设为首页
加入收藏
Boundary Layer Convergence Line Identification Algorithm for Weather Radar Based on R2CNN
-
摘要: 边界层辐合线是触发对流的中尺度天气系统之一, 边界层辐合线的精细化识别对于揭示其形成、演变及与其他系统相互作用机制至关重要。目前自动识别技术在适应边界层辐合线多样性(如尺度、强度和形状)方面存在局限。旋转区域卷积神经网络(R2CNN)可提高识别准确性、鲁棒性和泛化能力。综合考虑天气雷达型号和分辨率的多样性, 针对性构建识别数据集用于模型训练, 调整相应参数得到识别模型, 并利用交并比和置信度评估检验识别效果。结果表明:基于R2CNN的边界层辐合线识别算法在使用较低交并比阈值时命中率更高且空报率更低, 当置信度为0.7时, TS(threat score)评分最高。与现有的阵风锋识别算法(Machine Intelligence Gust Front Algorithm, MIGFA)效果相比, R2CNN在减少误报、提升命中率及平衡识别频率等关键性能方面优势显著, 适用于业务应用与推广。Abstract: Boundary layer convergence lines are recognized as one of the critical mesoscale weather systems triggered convection, and also affect low-altitude flight safety. The accurate and detailed identification of these lines is considered essential for revealing their formation, evolution, and interaction mechanisms with other weather systems. However, existing automatic identification technologies are limited in their ability to adapt the diverse characteristics of these lines, such as scale, intensity, and shape. The rotational region-based convolutional neural network (R2CNN) is employed to enhance the accuracy, robustness, and generalization of the identification process. A comprehensive identification dataset has been constructed for model training, considering the diversity of weather radar models and resolutions. Relevant parameters are adjusted to derive the optimized recognition model. The intersection over union (IoU) with confidence levels are employed to comprehensively assess and validate the identification results. Results indicate that the boundary layer convergence line recognition algorithm developed achieves a higher hit rate and a lower false alarm rate at lower IoU thresholds. At a confidence level of 0.7, the threat score (TS) reaches its maximum value.Compared to the existing Machine Intelligence Gust Front Algorithm (MIGFA), the model proposed in this study demonstrates significant advantages in reducing false alarms, improving hit rates, and achieving a balanced recognition frequency. Therefore, it is more suitable for operational applications and dissemination. This research not only provides a more effective method for identifying boundary layer convergence lines but also contributes to the improvement of low-altitude flight safety and advances meteorological detection technologies. The proposed method addresses limitations of existing technologies by effectively managing the diverse characteristics of boundary layer convergence lines. By incorporating rotational bounding boxes in the detection process, R2CNN model enhances the detection accuracy for objects with arbitrary orientations, which is particularly beneficial for meteorological phenomena that do not align with the standard axis. The constructed dataset includes a diverse collection of radar images from various models and resolutions, ensuring that the model is trained on a wide range of data and can generalize effectively to new, unseen data. Extensive experiments are conducted to evaluate the model's performance under different IoU thresholds and confidence levels. Findings demonstrate that at lower IoU thresholds, the model maintains high detection performance, indicating its robustness in practical applications where precise localization may be challenging. Furthermore, the superior performance of the proposed model compared to MIGFA indicates its potential for widespread adoption by meteorological agencies for better monitoring and forecasting.
