A METHOD FOR INTELLIGENT RECOGNITION OF RADAR SQUALL LINE
-
摘要: 提出一种雷达回波图像中飑线特征自动识别的方法。以多普勒天气雷达探测资料为主要数据源,对雷达探测到的基本反射率的空间分布和强度进行分析,通过数值预处理、高通滤波、二值化降噪、图像特征提取、目标物的中轴线提取,以及飑线形态分析等一系列步骤,实现对雷达飑线特征的智能识别。克服了回波高值区域不连通、碎块化对飑线自动识别造成的困难。通过4次强对流天气个例检验,飑线自动识别的准确率达到75%左右,尤其对呈现直线或劣弧状,且边界清晰的高值回波区,具有更高的识别成功率。该方法将以往需要由气象专业人员主观分析、判读雷达回波图像的工作自动化、客观化,可提高飑线识别、强对流天气预警相关业务的准确性和时效性。Abstract: In this paper, we propose an intelligent and early warning recognition method of squall line characteristics in weather radar images. Doppler weather radar data is used as the main data to analyze the spatial distribution and intensity of the base reflectivity detected by radar. Through data preprocessing, high-pass filtering, binarization noise reduction, image feature extraction, central axis extraction of target, and morphological analysis of the squall line, intelligent recognition and early warning of radar squall line characteristics is realized. The method overcomes the difficulty caused by the disconnection of the high-value area of the echo and the automatic identification of the squall line by the fragmentation. Through the test of 4 strong convective weather cases, the accuracy of the automatic identification of the squall line reaches about 75%. In particular, the high-value echo region, which exhibits straight or inferior arc and clear boundary, has higher identification rate. This method can automatically and objectively analyze the work of subjective analysis and judgment of radar echo images by meteorological professionals, and can improve the accuracy and effectiveness of squall line identification and strong convective weather warning services.
-
Key words:
- synoptic meteorology /
- squall line /
- weather radar /
- radar echo /
- squall line feature recognition
-
表 1 个例检验的参数取值
序号 参数名 参数值 参数意义 1 max_α 2 计算过程中处理的最高仰角 2 Filt_A 60 dBZ 回波图像高通滤波的阈值 3 SL 5 降噪处理时正方形区域的边长 4 k 0.25 降噪处理时的调节系数 5 r_i 20 二维矩阵MAT的纵向步长 6 ω_i 5 二维矩阵MAT的横向步长 7 n 57 计数器,与r_i配合使用 8 n﹒r_i 1140 MAT纵向的最大刻度 9 Trd 0.75 筛选MAT中的大值单元的判定阈值 10 Filt_B 5dBZ 下一次高通滤波时Filt_A降低的单位量 11 Filt_N 50 dBZ 高通滤波迭代计算的最小基本反射率 表 2 三项评价指标的得分汇总
地点 k Trd POD FAR CSI 南京 0.1 0.70 94.76% 84.69% 15.18% 0.2 0.75 92.67% 83.72% 16.08% 0.3 0.80 83.25% 84.65% 14.89% 合肥 0.1 0.70 92.56% 88.90% 11.00% 0.2 0.75 89.26% 88.44% 11.40% 0.3 0.80 81.82% 88.90% 10.83% -
[1] 陈淑琴, 章丽娜, 俞小鼎.浙北沿海连续3次飑线演变过程的环境条件[J].应用气象学报, 2017, 28(3):357-368. [2] 丁治英, 夏蘩, 高松, 等.一次飑线过程中雷达回波的组织结构变化及成因分析[J].热带气象学报, 2017, 33(3):323-333. [3] 李文娟, 郦敏杰, 李嘉鹏, 等.浙江省春季至夏初飑线分型及对比分析[J].热带气象学报, 2019, 35(4):480-490. [4] 徐凤梅, 王慧娟. 2009年6月3日商丘市强飑线天气过程及其对农业的影响[J].现代农业科技, 2011(11):306. [5] 竹利, 陈朝平, 陈茂强, 等.川北飑线成熟阶段灾害性大风成因个例分析[J].暴雨灾害, 2018, 37(2):164-173. [6] 李斌, 吴立广.自动气象站资料在模拟苏北一次飑线过程中的应用[J].热带气象学报, 2019, 35(6):789-800. [7] 方翀, 林隐静, 曹艳察, 等.华南地区西风带飑线和台风飑线环境场特征统计对比分析[J].热带气象学报, 2017, 33(6):965-974. [8] KELLY W E, RAND T W, UCKUN S, et al. Image processing for hazard recognition in on-board weather radar: U. S. Patent 6650275 B1[P]. 1-12. 2003-11-18. [9] 程凌舟, 何建新, 曾宪军.基于小波和Hu矩的飑线雷达回波识别[J].成都信息工程大学学报, 2017, 32(4):369-374. [10] 李哲, 李国翠, 刘黎平, 等.飑线优化识别及雷暴大风分析[J].高原气象, 2017, 36(3):801-810. [11] BLUESTEIN H B, MARX G T, JAIN M H. Formation of mesoscale lines of precipitation:nonsevere squall lines in Oklahoma during the spring[J]. Mon Wea Rev, 1987, 115(11):2719-2727. [12] ROTUNNO R, KLEMP J B, WEISMAN M L. A theory for strong, long-lived squall lines[J]. J Atmos Sci, 1988, 45(3): 463-485. [13] 田荟君, 潘玉洁, 刘佳, 等. 2014年5月31日华东地区一次暖区飑线过程的多普勒天气雷达分析[J].气象科学, 2018, 38(2):191-202. [14] 万夫敬, 江敦双, 赵传湖. 2017年8月6日山东罕见雷暴大风成因分析[J].海洋气象学报, 2018, 38(2):60-66. [15] 侯淑梅, 闵锦忠, 刁秀广, 等.飑线发展过程中回波合并的特征分析[J].大气科学学报, 2018, 41(3):367-376. [16] 张萍, 桂园园, 李文辉, 等.一次强飑线天气过程的雷达回波特征分析[J].江西科学, 2018, 36(1):106-111. [17] 吴瑞姣, 陶玮, 周昆, 等.江淮灾害性大风飑线的特征分析[J].气象, 2019, 45(2):155-165. [18] 李少英.混合方法同化地面站和雷达观测对飑线预报的影响[D].兰州: 兰州大学, 2018: 16-83. [19] 吴芳芳, 俞小鼎, 张志刚, 等.苏北地区超级单体风暴环境条件与雷达回波特征[J].气象学报, 2013, 71(2): 209-227. [20] 鲁蓉, 孙建华, 李德帅.初值水汽场对华南春季一次强飑线触发和维持影响的数值试验[J].热带气象学报, 2019, 35(1):37-50. [21] 王耀贵.图像高斯平滑滤波分析[J].计算机与信息技术, 2008(8):79-81+90. [22] KNOPP J, PRASAD M, WILLEMS G, et al. Hough transform and 3D SURF for robust three dimensional classification[M]. Computer Vision-ECCV, 2010:1-20. [23] YAO A, GALL J, GOOL L V. A Hough transform-based voting framework for action recognition[C]. Computer Vision & Pattern Recognition, 2010: 2-20. [24] DUDA R O, HART P E. Use of the hough transformation to detect lines and curves in pictures[J]. Communications of the ACM, 1972, 15(1):11-15. [25] 钟敏, 吴翠红, 王珊珊, 等. CINRAD/SA雷达两种识别跟踪产品的评估分析[J].气象, 2012, 38(6):722-727.