A Typhoon Center Location Method Based on Deep Neural Network
-
摘要: 台风中心定位的微小误差会对台风路径预报造成较大的偏离，因此精确定位台风中心是台风路径预测和灾害预报的重要步骤。台风云系随时间不断变化且风力强弱不一，在卫星云图中呈现了多样性和复杂性，现有基于神经网络的模型由于缺少对台风特征图像多维度参数的权重合理分配，在自动提取台风图像特征上受到了限制。为此，提出一种融合通道注意力与坐标注意力的神经网络模型（TY-LOCNet），首先搭建深度卷积神经网络模型提取台风特征；其次引入通道注意力机制从台风特征中捕获通道级别的信息，提升模型对重要通道的关注度；然后将通道注意力结果输入到坐标注意力机制中全局标定台风位置信息，使模型能够在较大的区域关注到台风的形态结构；此外，均方误差损失函数未能融合计算坐标导致定位精度低，因此提出距离损失函数(DISTLoss)通过距离回归提高模型定位精度。实验结果表明，TY-LOCNet的平均位置误差(MLE)、平均定位误差(MAE)和检测速度分别为3.502像素，0.292°和17 FPS，优于其他模型。台风中心定位模型TY-LOCNet可为台风预报提供实时性台风中心定位支持。Abstract: Minor errors in typhoon center position can cause significant deviations in typhoon path prediction, so accurately locating typhoon center is an important step in typhoon path prediction and disaster early warning. Typhoon cloud systems change continuously with varying wind strength, leading to diverse and complex satellite images. Existing models based on neural networks are limited in the automatic extraction of typhoon features due to the lack of reasonable weight allocation for multi-dimensional parameters in typhoon images. For this reason, this paper proposed a neural network model (TY-LOCNet) that integrated channel attention and coordinate attention. Firstly, a deep convolutional neural network model was built to extract typhoon characteristics. Secondly, the channel attention mechanism was introduced to capture channel-level information from typhoon characteristics and enhance the attention of the model on important channels. Moreover, the channel attention results were input into the coordinate attention mechanism to calibrate typhoon position information globally so that the model can focus on the morphological structure of typhoons in large areas. Furthermore, the mean square error loss function failed to fuse the calculated coordinates, resulting in low locating accuracy. Therefore, the distance loss function (DISTLoss) was proposed to improve the locating accuracy of the model through distance regression. Experimental results show that the mean location error, mean absolute error, and detection speed of TY-LOCNet were 3.502 pixels, 0.292°, and 17 FPS, respectively, outperforming other models. Therefore, the typhoon center location model TY-LOCNet may provide real-time information on typhoon center position for typhoon forecasting.
-
Key words:
- typhoon center location /
- attention mechanism /
- neural network /
- distance loss function
-
表 1 不同强度等级的台风中心风速和图片数量
台风级别 风速/(m·s-1) 图片数量 热带风暴 17.2~24.4 859 强热带风暴 24.5~32.6 1 477 台风 32.7~41.4 3 744 强台风 41.5~50.9 1 121 超强台风 ＞51.0 547 
下载: 导出CSV
表 2 五种注意力机制的试验结果
注意力机制 平均位置误差/像素 通道注意力 8.754 7 空间注意力 8.923 3 坐标注意力 8.194 7 自注意力 8.875 7 残差注意力 7.365 6 — 9.026 4
下载: 导出CSV
表 3 在不同位置使用双注意力机制模块的实验结果
位置 平均位置误差/像素 {1, 2, 3} 8.914 7 {3, 4, 5} 8.245 4 {4, 5, 6} 7.365 6 {1, 2, 3, 4, 5, 6} 7.744 0 ― 9.036 7
下载: 导出CSV
表 5 模型对2018年台风的定位误差
台风编号 台风名称 平均定位误差/° 台风编号 台风名称 平均定位误差/° 201801 布拉万 0.381 3 201816 贝碧嘉 0.406 6 201802 三巴 0.164 2 201817 赫克托 0.412 0 201803 杰拉华 0.242 2 201818 温比亚 0.385 9 201804 艾云尼 0.415 6 201819 苏力 0.106 0 201805 马力斯 0.348 9 201820 西马仑 0.263 5 201806 格美 0.389 8 201821 飞燕 0.181 1 201807 派比安 0.228 9 201822 山竹 0.117 0 201808 玛莉亚 0.132 5 201823 百里嘉 0.456 6 201809 山神 0.401 4 201824 潭美 0.183 2 201810 安比 0.317 7 201825 康妮 0.208 8 201811 悟空 0.285 7 201826 玉兔 0.181 9 201812 云雀 0.249 8 201827 桃芝 0.323 3 201813 珊珊 0.191 0 201828 万宜 0.215 4 201814 摩羯 0.502 3 201829 天兔 0.226 8 201815 丽琵 0.501 1 ― ― 平均: 0.290 3
下载: 导出CSV
表 6 四种模型性能对比
模型 每秒浮点运算次数 参数量/MB 检测速度/（幅·秒-1） Hourglass 8 256×106 14.99 5.01 TCLNet 6 980×106 1.09 7.01 TY-LOCNet 4 006×106 13.98 17.47 DRTCL 15 470×106 138.33 ＜2
下载: 导出CSV
表 7 不同输入尺寸的模型参数和误差对比
输入尺寸 每秒浮点运算次数 参数量/MB 检测速度（/幅·秒-1） 平均位置误差/像素 512×512 20 157×106 13.98 1.14 7.144 5 384×384 12 021×106 13.98 7.56 7.278 7 224×224 4 006×106 13.98 17.47 7.315 6 128×128 668×106 13.98 22.88 7.821 5
下载: 导出CSV
-
[1] 曾明剑, 于波, 周曾奎, 等. 卫星红外云图上台风中心定位技术研究和应用[J]. 热带气象学报, 2006, 22(3): 241-247. [2] 刘正光, 喻远飞, 吴冰, 等. 利用云导风矢量的台风中心自动定位[J]. 气象学报, 2003, 61(5): 636-640. [3] 王虹, 余建波, 陈明明, 等. 基于FY-2C气象卫星云图的台风分割方法的研究[J]. 计算机工程与应用, 2008, 44(20): 188-191. [4] 陈希, 李妍, 毛科峰, 等. 基于非线性最优目标函数的红外云图台风中心自动定位[J]. 热带气象学报, 2013, 29(1): 157-160. [5] 庞君如, 王让会, 李博. 基于密度矩阵和卫星修正的台风中心定位仿真[J]. 计算机仿真, 2019, 36(6): 398-402. [6] 刘佳, 王旭东. 一种利用FY-2卫星数据的台风中心定位方法[J]. 遥感信息, 2020, 35 (2): 29-33. [7] 王晓霞, 刘建强, 张毅, 等. 基于卫星微波散射计风场的台风中心定位方法比较研究[J]. 海洋预报, 2020, 37(5): 24-33. [8] WANG P, WANG P, WANG C, et al. A center location algorithm for tropical cyclone in satellite infrared images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13: 2 161-2 172. [9] YANG X, ZHAN Z, SHEN J. A deep learning based method for typhoon recognition and typhoon center location[C]//IGARSS 2019—2019 IEEE International Geoscience and Remote Sensing Symposium. IEEE, 2019: 9 871-9 874. [10] TOMPSON J J, JAIN A, LECUN Y, et al. Joint training of a convolutional network and a graphical model for human pose estimation[C]// 27th Internatianal Conference on Neural Information Processing Systems. Cambridge: MIT Press, 2014: 1 799-1 807. [11] TAN C. Learning to locate typhoon center using deep neural network[C]//2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS. IEEE, 2021: 4600-4603. [12] LI J, BIAN S, ZENG A, et al. Human pose regression with residual log-likelihood estimation[C]//Proceedings of the IEEE / CVF International Conference on Computer Vision. 2021: 11 025-11 034. [13] WANG P, YANG X, JI Z, et al. A Typhoon Center Location Method on Satellite Images Based on Deep Reinforcement Learning[C]// 2021 40th Chinese Control Conference. Shanghai: IEEE, 2021: 7 046-7 053. [14] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, SLC: IEEE, 2018: 7 132-7 141. [15] WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[C]//Proceedings of the European conference on computer vision (ECCV). Cham: Springer, 2018. [16] HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design[C]//Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 13 713-13 722. [17] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. California: ACM, 2017: 6 000−6 010. [18] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition, LA: IEEE, 2016: 1 063-6 919. [19] JAISWAL N, KISHTAWAL C M. Automatic determination of center of tropical cyclone in satellite-generated IR images[J]. IEEE Geoscience and Remote Sensing Letters, 2011, 8(3): 460-463. [20] WIMMERS A J, VELDEN C S, Objectively determining the rotational center of tropical cyclones in passive microwave satellite imagery[J], J Appl Meteor Climatol, 2010, 49(9): 2 013-2 034. -
点击查看大图
图(5) / 表(7)
计量
- 文章访问数: 73
- HTML全文浏览量: 11
- PDF下载量: 150
- 被引次数: 0
粤公网安备 4401069904700003号