-
图 1 边界层辐合线在测试集的识别效果
Fig. 1 Performance of boundary layer convergence line identification based on testing dataset
图 2 2017年8月4日10:34宜昌S波段天气雷达阵风锋观测特征与识别结果
(黑色框表示算法识别的阵风锋落区)
Fig. 2 Observation and identified gust front characteristics by S-band weather radar of Yichang at 1034 UTC 4 Aug 2017
(the black box denotes gust front region identified by algorithm)
图 3 2022年7月8日07:40南京六合X波段天气雷达边界层辐合线观测与识别结果
(黑色框表示算法识别的边界层辐合线落区)
Fig. 3 Observation and identified boundary layer convergence line by X-band weather radar of Luhe, Nanjing at 0740 UTC 8 Jul 2022
(the black box denotes boundary layer convergence line region identified by algorithm )
图 4 R2CNN和MIGFA方法识别阵风锋效果对比
Fig. 4 Comparative evaluation of gust front identification performance between R2CNN and MIGFA methods
图 5 2021年4月30日10:49宿迁天气雷达阵风锋识别效果对比
(黑色框表示R2CNN算法识别结果，橙色框表示MIGFA识别结果)
Fig. 5 Comparative evaluation of gust front identification by weather radar of Suqian at 1049 UTC 30 Apr 2021
(the black box denotes R2CNN algorithm, and the orange box denotes MIGFA algorithm)
-
[1] Doswell C A.Severe Convective Storms-An Overview.Severe Convective Storms.Boston:American Meteorological Society, 2001. [2] 俞小鼎, 周小刚, 王秀明. 雷暴与强对流临近天气预报技术进展. 气象学报, 2012, 70(3): 311-337. doi: 10.3969/j.issn.1004-4965.2012.03.003Yu X D, Zhou X G, Wang X M. Progress in forecast technology of thunderstorm and strong convection near weather. Acta Meteor Sinica, 2012, 70(3): 311-337. doi: 10.3969/j.issn.1004-4965.2012.03.003 [3] 郑永光, 周康辉, 盛杰, 等. 强对流天气监测预报预警技术进展. 应用气象学报, 2015, 26(6): 641-657.Zheng Y G, Zhou K H, Sheng J, et al. Progress in monitoring, forecasting and early warning technology of severe convective weather. J Appl Meteor Sci, 2015, 26(6): 641-657. [4] 孙继松, 戴建华, 何立富, 等. 强对流天气预报的基本原理与技术方法//中国强对流天气预报手册. 北京: 气象出版社, 2014: 208-209.Sun J S, Dai J H, He L F, et al. Basic Principles and Technical Methods of Severe Convective Weather Forecast//Manual of Severe Convective Weather Forecast in China. Beijing: China Meteorological Press, 2014: 208-209. [5] 苏爱芳, 郑永光, 张宁, 等. 边界层辐合线触发深厚湿对流研究进展. 气象学报, 2022, 80(2): 177-189.Su A F, Zheng Y G, Zhang N, et al. A review of research on boundary convergence lines triggering of deep and moist convection. Acta Meteor Sinica, 2022, 80(2): 177-189. [6] Doswell C A. The distinction between large-scale and mesoscale contribution to severe convection: A case study example. Wea Forecasting, 1987, 2(1): 3-16. doi: 10.1175/1520-0434(1987)002<0003:TDBLSA>2.0.CO;2 [7] Wilson J W, Schreiber W E. Initiation of convective storms at radar-observed boundary-layer convergence lines. Mon Wea Rev, 1986, 114(12): 2516-2536. doi: 10.1175/1520-0493(1986)114<2516:IOCSAR>2.0.CO;2 [8] Huang Y P, Meng Z Y, Li W B, et al. General features of radar-observed boundary layer convergence lines and their associated convection over a sharp vegetation-contrast area. Geophys Res Lett, 2019, 46(5): 2865-2873. doi: 10.1029/2018GL081714 [9] 刁秀广, 车军辉, 李静, 等. 边界层辐合线在局地强风暴临近预警中的应用. 气象, 2009, 35(2): 29-33.Diao X G, Che J H, Li J, et al. Application of boundary convergence line in nowcasting warning of severe convective storm. Meteor Mon, 2009, 35(2): 29-33. [10] Liu H J, Meng Z Y, Zhu Y N, et al. Convection initiation associated with a boundary layer convergence line over a real-world sharp vegetation-contrast area. Mon Wea Rev, 2023, 151(5): 1189-1212. doi: 10.1175/MWR-D-22-0083.1 [11] Mueller C K, Carbone R E. Dynamics of a thunderstorm outflow. J Atmos Sci, 1987, 44(15): 1879-1898. doi: 10.1175/1520-0469(1987)044<1879:DOATO>2.0.CO;2 [12] Harrison S J, Mecikalski J R, Knupp K R. Analysis of outflow boundary collisions in north-central Alabama. Wea Forecasting, 2009, 24(6): 1680-1690. doi: 10.1175/2009WAF2222268.1 [13] 张曦, 黄兴友, 刘新安, 等. 北京大兴国际机场相控阵雷达强对流天气监测. 应用气象学报, 2022, 33(2): 192-204.Zhang X, Huang X Y, Liu X A, et al. The hazardous convective storm monitoring of phased-array antenna radar at Daxing International Airport of Beijing. J Appl Meteor Sci, 2022, 33(2): 192-204. [14] 郑永光, 张春喜, 陈炯, 等. 用NCEP资料分析华北暖季对流性天气的气候背景. 北京大学学报(自然科学版), 2007, 43(5): 600-608. doi: 10.3321/j.issn:0479-8023.2007.05.003Zheng Y G, Zhang C X, Chen J, et al. Climatic background of warm-season convective weather in North China based on the NCEP analysis. Acta Scientiarum Naturalium Universitatis Pekinensis(Nat Ed), 2007, 43(5): 600-608. doi: 10.3321/j.issn:0479-8023.2007.05.003 [15] 方祖亮, 俞小鼎, 王秀明. 东北暖季干线统计分析. 气象学报, 2020, 78(2): 260-276.Fang Z L, Yu X D, Wang X M. Statistical analysis of drylines in Northeast China. Acta Meteor Sinica, 2020, 78(2): 260-276. [16] 俞小鼎, 王秀明, 李万莉, 等. 雷暴与强对流临近预报. 北京: 气象出版社, 2020.Yu X D, Wang X M, Li W L, et al. Thunderstorm and Severe Convection Nowcasting. Beijing: China Meteorological Press, 2020. [17] Hughes C P, Veron D E. A characterization of the Delaware sea breeze using observations and modeling. J Appl Meteor Climatol, 2018, 57(7): 1405-1421. doi: 10.1175/JAMC-D-17-0186.1 [18] 王彦, 李胜山, 郭立, 等. 渤海湾海风锋雷达回波特征分析. 气象, 2006, 32(12): 23-27.Wang Y, Li S S, Guo L, et al. Doppler radar echo features of sea breeze front in Bohai Bay. Meteor Mon, 2006, 32(12): 23-27. [19] LeMone M A. The structure and dynamics of horizontal roll vortices in the planetary boundary layer. J Atmos Sci, 1973, 30(6): 1077-1091. doi: 10.1175/1520-0469(1973)030<1077:TSADOH>2.0.CO;2 [20] 郭飞燕, 刁秀广, 褚颖佳, 等. 弱垂直风切变环境下强下击暴流双偏振雷达特征. 应用气象学报, 2023, 34(6): 681-693.Guo F Y, Diao X G, Chu Y J, et al. Dual polarization radar characteristics of severe downburst occurred in weak vertical wind shear. J Appl Meteor Sci, 2023, 34(6): 681-693. [21] 李劲, 顾松山. 2009年6月5日安徽致灾大风天气过程分析. 安徽农业科学, 2010, 38(14): 7443-7445. doi: 10.3969/j.issn.0517-6611.2010.14.117Li J, Gu S S. Analysis of a disaster-causing wind process in Anhui on June 5th, 2009. J Anhui Agric Sci, 2010, 38(14): 7443-7445. doi: 10.3969/j.issn.0517-6611.2010.14.117 [22] 何娜, 丁青兰, 俞小鼎, 等. 