[1]DU Yanling,WANG Lili,HUANG Dongmei,et al.Improved R2CNN ocean eddy automatic detection with a dense feature pyramid[J].CAAI Transactions on Intelligent Systems,2023,18(2):341-351.[doi:10.11992/tis.202112019]
Copy

Improved R2CNN ocean eddy automatic detection with a dense feature pyramid

CAAI Transactions on Intelligent Systems[ISSN 1673-4785/CN 23-1538/TP] Volume: 18 Number of periods: 2023 2 Page number: 341-351 Column: 学术论文—机器感知与模式识别 Public date: 2023-05-05
Title:
Improved R2CNN ocean eddy automatic detection with a dense feature pyramid
Author(s):
DU Yanling1;  WANG Lili1;  HUANG Dongmei1; 2;  CHEN Ke3;  HE Qi1
1. College of Information Technology, Shanghai Ocean University, Shanghai 201306, China;
2. College of Electronical and Information Engineering, Shanghai University of Electric Power, Shanghai 200090, China;
3. East China Sea Forecast Center, Ministry of Natural Resources, Shanghai 200136, China
Keywords:
deep learning; object detection; ocean eddy; dense feature pyramid network; convolution neural network; feature fusion; R2CNN; pattern recognition
CLC:
TP183; TP391
DOI:
10.11992/tis.202112019
Abstract:
The lack of understanding of the evolution process of ocean eddies is the main obstacle to current physical ocean research. To gain a better understanding of the mechanism of its generation, development, and variation process, as well as its interaction with multi-scale ocean processes, the automatic detection of ocean eddies is crucial. However, existing horizontal detection methods struggle to accurately detect ocean eddies due to their scale diversity, irregular shape, and dense distribution, leading to significant redundancy and overlap in the detection area. To address this issue, we propose a multi-scale rotation-dense feature pyramid network. The feature pyramid network is enhanced through dense connection (DFPN) for multi-scale high-level semantic feature extraction and fusion, improving feature propagation and reuse. To accommodate the dense distribution of ocean eddies, the R2CNN network is improved, and a multi-scale RoI Align mechanism is introduced to preserve feature semantics and spatial information integrity, thus improving model detection performance. We construct the ocean eddy dataset using sea-level anomaly data, which is preprocessed into VOC format for training. The parameters are adjusted to obtain the optimal detection model, and experimental results show that the model’s detection accuracy can reach 96.4%. The proposed model demonstrates good generalisability through the automatic detection of ocean eddies in the Pacific and Atlantic Oceans.
References:
[1] 张盟, 杨玉婷, 孙鑫, 等. 基于深度卷积网络的海洋涡旋检测模型[J]. 南京航空航天大学学报, 2020, 52(5): 708–713
ZHANG Meng, YANG Yuting, SUN Xin, et al. Ocean eddy detection model based on deep convolution neural network[J]. Journal of Nanjing University of aeronautics & astronautics, 2020, 52(5): 708–713
[2] 沈飙, 陈扬, 杨琛, 等. 海洋科学中尺度涡的计算机视觉检测和分析方法[J]. 数据与计算发展前沿, 2020, 2(6): 30–41
SHEN Biao, CHEN Yang, YANG Chen, et al. Computer vision detection and analysis of mesoscale eddies in marine science[J]. Frontiers of data & computing, 2020, 2(6): 30–41
[3] 董昌明, 蒋星亮, 徐广珺, 等. 海洋涡旋自动探测几何方法、涡旋数据库及其应用[J]. 海洋科学进展, 2017, 35(4): 439–453
DONG Changming, JIANG Xingliang, XU Guangjun, et al. Automated eddy detection using geometric approach, eddy datasets and their application[J]. Advances in marine science, 2017, 35(4): 439–453
[4] NENCIOLI F, DONG Changming, DICKEY T, et al. A vector geometry-based eddy detection algorithm and its application to a high-resolution numerical model product and high-frequency radar surface velocities in the southern California bight[J]. Journal of atmospheric and oceanic technology, 2010, 27(3): 564–579.
[5] CHELTON D B, SCHLAX M G, SAMELSON R M. Global observations of nonlinear mesoscale eddies[J]. Progress in oceanography, 2011, 91(2): 167–216.
[6] ALPERS W, BRANDT P, LAZAR A, et al. A small-scale oceanic eddy off the coast of West Africa studied by multi-sensor satellite and surface drifter data[J]. Remote sensing of environment, 2013, 129: 132–143.
[7] FAGHMOUS J H, CHAMBER Y, BORIAH S, et al. A novel and scalable spatio-temporal technique for ocean eddy monitoring[C]//AAAI’12: Proceedings of the Twenty-Sixth AAAI Conference on Artificial Intelligence. New York: ACM, 2012: 281?287.
[8] ASHKEZARI M D, HILL C N, FOLLETT C N, et al. Oceanic eddy detection and lifetime forecast using machine learning methods[J]. Geophysical research letters, 2016, 43(23): 12234–12241.
[9] KARIMOVA S. Spiral eddies in the Baltic, Black and Caspian Seas as seen by satellite radar data[J]. Advances in space research, 2012, 50(8): 1107–1124.
[10] ISERN-FONTANET J, GARCíA-LADONA E, FONT J. Identification of marine eddies from altimetric maps[J]. Journal of atmospheric and oceanic technology, 2003, 20(5): 772–778.
[11] YI J, DU Y, HE Z, et al. Enhancing the accuracy of automatic eddy detection and the capability of recognizing the multi-core structures from maps of sea level anomaly[J]. Ocean science, 2014, 10(1): 39–48.
[12] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770?778.
[13] LIN T Y, DOLLáR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 936?944.
[14] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 2999?3007.
[15] REN Shaoqing, HE Kaiming, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(6): 1137–1149.
[16] DUO Zijun, WANG Wenke, WANG Huizan. Oceanic mesoscale eddy detection method based on deep learning[J]. Remote sensing, 2019, 11(16): 1921.
[17] JIANG Yingying, ZHU Xiangyu, WANG Xiaobing, et al. R2CNN: rotational region CNN for orientation robust scene text detection[EB/OL]. (2017?06?30)[2020?01?01].https://arxiv.org/abs/1706.09579.
[18] YANG Xue, SUN Hao, FU K, et al. Automatic ship detection in remote sensing images from google earth of complex scenes based on multiscale rotation dense feature pyramid networks[J]. Remote sens, 2018, 10: 132.
[19] SIMONYAN K , ? ZISSERMAN A . Very deep convolutional networks for large-scale image recognition[EB/OL]. (2014?10?04)[2020?01?01].https://www.semanticscholar.org/paper/Very-Deep-Convolutional-Networks-for-Large-Scale-Simonyan-Zisserman/eb42cf88027de515750f230b23b1a057dc782108.
[20] 单志勇, 张鐘月. 基于改进Faster R-CNN算法的行人检测[J]. 现代计算机, 2021, 27(23): 124–128
SHAN Zhiyong, ZHANG Zhongyue. Pedestrian detection based on improved faster R-CNN algorithm[J]. Modern computer, 2021, 27(23): 124–128
[21] MA Jianqi, SHAO Weiyuan, YE Hao, et al. Arbitrary-oriented scene text detection via rotation proposals[J]. IEEE transactions on multimedia, 2018, 20(11): 3111–3122.
[22] 戴媛, 易本顺, 肖进胜, 等. 基于改进旋转区域生成网络的遥感图像目标检测[J]. 光学学报, 2020, 40(1): 270–280
DAI Yuan, YI Benshun, XIAO Jinsheng, et al. Object detection of remote sensing image based on improved rotation region proposal network[J]. Acta optica sinica, 2020, 40(1): 270–280
[23] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus: IEEE, 2014: 580?587.
[24] HE Kaiming, GKIOXARI G, DOLLáR P, et al. Mask R-CNN[C]//IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 2980?2988.
[25] 李广帅, 苏娟, 李义红. 基于改进Faster R-CNN的SAR图像飞机检测算法[J]. 北京航空航天大学学报, 2021, 47(1): 159–168
LI Guangshuai, SU Juan, LI Yihong. An aircraft detection algorithm in SAR image based on improved Faster R-CNN[J]. Journal of Beijing University of aeronautics and astronautics, 2021, 47(1): 159–168
Similar References:

Memo

-

Last Update: 1900-01-01

Copyright © CAAI Transactions on Intelligent Systems