基于多通道卫星云图的台风中心定位方法研究
|
摘要
台风具有突发性强、破坏力大的特点,是世界上最严重的自然灾害之一我国又是世界上台风灾害最为严重的国家之一。台风给我国东南沿海各省市的工农业生产、交通运输和人民生命财产的安全造成严重威胁和极大损失。正确面对台风、避免台风的灾害是一事关国家和人民生命安全的问题,台风预报也就成为了气象预报的重点之一。确定台风中心位置对台风预报来说具有至关重要的作用。静止卫星云图可以连续监视热带云系的演变,揭示和发现新的天气事实,已成为分析和预报台风的重要工具之一。本文基于静止卫星云图研究台风中心定位方法,主要包括如下三个方面的研究工作：
(1)基于非下采样轮廓波变换(NSCT)和能量准则的多通道卫星云图融合。图像融合对于台风中心定位而言是一个非常重要的预处理。本文基于NSCT和能量熵提出一种多通道卫星云图融合算法。NSCT具有良好的多分辨率、平移不变性和多方向性等特点,它能够对图像的边缘和轮廓给出一种最优渐进表示。局部能量在表示和定位各类图像的特征时具有良好的鲁棒性。本文对静止卫星云图进行多尺度NSCT分解,然后分别计算高频和低频系数的局部能量和能量熵,通过计算能量熵选择一些新的系数。最后通过逆NSCT变换得到融合后的卫星云图。文中讨论了基于小波变换、Contourlet变换以及NSCT融合技术与本文提出的融合方法的性能比较。通过多个融合例子证明本文所提出的方法能够获得丰富的方向信息,并且对噪声有较强的鲁棒性。
(2)基于边界特征的台风主体云系自动识别。台风在不同的发展阶段具有不同的特点,诸如纹理、形状、面积等会发生变化。基于上述这些特征不能将各个发展阶段的台风主体云系自动识别出来。台风在各个发展阶段都有一定的涡旋特性,而非台风云系一般没有这种性质。基于此,利用融合后的卫星云图提取云系的边界特征,并统计边界云的旋转程度,进而自动识别台风主体云系。首先,利用Bezier直方图的曲率曲线获得卫星云图两次分割的两个阂值,分别对融合后的卫星云图进行一次和二次分割；利用分割后的二值化卫星云图结合台风的几何特征,如旋转、面积和形状等特征,自动识别台风主体云系。实验结果表明：利用所提出来的方法能够有效地将各个发展阶段的台风主体云系从卫星云图中识别出来。
(3)基于分形特征和梯度信息的台风中心定位。除了消亡后期的台风云系外,其他各个时期的台风大都有密闭云区,而不管是哪种类型的台风,台风中心绝大部分都位于密闭云区域内。基于此,本文利用分形维数与灰度-梯度共生矩阵的三个二次统计参数结合,定出台风密闭云区域,而后基于密闭云区内台风中心区域梯度信息最丰富的特点,定出台风中心。无论是有眼台风还是无眼台风均可以用该方法定位台风中心,并能获得较高的定位精度。
Typhoon is one of the most serious natural disasters all over the world. China is one of countries that are seriously suffered from typhoon. In many east and south provinces of China, typhoon has threatened the industry and agriculture and people's lives and properties.The problem how to face and avoid this disaster or how to turn the harm into the benefits is very important to our country and people's safety. Typhoon forecasts become one of the most important things of weather forecasts. Center location of typhoon is very important to typhoon forecasts.Stationary satellite cloud images can continuously monitor the tropical cloud system, they can also find new weather facts. It is one of the most important tools in typhoon analyzing and forecasting. Center location of typhoon is studied by the stationary satellite cloud images.Three main research works have been done as follows:
(1)Multi-channel satellite cloud image fusion based on NonSubsampled Contourlet Transform (NSCT) combined with energy entropy. Image fusion is one of the most important processing for locating center. A multi-channel satellite cloud image fusion algorithm is proposed by combining NSCT with energy entropy. NSCT has characteristics of good multi-resolution, shift-invariance and high directionality. It can give an asymptotic optimal representation of edges and contours in image. The local energy is robust in the representing and locating of all kinds of image features. In this paper, NSCT is used to perform a multi-scale decomposition to a stationary satellite cloud image. Secondly, the local energy and the local energy entropy of the high and low frequency-coefficients are calculated separately. Then we choose new coefficients based on the weighting coefficients, which is calculated by energy entropy. Finally, the fused image is generated by reverse NSCT. Compared with Wavelet Transform, Contourlet Transform and NSCT, the issue regarding evaluation of fusion result is also discussed. Some image fusion examples illustrate that the proposed algorithm in this paper gets abundant information of direction and great robustness to noise.
