, 等. 复杂地层情况下全井周电成像图像修复方法[J]. 石油物探, 2018, 57(1): 148-153. ZHANG X, ZHANG M, XIAO X L, et al. Image inpainting for fullbore electrical imaging logging in complex formations[J]. Geophysical Prospecting for Petroleum, 2018, 57(1): 148-153(in Chinese).
[18]	罗歆, 闫建平, 王敏, 等. FMI测井图像井壁复原方法优化及应用[J]. 测井技术, 2021, 45(4): 386-393. LUO X, YAN J P, WANG M, et al. Optimization and application of borehole wall restoration method of FMI logging image[J]. Well Logging Technology, 2021, 45(4): 386-393(in Chinese).
[19]	陈长胜, 袁瑞, 王超, 等. 基于图像分解的微电阻率成像测井图像修复方法[J]. 长江大学学报(自然科学版), 2019, 16(1): 95-99. CHEN C S, YUAN R, WANG C, et al. Image restoration method for micro-resistivity imaging logging based on image decomposition[J]. Journal of Yangtze University (Natural Science Edition), 2019, 16(1): 95-99.
[20]	李振苓, 沈金松, 李思, 等. 成像测井电导率图像空白带奇异谱插值和缝洞孔隙度分离方法[J]. 测井技术, 2017, 41(1): 33-40. LI Z L, SHEN J S, LI S, et al. Singular spectral interpolation of blank strips in formation micro-scanner conductivity image and separation of porosity between fracture and Karst cave[J]. Well Logging Technology, 2017, 41(1): 33-40(in Chinese).
[21]	WANG Z F, GAO N, ZENG R, et al. A gaps filling method for electrical logging images based on a deep learning model[J]. Well Logging Technology, 2019, 43(6): 578-582.
[22]	袁晓涛, 马旭成, 肖仕军, 等. 融合多层语义特征的测井电成像空白条带填充深度神经网络方法[J]. 西安石油大学学报(自然科学版), 2022, 37(6): 133-139. YUAN X T, MA X C, XIAO S J, et al. A deep neural network method for filling blank strips in electrical logging images based on fusion of multi-level semantic features[J]. Journal of Xi’an Shiyou University (Natural Science Edition), 2022, 37(6): 133-139(in Chinese).
[23]	贾文玉, 田素月, 孙耀庭, 成像测井技术与应用[M]. 北京: 石油工业出版社, 2000. JIA W Y, TIAN S Y, SUN Y T. Imaging logging technology and its application[M]. Beijing: Petroleum Industry Press, 2000(in Chinese).
[24]	ISOLA P, ZHU J Y, ZHOU T H, et al. Image-to-image translation with conditional adversarial networks[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2017: 5967-5976.
[25]	WANG T C, LIU M Y, ZHU J Y, et al. High-resolution image synthesis and semantic manipulation with conditional GANs[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2018: 8798-8807.
[26]	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2016: 770-778.