Intelligent detection of “bead-shaped” abnormal reflections in carbonate reservoir caves based on Yolox algorithm
ZHANG Ao1,2, LI Zongjie3, LIU Jun3, YAN Xingyu1,2, LI Wei3, GU Hanming1,2
1. School of Geophysics & Geomatics, China University of Geosciences(Wuhan), Wuhan, Hubei 430074, China; 2. Hubei Subsurface Multiscale Image Key Laboratory, Wuhan, Hubei 430074, China; 3. Research Institute of Exploration and Development, NorthwestOilfield Branch Co., SINOPEC, Urumqi, Xinjiang 830011, China
Abstract:The traditional identification method of carbonate reservoir caves is mainly based on the analysis of seismic reflection characteristics, which has high data requirements, weak universality, and low efficiency and is affected by subjective factors. The research objects of the methods using the feature extraction capability of convo-lutional neural networks (CNNs) to identify geological structures are mainly large-scale structures such as salt domes, faults, and strata, but they can easily misjudge small-scale structures such as caves. Due to the different reflection characteristics of different scales of caves on the seismic profile, there is a certain mapping relationship between caves and the "bead shape." Therefore, a relatively large-scale "bead shape" can be first identified on the seismic profile, and then the cave can be identified according to the mapping relationship between "bead shape" and caves. Therefore, the network structure of the Yolox-based "bead-shaped" object detection model is proposed, which mainly includes the feature extraction module, feature enhancement module, and Decoupling Head module. Effective features of different scales are obtained by feature extraction after a seismic profile is input,and then the feature enhancement network is input for multi-scale feature fusion. Finally, the information on the detection frame is obtained by the Decoupling Head. The position of the "bead-shaped" anomalous reflection boundary is obtained after decoding, and the detection frame is output. The test results of synthetic seismic data and actual seismic data show that ① the traditional method based on amplitude can hardly identify the "bead shape" when it is at and near a strong event, and the high-value part cannot reflect the actual range of the "bead shape." ② The recognition results of the U-Net model reflect the position of the "bead shape" but cannot obtain the boundary coordinates, which is prone to consider two close "bead shapes" as one. Hence, it has low identification accuracy and is exposed to errors or misses. ③ The detection frame in the identification result of the Yolox model reflects the position and actual size of the "bead shape", and its specific coordinates can be obtained. Therefore, the identification effect of the Yolox model is better than that of traditional methods and U-Net models.
李胜军,王福焕,雍学善,等.塔中地区碳酸盐岩溶洞地震响应机理及影响因素[J].新疆石油地质,2011,32(3):305-307.LI Shengjun,WANG Fuhuan,YONG Xueshan,et al.Seismic response mechanism of carbonate caverns and influence factors in Tazhong area[J].Xinjiang Petroleum Geology,2011,32(3): 305-307.
[3]
马吉静.高密度电阻率法的异常识别和推断——以溶洞探测和寻找地下水为例[J].地球物理学进展,2019,34(4):1489-1498.MA Jijing.Anomaly identification and inference of high density resistivity method:take karst cave exploration and groundwater exploration as an example[J].Progress in Geophysics,2019,34(4):1489-1498.
[4]
胡华锋,鲍典,马灵伟,等.基于高频衰减梯度的碳酸盐岩溶洞储集体规模识别方法——以塔河油田岩溶古河道溶洞储集体为例[J].石油物探,2018,57(6): 892-901.HU Huafeng,BAO Dian,MA Lingwei,et al.Scale characterization of carbonate karst cave reservoirs based on high-frequency attenuation gradient:A case study from karst paleochannels reservoirs in Tahe oilfield,China[J]. Geophysical Prospecting for Petroleum,2018,57(6):892-901.
[5]
朱博华,杨江峰,吕秋玲,等.基于分频成像的小尺度溶洞检测[J].石油地球物理勘探,2018,53(3):538-544.ZHU Bohua,YANG Jiangfeng,LYU Qiuling,et al.Small-scale cave detection based on frequency-segmented imaging[J].Oil Geophysical Prospecting,2018,53(3):538-544.
[6]
SHI Y,WU X,FOMEL S.Automatic salt-body classification using deep-convolutional neural network[C].SEG Technical Program Expanded Abstracts,2018,37:1971-1975.
[7]
WU X,LIANG L,SHI Y,et al. FaultSeg3D:Using synthetic data sets to train an end-to-end convolutional neural network for 3D seismic fault segmentation[J].Geophysics,2019,84(3):IM35-IM45.
[8]
Mukhopadhyay P,Mallick S.Bayesian deep learning for seismic facies classification and its uncertainty estimation [C].SEG Technical Program Expanded Abstracts,2019,38: 2488-2492.
[9]
陈康,狄贵东,张佳佳,等.基于改进U-Net卷积神经网络的储层预测[J].CT理论与应用研究,2021,30(4):403-415.CHEN Kang,DI Guidong,ZHANG Jiajia,et al.Reser- voir prediction based on improved U-Net convolutional neural network[J].CT Theory and Applications,2021,30(4):403-415.
[10]
张政,严哲,顾汉明.基于残差网络与迁移学习的断层自动识别[J]. 石油地球物理勘探,2020,55(5):950-956.ZHANG Zheng,YAN Zhe,GU Hanming.Automatic fault recognition with residual network and transfer learning[J]. Oil Geophysical Prospecting,2020,55(5):950-956.
[11]
王子健,伍新明,杜玉山,等.基于深度学习的地震断层检测与断面组合[J].油气地质与采收率,2022,29(1):69-79.WANG Zijian,WU Xinming,DU Yushan,et al.Deep learning-based seismic fault detection and surface combination[J].Petroleum Geology and Recovery Efficiency,2022,29(1): 69-79
[12]
闫星宇,李宗杰,顾汉明,等.基于深度卷积神经网络的地震数据溶洞识别[J].石油地球物理勘探,2022,57(1):1-11.YAN Xingyu,LI Zongjie,GU Hanming,et al.Identification of Karst caves in seismic data based on deep convolutional neural network[J].Oil Geophysical Prospecting,2022,57(1): 1-11.
[13]
王立华,魏建新,狄帮让.溶洞物理模型地震响应及其属性分析[J].石油地球物理勘探,2008,43(3): 291-296.WANG Lihua,WEI Jianxin,DI Bangrang.Seismic response of karst cave physical model and analysis of its attributes[J]. Oil Geophysical Prospecting,2008,43(3):291-296.
REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once: Unified, real-time object detection[C].2016 IEEE Conference on Computer Vision and Pattern Recognition, 2016,779-788.
[16]
GE Z,LIU S,WANG F,et al.Yolox: Exceeding yolo series in 2021[EB/OL].arXiv:2107.08430,2021.https://ui.adsabs.harvard.edu/abs/2021arXiv210708430G/abstract.
[17]
于娟,罗舜.基于YOLOv5的违章建筑检测方法[J].计算机工程与应用,2021,57(20):236-244.YU Juan,LUO Shun. Detection method of illegal building based on YOLOv5[J].Computer Engineering and Applications,2021,57(20): 236-244.
[18]
蒋昕昊,蔡伟,杨志勇,等.基于YOLO-IDSTD算法的红外弱小目标检测[J].红外与激光工程,2022,51(3): 502-511.JIANG Xinhao,CAI Wei,YANG Zhiyong,et al.Infrared dim and small target detection based on YOLO-IDSTD algorithm[J].Infrared and Laser Engineering,2022,51(3): 502-511.
[19]
HE K,ZHANG X,REN S,et al.Deep residual lear-ning for image recognition[C].2016 IEEE Conference on Computer Vision and Pattern Recognition,2016,770-778.
[20]
WANG C,LIAO H M,WU Y,et al.CSPNet: A new backbone that can enhance learning capability of CNN[C].2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops,2020,doi: 10.1109/CVPRW50498.2020.00203.
[21]
HE K,ZHANG X,REN S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(9):1904-1916.
[22]
LIN T,DOLLÁR P,GIRSHICK R,et al.Feature py-ramid networks for object detection[C].2017 IEEE Conference on Computer Vision and Pattern Recognition,2017,doi:10.1109/CVPR.2017.106.
[23]
GE Z,LIU S,LI Z,et al.Ota:Optimal transport assignment for object detection[C].2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition,2021,doi:10.1109/CVPR46437.2021.00037.
[24]
KINGMA D P,BA J L.Adam:A method for stochastic optimization[C].ICLR 2015,2015,doi:10.48550/arXiv.1412.6980.
[25]
魏明军,周太宇,纪占林,等.基于Mask-YOLO的复杂场景口罩佩戴检测[J].应用科学学报,2022,40(1):93-104.WEI Mingjun,ZHOU Taiyu,JI Zhanlin,et al.Mask wearing detection in complex scenes based on Mask-YOLO[J].Journal of Applied Sciences,2022,40(1):93-104.