(2) Auto-recognition of typhoon cloud based on boundary features.In different development stages, typhoon has different characteristics such as texture, shape, area, and so on. We couldn't auto-recognize the typhoon clouds in all the stages based only these features.While in these developing stages, typhoon all has helicity but non-typhoon has not. Based on this, we extract boundary features of clouds and statistic the rotation degree of boundary clouds in the fused satellite cloud image.In this paper, we use curvature curve of Bezier histogram to obtain two segmentation thresholds to respectively segment the fused satellite cloud image, and combine typhoon's geometric features, such as rotation, area and shape to automatically recognize the typhoon. Experiment results show that typhoon can be efficiently recognized by the proposed algorithm in all different developing stages.
(3)Center locating of typhoon based on fractal and texture features.Various periods of typhoon all have a region with dense cloud, except parts of latter disappearance typhoon cloud. No matter what kind of typhoon, center is located in the region of the dense cloud. Based on this,following work has bee done to get typhoon center:At first, we combine the fractal dimension and Gray-Gradient Co-occurrence Matrix of three second statistical parameters to locate dense cloud region of typhoon; Then, based on center of typhoon has the most abundant Gradient information in region of the dense cloud, we get the center with Gauss and Canny algorithm.Whether typhoon has an Eyed or not, the center region of typhoon all can be located by the proposed algorithm, and high location accuracy can be obtained.
(1)基于非下采样轮廓波变换(NSCT)和能量准则的多通道卫星云图融合。图像融合对于台风中心定位而言是一个非常重要的预处理。本文基于NSCT和能量熵提出一种多通道卫星云图融合算法。NSCT具有良好的多分辨率、平移不变性和多方向性等特点,它能够对图像的边缘和轮廓给出一种最优渐进表示。局部能量在表示和定位各类图像的特征时具有良好的鲁棒性。本文对静止卫星云图进行多尺度NSCT分解,然后分别计算高频和低频系数的局部能量和能量熵,通过计算能量熵选择一些新的系数。最后通过逆NSCT变换得到融合后的卫星云图。文中讨论了基于小波变换、Contourlet变换以及NSCT融合技术与本文提出的融合方法的性能比较。通过多个融合例子证明本文所提出的方法能够获得丰富的方向信息,并且对噪声有较强的鲁棒性。
(2)基于边界特征的台风主体云系自动识别。台风在不同的发展阶段具有不同的特点,诸如纹理、形状、面积等会发生变化。基于上述这些特征不能将各个发展阶段的台风主体云系自动识别出来。台风在各个发展阶段都有一定的涡旋特性,而非台风云系一般没有这种性质。基于此,利用融合后的卫星云图提取云系的边界特征,并统计边界云的旋转程度,进而自动识别台风主体云系。首先,利用Bezier直方图的曲率曲线获得卫星云图两次分割的两个阂值,分别对融合后的卫星云图进行一次和二次分割；利用分割后的二值化卫星云图结合台风的几何特征,如旋转、面积和形状等特征,自动识别台风主体云系。实验结果表明：利用所提出来的方法能够有效地将各个发展阶段的台风主体云系从卫星云图中识别出来。
(3)基于分形特征和梯度信息的台风中心定位。除了消亡后期的台风云系外,其他各个时期的台风大都有密闭云区,而不管是哪种类型的台风,台风中心绝大部分都位于密闭云区域内。基于此,本文利用分形维数与灰度-梯度共生矩阵的三个二次统计参数结合,定出台风密闭云区域,而后基于密闭云区内台风中心区域梯度信息最丰富的特点,定出台风中心。无论是有眼台风还是无眼台风均可以用该方法定位台风中心,并能获得较高的定位精度。
Typhoon is one of the most serious natural disasters all over the world. China is one of countries that are seriously suffered from typhoon. In many east and south provinces of China, typhoon has threatened the industry and agriculture and people's lives and properties.The problem how to face and avoid this disaster or how to turn the harm into the benefits is very important to our country and people's safety. Typhoon forecasts become one of the most important things of weather forecasts. Center location of typhoon is very important to typhoon forecasts.Stationary satellite cloud images can continuously monitor the tropical cloud system, they can also find new weather facts. It is one of the most important tools in typhoon analyzing and forecasting. Center location of typhoon is studied by the stationary satellite cloud images.Three main research works have been done as follows:
(1)Multi-channel satellite cloud image fusion based on NonSubsampled Contourlet Transform (NSCT) combined with energy entropy. Image fusion is one of the most important processing for locating center. A multi-channel satellite cloud image fusion algorithm is proposed by combining NSCT with energy entropy. NSCT has characteristics of good multi-resolution, shift-invariance and high directionality. It can give an asymptotic optimal representation of edges and contours in image. The local energy is robust in the representing and locating of all kinds of image features. In this paper, NSCT is used to perform a multi-scale decomposition to a stationary satellite cloud image. Secondly, the local energy and the local energy entropy of the high and low frequency-coefficients are calculated separately. Then we choose new coefficients based on the weighting coefficients, which is calculated by energy entropy. Finally, the fused image is generated by reverse NSCT. Compared with Wavelet Transform, Contourlet Transform and NSCT, the issue regarding evaluation of fusion result is also discussed. Some image fusion examples illustrate that the proposed algorithm in this paper gets abundant information of direction and great robustness to noise.
(2) Auto-recognition of typhoon cloud based on boundary features.In different development stages, typhoon has different characteristics such as texture, shape, area, and so on. We couldn't auto-recognize the typhoon clouds in all the stages based only these features.While in these developing stages, typhoon all has helicity but non-typhoon has not. Based on this, we extract boundary features of clouds and statistic the rotation degree of boundary clouds in the fused satellite cloud image.In this paper, we use curvature curve of Bezier histogram to obtain two segmentation thresholds to respectively segment the fused satellite cloud image, and combine typhoon's geometric features, such as rotation, area and shape to automatically recognize the typhoon. Experiment results show that typhoon can be efficiently recognized by the proposed algorithm in all different developing stages.
(3)Center locating of typhoon based on fractal and texture features.Various periods of typhoon all have a region with dense cloud, except parts of latter disappearance typhoon cloud. No matter what kind of typhoon, center is located in the region of the dense cloud. Based on this,following work has bee done to get typhoon center:At first, we combine the fractal dimension and Gray-Gradient Co-occurrence Matrix of three second statistical parameters to locate dense cloud region of typhoon; Then, based on center of typhoon has the most abundant Gradient information in region of the dense cloud, we get the center with Gauss and Canny algorithm.Whether typhoon has an Eyed or not, the center region of typhoon all can be located by the proposed algorithm, and high location accuracy can be obtained.
引文
[1]王丽军.基于嵌入式隐马尔可夫模型与交叉熵的台风中心定位[D].天津：天津大学,2006.
[2]陈渭民.卫星气象学[M].北京气象出版社,1989：323-370.
[3]Hasler AF.,Palaniappan K.,Kambhammetu C.,Black P.,Uhlhorn E., Chesters D. High-resolution wind field within the inner core and eye of a mature tropical cyclone from GOES 1-min images [J].Bulletin of The American Meteorological Society,1999,80(1):97.
[4]Candes E.J.Ridgelets:theory and Applications [D].USA:Department of Statistics, Stanford University.1998.