北京及周边地区雷暴阵风锋特征统计分析. 气象学报, 2020, 78(2): 250-259.He N, Ding Q L, Yu X D, et al. Statistical analysis of thunderstorm gust front characteristics in Beijing and surrounding areas. Acta Meteor Sinica, 2020, 78(2): 250-259. [23] 徐芬, 杨吉, 郑媛媛, 等. MIGFA阵风锋识别算法改进与检验. 气象, 2016, 42(1): 44-53.Xu F, Yang J, Zheng Y Y, et al. Improvement of the MIGFA technique for identifying gust front and its verification. Meteor Mon, 2016, 42(1): 44-53. [24] Hermes L G, Witt A, Smith S D, et al. The gust-front detection and wind-shift algorithms for the terminal Doppler weather radar system. J Atmos Oceanic Technol, 1993, 10(5): 693-709. doi: 10.1175/1520-0426(1993)010<0693:TGFDAW>2.0.CO;2 [25] Yuan Y, Wang P, Wang D, et al. An algorithm for automated identification of gust fronts from Doppler radar data. J Meteor Res, 2018, 32(3): 444-455. doi: 10.1007/s13351-018-7089-7 [26] 何立富, 陈涛, 孔期. 华南暖区暴雨研究进展. 应用气象学报, 2016, 27(5): 559-569.He L F, Chen T, Kong Q. A review of studies on prefrontal torrential rain in South China. J Appl Meteor Sci, 2016, 27(5): 559-569. [27] 吴海英, 孙继松, 慕瑞琪, 等. 一次强对流过程中两种不同类型风暴导致的极端对流大风分析. 气象学报, 2023, 81(2): 205-217.Wu H Y, Sun J S, Mu R Q, et al. Analysis of extreme convective gusts caused by two types of weather storms during a strong convection event. Acta Meteor Sinica, 2023, 81(2): 205-217. [28] Uyeda H, Zrnić D S. Automatic detection of gust fronts. J Atmos Oceanic Technol, 1986, 3(1): 36-50. doi: 10.1175/1520-0426(1986)003<0036:ADOGF>2.0.CO;2 [29] Klingle D L, Smith D R, Wolfson M M. Gust front characteristics as detected by Doppler radar. Mon Wea Rev, 1987, 115(5): 905-918. doi: 10.1175/1520-0493(1987)115<0905:GFCADB>2.0.CO;2 [30] Smalley D J, Bennett B J, Frankel R. MIGFA: The Machine Intelligent Gust Front Algorithm for NEXRAD. 32nd Conference on Radar Meteorology, 2005. [31] Mueller C, Saxen T, Roberts R, et al. NCAR auto-nowcast system. Wea Forecasting, 2003, 18(4): 545-561. doi: 10.1175/1520-0434(2003)018<0545:NAS>2.0.CO;2 [32] Hwang Y, Yu T Y, Lakshmanan V, et al. Neuro-fuzzy gust front detection algorithm with S-band polarimetric radar. IEEE Trans Geosci Remote Sens, 2017, 55(3): 1618-1628. doi: 10.1109/TGRS.2016.2628520 [33] 王楠, 刘黎平, 徐宝祥, 等. 利用多普勒雷达资料识别低空风切变和辐合线方法研究. 应用气象学报, 2007, 18(3): 314-320.Wang N, Liu L P, Xu B X, et al. Recognizing low-altitude wind shear and convergence line with Doppler radar. J Appl Meteor Sci, 2007, 18(3): 314-320. [34] 韦林祯. 基于多普勒雷达的边界层辐合线的识别. 天津: 天津大学, 2010.Wei L Z. Automatic Recognition of Boundary Convergent Lines Based on Doppler Radar. Tianjin: Tianjin University, 2010. [35] 冯卫家. 基于边界层辐合线的强对流天气预测. 天津: 天津大学, 2012.Feng W J. Convective Weather Forecast Based on Boundary Layer Convergence Lines. Tianjin: Tianjin University, 2012. [36] 郑佳锋. 新一代天气雷达阵风锋识别算法研究. 成都: 成都信息工程学院, 2013.Zheng J F. An Algorithm Research of Gust Front Identification for CINRAD. Chengdu: Chengdu University of Information Technology, 2013. [37] 涂朝勇. 西安咸阳机场阵风锋自动识别及其应用研究. 成都: 成都信息工程大学, 2018.Tu C Y. Research on Automatic Identification of Gust Fronts at Xi'an Xianyang Airport and its Application. Chengdu: Chengdu University of Information Technology, 2018. [38] 冷亮, 肖艳姣, 吴涛. 基于数学形态学的阵风锋识别算法. 气象科技, 2016, 44(1): 1-6.Leng L, Xiao Y J, Wu T. Automatic recognition of gust fronts based on mathematical morphology. Meteor Sci Technol, 2016, 44(1): 1-6. [39] 王一童, 王秀明, 俞小鼎. 产生致灾大风的超级单体回波特征. 应用气象学报, 2022, 33(2): 180-191.Wang Y T, Wang X M, Yu X D. Radar characteristics of straight-line damaging wind producing supercell storms. J Appl Meteor Sci, 2022, 33(2): 180-191. [40] 张林, 李峰, 吴蕾, 等. CINRAD/SAD双偏振雷达非降水回波识别技术. 应用气象学报, 2022, 33(6): 724-735.Zhang L, Li F, Wu L, et al. Non-precipitation identification technique for CINRAD/SAD dual polarimetric weather radar. J Appl Meteor Sci, 2022, 33(6): 724-735. [41] 徐玥, 邵美荣, 唐凯, 等. 2021年黑龙江两次超级单体龙卷过程多尺度特征. 应用气象学报, 2022, 33(3): 305-318.Xu Y, Shao M R, Tang K, et al. Multiscale characteristics of two supercell tornados of Heilongjiang in 2021. J Appl Meteor Sci, 2022, 33(3): 305-318. [42] 白华, 袁潮, 潘晓, 等. 辽宁省温带气旋龙卷的环境参数特征. 应用气象学报, 2023, 34(1): 104-116.Bai H, Yuan C, Pan X, et al. Environmental characteristics of extratropical cyclone tornadoes in Liaoning. J Appl Meteor Sci, 2023, 34(1): 104-116. [43] 郑佳锋, 张杰, 朱克云, 等. 阵风锋自动识别与预警. 应用气象学报, 2013, 24(1): 117-125.Zheng J F, Zhang J, Zhu K Y, et al. Automatic identification and alert of gust fronts. J Appl Meteor Sci, 2013, 24(1): 117-125. [44] 郭雪星, 瞿建华, 叶凌梦, 等. 基于朴素贝叶斯的FY-4A/AGRI云检测方法. 应用气象学报, 2023, 34(3): 282-294.Guo X X, Qu J H, Ye L M, et al. FY-4A/AGRI cloud detection method based on naive Bayesian algorithm. J Appl Meteor Sci, 2023, 34(3): 282-294. [45] 李哲, 吴翀, 刘黎平, 等. 双偏振相控阵雷达误差评估与相态识别方法. 应用气象学报, 2022, 33(1): 16-28.Li Z, Wu C, Liu L P, et al. Error evaluation and hydrometeor classification method of dual polarization phased array radar. J Appl Meteor Sci, 2022, 33(1): 16-28. [46] 袁凯, 李武阶, 庞晶. 基于决策树算法的鄂东地区冰雹识别技术. 应用气象学报, 2023, 34(2): 234-245.Yuan K, Li W J, Pang J. Hail identification technology in eastern Hubei based on decision tree algorithm. J Appl Meteor Sci, 2023, 34(2): 234-245. [47] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks. Commun ACM, 2017, 60(6): 84-90. [48] 尹晓燕, 胡志群, 郑佳锋, 等. 利用深度学习填补双偏振雷达回波遮挡. 应用气象学报, 2022, 33(5): 581-593.Yin X Y, Hu Z Q, Zheng J F, et al. Filling in the dual polarization radar echo occlusion based on deep learning. J Appl Meteor Sci, 2022, 33(5): 581-593. [49] 张林, 吴蕾, 李峰, 等. 基于深度学习的天气雷达异常数据识别技术. 