[5]Candes E. J. Monoscale Ridgelet for the Representation of Images with Edges [R]. USA:Department of Statistics,Stanford University.1999.
[6]Candes E. J.,Donoho D. L. Curvelets [R].USA:Department of Statistics, Stanford University.1998.
[7]Do M.N.,Vetterli M. Contourlets, in Beyond Wavelets [J].New York:Academic Press,2002.
[8]Do M.N.,Vetterli M. Contourlets:a new directional multi-resolution image representation. Signal, Systems and Computers,2002,1:497-501.
[9]于波,冯民学,陈必云.模糊神经网络在台风云系图像识别中的应用[J].气象,1996,22(1)：22-25.
[10]刘正光,林孔元,郭爱民,程彦.卫星云图形态特征提取[J].计算机研究与发展,1997,34(9)：689-693
[11]王虹,余建波,陈明明,刘嵩鹤,关爱贤,童水兵.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.
[12]刘凯,黄峰,罗坚.台风卫星云图分割方法研究[J].微机发展,2001,卷1：54—55.
[13]刘凯,黄峰,罗坚.基于纹理特征的卫星云图台风自动识别方法[J].微型机与应用,2001,卷9：48—49.
[14]郝玉龙,程宝义,范茵,张宏军.基于纹理方向整体分布特征的台风自动识别方法[J].中国图象图形学报,2002,7(12)：1319-1322.
[15]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
[16]师春香,吴蓉璋,项续康.多阈值和神经网络卫星云图云系自动分割试验[J].应用气象学,2001,12(1)：70-78.
[17]李俊,周凤仙.气象卫星台风云图的自动识别方法及其应用[J].应用气象学报,1992,3(4)：402—409
[18]李艳兵,李元祥,瞿景秋.卫星云图形态特征提取和表示的一种方法[J].南京气象学院学报,2006,29(5)：682-687.
[19]薛俊韬,刘正光,刘还珠.小波变换在云图边缘处理中的应用[J].天津大学学报,2002,35(6)：736-739.
[20]吴冰.台风云系中心定位智能系统研究[D].天津：天津大学,2002.
[21]刘凯,黄峰,罗坚.无眼台风自动定位方法研究[J].信息与控制,2001,30(6)：543-546.
[22]刘正光,喻远飞,吴冰.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.
[23]谢俊元,艾早阳,高勇.台风中心定位中的螺旋线自动识别方法[J].软件学报,1997：398-403.
[24]王燕燕,叶臻,孙慰迟.台风中心的旋转定位[J].中国图像图形学报,2002,7(5)：491-494.
[25]Harris, Hurray J.R.,Hirose T.,HIS transformation for the integration of radar imagery with other remotely sensed Data [J].Photogrammetry Engineering and Remote Sensing,1990,56(12):87-93.
[26]Chavez P.S.,Sides S.C.,Anderson J.A. Comparison of there different methods to merge Multi-resolution and Multi-spectral data:Landsat TM and SPOT panchromatic [J].Photogrammetry Engineering and Remote Sensing,1991,57(3): 295-303.
[27]Giorgio B.,Schwartz E. L. Improved cross-correlation for template matching on the laplacian pyramid [J].Pattern Recognition Letters,1998, (19):765-770
[28]Guy F.,Oliver A.,Florence L.,Selma C.Multi-resolution analysis and reconstruction by a morphological pyramid in the remote sensing of terrestrial surfaces [J].Signal Processing,2001, (81):2171-2185.
[29]Alexander T.,van Ruyven L.J.,Valeton J.M.Merging thermal and visual images by a contrast pyramid [J].Optical Engineering,1989,28(7):789-792.
[30]郭雷,李辉辉,鲍永生.图像融合[M].北京：电子工业出版社,2008：229-248
[31]Da Cunha A. L.,Zhou J. P.,Do M.N. The nonsubsampled contourlet transform: theory, design, and applications [J].IEEE Transaction on Image Processing,2006, 15(10):3089-3101
[32]Bamberger R. H.,Smith M.J.T. A filter bank for the directional decomposition of images:Theory and design [J].IEEE Trans Signal Processing,1992,40 (4): 882-893.
[33]Owens R. Feature detection via phase congruency [D].Australia:Dept. of Computer Science, the University of Western Australia,1996.
[34]Venkatesh S.,Owens R. An energy feature detection scheme [J].The International Conference on Image Processing, Singapore,1989:553-557.
[35]金梅,张长江,李冰.红外图像目标快速分割方法[J].激光与红外,2004,34(3)：222-224.
[36]程帅.基于小波分析的遥感图像增强研究[D].成都：成都理工大学,2006.
[37]姜青香,刘慧平.利用纹理分析方法提取TM图像信息[J].遥感学报,2004,8(5)：458-464.
[38]洪继光.灰度——梯度共生矩阵纹理分析方法[J].自动化学报,1984,10(1)：22-25.
[39]余建波.基于气象卫星云图的云类识别及台风分割和中心定位研究[D].武汉：武汉理工大学,2008.
[40]孔秀梅.形成期台风螺旋云带的提取、描述及中心定位的研究[D].天津：天津大学,2003.
[41]王振会,曾维麟.卫星云迹风微机客观导出系统[J].南京气象学院学报,1996,19(1)：69-75.
[42]王振会,许建明,G.Kelly.基于傅立叶相位分析的卫星云图导风技术[J].气象科学,2004,24(1).
[43]Wong K. Y.,Yip C.L.,Li P. W. Automatic tropical cyclone eye fix using genetic algorithm [J].Expert Systems with Applications,2008,34(1):643-656.
[44]Chaudhur I. B.,Sarkar N. Texture segmentation using fractal dimension [J].IEEE Transactions Pattern Analysis and Machine Intelligence,1995,17 (1):72-77.
[2]陈渭民.卫星气象学[M].北京气象出版社,1989：323-370.
[3]Hasler AF.,Palaniappan K.,Kambhammetu C.,Black P.,Uhlhorn E., Chesters D. High-resolution wind field within the inner core and eye of a mature tropical cyclone from GOES 1-min images [J].Bulletin of The American Meteorological Society,1999,80(1):97.
[4]Candes E.J.Ridgelets:theory and Applications [D].USA:Department of Statistics, Stanford University.1998.
[5]Candes E. J. Monoscale Ridgelet for the Representation of Images with Edges [R]. USA:Department of Statistics,Stanford University.1999.
[6]Candes E. J.,Donoho D. L. Curvelets [R].USA:Department of Statistics, Stanford University.1998.
[7]Do M.N.,Vetterli M. Contourlets, in Beyond Wavelets [J].New York:Academic Press,2002.
[8]Do M.N.,Vetterli M. Contourlets:a new directional multi-resolution image representation. Signal, Systems and Computers,2002,1:497-501.
[9]于波,冯民学,陈必云.模糊神经网络在台风云系图像识别中的应用[J].气象,1996,22(1)：22-25.
[10]刘正光,林孔元,郭爱民,程彦.卫星云图形态特征提取[J].计算机研究与发展,1997,34(9)：689-693
[11]王虹,余建波,陈明明,刘嵩鹤,关爱贤,童水兵.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.
[12]刘凯,黄峰,罗坚.台风卫星云图分割方法研究[J].微机发展,2001,卷1：54—55.
[13]刘凯,黄峰,罗坚.基于纹理特征的卫星云图台风自动识别方法[J].微型机与应用,2001,卷9：48—49.
[14]郝玉龙,程宝义,范茵,张宏军.基于纹理方向整体分布特征的台风自动识别方法[J].中国图象图形学报,2002,7(12)：1319-1322.
[15]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
[16]师春香,吴蓉璋,项续康.多阈值和神经网络卫星云图云系自动分割试验[J].应用气象学,2001,12(1)：70-78.
[17]李俊,周凤仙.气象卫星台风云图的自动识别方法及其应用[J].应用气象学报,1992,3(4)：402—409
[18]李艳兵,李元祥,瞿景秋.卫星云图形态特征提取和表示的一种方法[J].南京气象学院学报,2006,29(5)：682-687.