应用气象学报, 2023, 34(6): 694-705.Zhang L, Wu L, Li F, et al. Indentification of weather radar abnormal data based on deep learning. J Appl Meteor Sci, 2023, 34(6): 694-705. [50] Girshick R, Donahue J, Darrell T, et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation//IEEE Conference on Computer Vision and Pattern Recognition, 2014: 580-587. [51] Girshick R. Fast R-CNN//IEEE International Conference on Computer Vision(ICCV), 2015: 1440-1448. [52] Ren S Q, He K M, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Trans Pattern Anal Mach Intell, 2017, 39(6): 1137-1149. [53] Lin T Y, Dollár P, Girshick R, et al. Feature Pyramid Networks for Object Detection//IEEE Conference on Computer Vision and Pattern Recognition(CVPR). 2017: 936-944. [54] 徐月飞, 赵放, 毛程燕, 等. 基于深度卷积神经网络的阵风锋识别算法. 暴雨灾害, 2020, 39(1): 81-88.Xu Y F, Zhao F, Mao C Y, et al. Gust front detection algorithm based on deep convolutional neural network. Torrential Rain Disasters, 2020, 39(1): 81-88. [55] Bose S R, Kumar V S. Efficient inception V2 based deep convolutional neural network for real-time hand action recognition. IET Image Process, 2020, 14(4): 688-696. [56] 郎子鑫, 胡金蓉, 罗月梅, 等. Mask RCNN-CBAM: 基于双维度注意力机制的阵风锋自动识别研究. 气候变化研究快报, 2023, 12(3): 576-588.Lang Z X, Hu J R, Luo Y M, et al. Mask RCNN-CBAM: Automatic identification of gust front based on two-dimensional attention mechanism. Climate Change Res Lett, 2023, 12(3): 576-588. [57] Tian H Y, Hu Z Q, Wang F Z, et al. Radar echo recognition of gust front based on deep learning. Remote Sens, 2024, 16(3). DOI: 10.3390/RS16030439. [58] Ronneberger O, Fischer P, Brox T. U-net: Convolutional Networks for Biomedical Image Segmentation//Lecture Notes in Computer Science. Cham: Springer International Publishing, 2015: 234-241. [59] Jiang Y Y, Zhu X Y, Wang X B, et al. R2CNN: Rotational region CNN for orientation robust scene text detection. arXiv, 2017. DOI: 10.48550/arXiv1706.09579. [60] Ma J Q, Shao W Y, Ye H, et al. Arbitrary-oriented scene text detection via rotation proposals. IEEE Trans Multimed, 2018, 20(11): 3111-3122. [61] 付涵, 范湘涛, 严珍珍, 等. 基于深度学习的遥感图像目标检测技术研究进展. 遥感技术与应用, 2022, 37(2): 290-305.Fu H, Fan X T, Yan Z Z, et al. Progress of object detection in remote sensing images based on deep learning. Remote Sens Technol Appl, 2022, 37(2): 290-305. [62] 蓝鑫, 吴淞, 伏博毅, 等. 深度学习的遥感图像旋转目标检测综述. 计算机科学与探索, 2024, 18(4): 861-877.Lan X, Wu S, Fu B Y, et al. Survey on deep learning in oriented object detection in remote sensing images. J Front Comput Sci Technol, 2024, 18(4): 861-877. [63] 张磊, 张永生, 于英, 等. 遥感图像倾斜边界框目标检测研究进展与展望. 遥感学报, 2022, 26(9): 1723-1743.Zhang L, Zhang Y S, Yu Y, et al. Survey on object detection in tilting box for remote sensing images. Natl Remote Sens Bull, 2022, 26(9): 1723-1743. [64] He K M, Zhang X Y, Ren S Q, et al. Deep Residual Learning for Image Recognition//IEEE Conference on Computer Vision and Pattern Recognition(CVPR). 2016: 770-778. [65] Sutherland I E, Hodgman G W. Reentrant polygon clipping. Commun ACM, 1974, 17(1): 32-42. [66] Amari S I. Backpropagation and stochastic gradient descent method. Neurocomputing, 1993, 5(4/5): 185-196. -
下载:
图(6)
计量
- 摘要浏览量: 129
- HTML全文浏览量: 11
- PDF下载量: 58
- 被引次数: 0