[19]薛俊韬,刘正光,刘还珠.小波变换在云图边缘处理中的应用[J].天津大学学报,2002,35(6)：736-739.
[20]吴冰.台风云系中心定位智能系统研究[D].天津：天津大学,2002.
[21]刘凯,黄峰,罗坚.无眼台风自动定位方法研究[J].信息与控制,2001,30(6)：543-546.
[22]刘正光,喻远飞,吴冰.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.
[23]谢俊元,艾早阳,高勇.台风中心定位中的螺旋线自动识别方法[J].软件学报,1997：398-403.
[24]王燕燕,叶臻,孙慰迟.台风中心的旋转定位[J].中国图像图形学报,2002,7(5)：491-494.
[25]Harris, Hurray J.R.,Hirose T.,HIS transformation for the integration of radar imagery with other remotely sensed Data [J].Photogrammetry Engineering and Remote Sensing,1990,56(12):87-93.
[26]Chavez P.S.,Sides S.C.,Anderson J.A. Comparison of there different methods to merge Multi-resolution and Multi-spectral data:Landsat TM and SPOT panchromatic [J].Photogrammetry Engineering and Remote Sensing,1991,57(3): 295-303.
[27]Giorgio B.,Schwartz E. L. Improved cross-correlation for template matching on the laplacian pyramid [J].Pattern Recognition Letters,1998, (19):765-770
[28]Guy F.,Oliver A.,Florence L.,Selma C.Multi-resolution analysis and reconstruction by a morphological pyramid in the remote sensing of terrestrial surfaces [J].Signal Processing,2001, (81):2171-2185.
[29]Alexander T.,van Ruyven L.J.,Valeton J.M.Merging thermal and visual images by a contrast pyramid [J].Optical Engineering,1989,28(7):789-792.
[30]郭雷,李辉辉,鲍永生.图像融合[M].北京：电子工业出版社,2008：229-248
[31]Da Cunha A. L.,Zhou J. P.,Do M.N. The nonsubsampled contourlet transform: theory, design, and applications [J].IEEE Transaction on Image Processing,2006, 15(10):3089-3101
[32]Bamberger R. H.,Smith M.J.T. A filter bank for the directional decomposition of images:Theory and design [J].IEEE Trans Signal Processing,1992,40 (4): 882-893.
[33]Owens R. Feature detection via phase congruency [D].Australia:Dept. of Computer Science, the University of Western Australia,1996.
[34]Venkatesh S.,Owens R. An energy feature detection scheme [J].The International Conference on Image Processing, Singapore,1989:553-557.
[35]金梅,张长江,李冰.红外图像目标快速分割方法[J].激光与红外,2004,34(3)：222-224.
[36]程帅.基于小波分析的遥感图像增强研究[D].成都：成都理工大学,2006.
[37]姜青香,刘慧平.利用纹理分析方法提取TM图像信息[J].遥感学报,2004,8(5)：458-464.
[38]洪继光.灰度——梯度共生矩阵纹理分析方法[J].自动化学报,1984,10(1)：22-25.
[39]余建波.基于气象卫星云图的云类识别及台风分割和中心定位研究[D].武汉：武汉理工大学,2008.
[40]孔秀梅.形成期台风螺旋云带的提取、描述及中心定位的研究[D].天津：天津大学,2003.
[41]王振会,曾维麟.卫星云迹风微机客观导出系统[J].南京气象学院学报,1996,19(1)：69-75.
[42]王振会,许建明,G.Kelly.基于傅立叶相位分析的卫星云图导风技术[J].气象科学,2004,24(1).
[43]Wong K. Y.,Yip C.L.,Li P. W. Automatic tropical cyclone eye fix using genetic algorithm [J].Expert Systems with Applications,2008,34(1):643-656.
[44]Chaudhur I. B.,Sarkar N. Texture segmentation using fractal dimension [J].IEEE Transactions Pattern Analysis and Machine Intelligence,1995,17 (1):72-77.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |