强单体客观建模方法的研究
|
摘要
冰雹是主要气象灾害之一,冰雹云的发生发展具有突发性,移动迅速,生命史短等特点,目前,预报员多采用多普勒雷达回波的系列产品对其监测和预报。雹云的发生发展具有规律性,雹云的回波强度及形态特征具有某些特殊性,这些是人工预报的直观视觉依据,也是本文展开基于冰雹云图像特征的自动识别研究的基础。
目前,围绕冰雹系统识别的研究工作多在机理分析、预报因子、主观建模等方面展开,而基于多普勒雷达回波反射率图像的冰雹云体特征及雹云演变规律的自动识别方法的研究尚未见到相关报道。
数字图像中,被识别对象的形态特征的自动分析多用于和医学相关的领域,所研究的对象多为刚体或有固定形状,其各个部分之间的相对运动不很明显,而冰雹天气过程的云体内部相对运动比较剧烈、层次不是很分明且不同部分之间过渡缓慢,这就使已有的形态特征提取方法以及运动体跟踪方法对雹云而言不十分有效。
本文就雹云的自动识别问题在如下方面提出新的思路和解决方案并取得满意结果:
1.首先,对冰雹云雷达图像展开模式分析,并就典型雹云的特殊形态和特定组成构建了一种三层模型,即云体核、以核为中心的主体层和主体外侧的过渡层。在这三层模型中,其核的强度高、主体层呈钩状、过渡层极窄。
2.为提取雹云模型中主体层钩状特征和过渡层的梯度特征,构建了一种“探针”模型,并成功设计出基于“探针”模型的雹云形态特征和梯度特征的提取算法;进而,采用支持向量机的方法训练出用以识别雹云的分类模型,该模型在测试集中表现出来的较高的识别率表明雹云模型的合理性和所提特征的有效性。
3.为尽可能提早预报,从雹云发生发展规律的角度展开了对时间序列图像的跟踪分析,提出了一个能够提取雷达反射率图中冰雹云动态发展规律的算法模型,与上述特征提取思路不同,该模型主要反映云体的移动速度、方向以及云体的发展、碰撞及复合等。
4.将回波反射率图数据与径向速度图数据相融合,提出了一种去除超折射噪声的简洁而有效的方法。
以上研究所涉及算法均已编程实现,并在样本测试过程中表现出较强的稳定性和良好的测试效果。
Hailstone is a main disaster in the Meteorological area. The hailstorm arises suddenly, moves fast, and disappears very quickly. Now, weather operators monitor it and make hailstone forecast mainly by reflectivity images of Doppler radar. The development of the hailstone is of regularity that the hailstone’s reflectivity intensity and its morphological features make them difference, which help weather operators make forecast. These features are gist of the research on the hailstone automatic detection based on hailstone’s radar image, which is done in this dissertation.
Recently, the research on hailstone’s detection is mainly throw its light on mechanics analysis, forecasting factor and subjective model. There is no similar method to the one described in this dissertation reported yet.
The automatic analysis of the images’morphological character is mainly used in medical field with objects to be the rigid body which have rigid fixing shape and little movement among parts of it, but the radar reflectivity image of hail clouds is different. In the hail cloud, movement among parts of which is violent and the interfaces of these parts are not very clear, which makes traditional morphological character extraction and the moving object tracing methods inefficient with the hail cloud.
In this dissertation, some new ideas and methods are brought forward.
1. With the pattern analysis of hail cloud in radar reflectivity, a model is constructed in which the image of cloud is separated into three layers: core, body and the transition band around of the body according to typical hail cloud reflectivity image. As to the hail clouds, they have some visual features that the reflectivity of the core is very high, the body is hook-like and the transition band is thin.
2. To extraction these features of the model constructed, another model named“probe”is introduced. A method for feature extraction, including the gradient information of the transition band and the attribute about its morphological character especially those about the hook-like body, is proposed based on the“probe”model for hail auto-forecasting. Furthermore, a Support Vector Machine classifier is trained with features extracted from the probe model. The classifier is shown that its accuracy is satisfactory, which testify the efficiency of the method in this dissertation.
3. For predicting the hailstone as early as most, the process of the hailstone clouds development is researched using the time series method and a corresponding algorithm model is constructed. In the algorithm model the velocity and direction of cloud’s movement, the separation and the combination in hail clouds developing process are taken into account.
4. A method is put forward to filter the super refraction echo in reflectivity image by using both of reflectivity image and radial speed image.
All algorithms are coded with C++ language. It is stable and effective, which is testified by test.
目前,围绕冰雹系统识别的研究工作多在机理分析、预报因子、主观建模等方面展开,而基于多普勒雷达回波反射率图像的冰雹云体特征及雹云演变规律的自动识别方法的研究尚未见到相关报道。
数字图像中,被识别对象的形态特征的自动分析多用于和医学相关的领域,所研究的对象多为刚体或有固定形状,其各个部分之间的相对运动不很明显,而冰雹天气过程的云体内部相对运动比较剧烈、层次不是很分明且不同部分之间过渡缓慢,这就使已有的形态特征提取方法以及运动体跟踪方法对雹云而言不十分有效。
本文就雹云的自动识别问题在如下方面提出新的思路和解决方案并取得满意结果:
1.首先,对冰雹云雷达图像展开模式分析,并就典型雹云的特殊形态和特定组成构建了一种三层模型,即云体核、以核为中心的主体层和主体外侧的过渡层。在这三层模型中,其核的强度高、主体层呈钩状、过渡层极窄。
2.为提取雹云模型中主体层钩状特征和过渡层的梯度特征,构建了一种“探针”模型,并成功设计出基于“探针”模型的雹云形态特征和梯度特征的提取算法;进而,采用支持向量机的方法训练出用以识别雹云的分类模型,该模型在测试集中表现出来的较高的识别率表明雹云模型的合理性和所提特征的有效性。
3.为尽可能提早预报,从雹云发生发展规律的角度展开了对时间序列图像的跟踪分析,提出了一个能够提取雷达反射率图中冰雹云动态发展规律的算法模型,与上述特征提取思路不同,该模型主要反映云体的移动速度、方向以及云体的发展、碰撞及复合等。
4.将回波反射率图数据与径向速度图数据相融合,提出了一种去除超折射噪声的简洁而有效的方法。
以上研究所涉及算法均已编程实现,并在样本测试过程中表现出较强的稳定性和良好的测试效果。
Hailstone is a main disaster in the Meteorological area. The hailstorm arises suddenly, moves fast, and disappears very quickly. Now, weather operators monitor it and make hailstone forecast mainly by reflectivity images of Doppler radar. The development of the hailstone is of regularity that the hailstone’s reflectivity intensity and its morphological features make them difference, which help weather operators make forecast. These features are gist of the research on the hailstone automatic detection based on hailstone’s radar image, which is done in this dissertation.
Recently, the research on hailstone’s detection is mainly throw its light on mechanics analysis, forecasting factor and subjective model. There is no similar method to the one described in this dissertation reported yet.
The automatic analysis of the images’morphological character is mainly used in medical field with objects to be the rigid body which have rigid fixing shape and little movement among parts of it, but the radar reflectivity image of hail clouds is different. In the hail cloud, movement among parts of which is violent and the interfaces of these parts are not very clear, which makes traditional morphological character extraction and the moving object tracing methods inefficient with the hail cloud.
In this dissertation, some new ideas and methods are brought forward.
1. With the pattern analysis of hail cloud in radar reflectivity, a model is constructed in which the image of cloud is separated into three layers: core, body and the transition band around of the body according to typical hail cloud reflectivity image. As to the hail clouds, they have some visual features that the reflectivity of the core is very high, the body is hook-like and the transition band is thin.
2. To extraction these features of the model constructed, another model named“probe”is introduced. A method for feature extraction, including the gradient information of the transition band and the attribute about its morphological character especially those about the hook-like body, is proposed based on the“probe”model for hail auto-forecasting. Furthermore, a Support Vector Machine classifier is trained with features extracted from the probe model. The classifier is shown that its accuracy is satisfactory, which testify the efficiency of the method in this dissertation.
3. For predicting the hailstone as early as most, the process of the hailstone clouds development is researched using the time series method and a corresponding algorithm model is constructed. In the algorithm model the velocity and direction of cloud’s movement, the separation and the combination in hail clouds developing process are taken into account.
4. A method is put forward to filter the super refraction echo in reflectivity image by using both of reflectivity image and radial speed image.
All algorithms are coded with C++ language. It is stable and effective, which is testified by test.
引文
[1]辽宁省气象局,辽宁省气象志,辽宁省气象出版社,2002
[2]游景炎,华北冰雹发生条件的合成对比分析,强对流天气文集,气象出版社,1983
[3]隋东,沈阳地区一次冰雹强对流天气过程形成机制的数值模拟:[硕士学位论文],南京:南京气象学院,2003
[4]Romero R.Observations and fine grid simulation of a convective outbreak in northeastern Spain: Importance of diurnal forcing and convective cold pools. Monthly weather revies,2001,129(9),2157~2182
[5]Fiuley C.A.Numerial simulation of tornadogenesis in a high-precipitaton supercell Part I:Storm evolution and transition into a bow echo. Journal of theatmaspheric sciences,2001,58(13),1597-1679
[6] Auer Jr., A. H., 1994: Hail Recognition through the Combined Use of Radar Reflectivity and Cloud-Top Temperatures, Mon. Wea, Rev,122, 2218-2221
[7]段勇,贺敬安,冰雹云雷达回波自动识别系统,成都信息工程学院学报,2001,16 卷,2 期,110-116
[8]王耀生,人工智能、模式识别在气象领域应用的现状与展望,气象人工智能专辑,1994
[9]中国人民解放军总参谋部气象局,多普勒天气雷达资料分析与应用,北京:解放军出版社,2000.1
[10]张培昌,杜秉玉,戴铁丕,雷达气象学,北京:气象出版社, 2001.1
[11]Rafael C. Gonzalez, Richard E. Woods, 数字图像处理(阮秋琦,阮宇智等译),北京:电子工业出版社,2003
[12]王润生,图像理解,长沙:国防科技大学出版社,1995:184～207
[13] F H Freeman. On the encoding of arbitrary geometric configurations[J]. IRE Trans Electron Comput, 1961: EC-10: 260~268
[14] C T Zahn, R Z Roskies. Fourier descriptors for plane closed curves[J]. IEEE Trans on Computers, 1972: 21: 269~81
[15]王伟智,刘秉瀚,郑智勇,基于形态特征的淋巴瘤病理图像分级分割方法,中国体视学与图像分析,2005,10 卷,1 期,61-65
[16]汪源源,沈嘉琳,王涌,王怡,基于形态特征判别超声图像中乳腺肿瘤的良恶性,光学精密工程,2006,14 卷,2 期,333-340
[17]HAO Bao-qing, YIN Guang-fu, QIN Tian-ying, WANG Yan, Study on erythrocyte morphological features by computerized image processing, Journal of Southeast University of Nationality ? Natural Science Edition, 2005, 31(3), 390-395
[18]孙即祥等,模式识别中的特征提取与计算机视觉不变量,北京:国防工业出版社,2001:115～172
[19]边肇祺,张学工等,模式识别,北京:清华大学出版社,2000.1
[20]Vladimir N. Vapnik,统计学习理论的本质(张学工译),北京:清华大学出版社,2000.9
[21] Chih-Chung Chang and Chih-Jen Lin, LIBSVM : a library for support vector machines, 2001. Software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm
[22] Keerthi, S. S. and C.-J. Lin (2003). Asymptotic behaviors of support vector machineswith Gaussian kernel. Neural Computation 15 (7), 1667–1689.
[23] Lin, H.-T. and C.-J. Lin (2003). A study on sigmoid kernels for SVM and the training of non-PSD kernels by SMO-type methods. Technical report, Department of Computer Science and Information Engineering, National Taiwan University. Available at http://www.csie.ntu.edu.tw/~cjlin/papers/tanh.pdf.
[24] Chih-Wei Hsu, Chih-Chung Chang, and Chih-Jen Lin. A Practical Guide to Support Vector Classification, 2001.
[25] Michie, D., D. J. Spiegelhalter, and C. C. Taylor (1994). Machine Learning, Neural and Statistical Classification. Englewood Cliffs, N.J.: Prentice Hall. Data available at http://www.ncc.up.pt/liacc/ML/statlog/datasets.html.
[26]Fiuley C.A.Numerial simulation of tornadogenesis in a high-precipitaton supercell Part I:Storm evolution and transition into a bow echo. Journal of theatmaspheric sciences,2001,58(13),1597-1679
[27]Anil K. Jain, Robert P. W. Duin, and Jianchang Mao, "Statistical Patern Recognition: A Review", IEEE Transaction on Pattern Analysis and Machine Intelligence, Vol.22, No. 1, 2000.1
[28] Chun-Hung Lin and Ja-Ling Wu, Automatic Facial Feature Extraction by Genetic Algorithms, IEEE Transactions on Image Processing, vol. 8, No. 6, june, 1999, 834-844
[29]李智勇,动态图像分析,北京:国防工业出版社,1999
[30]郑南宁,计算机视觉与模式识别,北京:国防工业出版社,1998
[31]王天珍.计算机视觉研究发展.武汉汽车工业大学学报,1998 年 2 月:56-61
[32]Brooks, H. E. Dowell C. A. The role of midtroposphric winds in the evolution and maintenance of low-level mesocyclones. Mon. Wea.Rev,1994,122,126-136
[33]Liwei Wang, Xiao Wang, and Jufu Feng, On Image Matrix Based Feature Extraction Algorithms, IEEE Transactions on Systems, Man, And Cybernetics—part B: CYBERNETICS, VOL. 36, NO. 1, FEBRUARY 2006,194-197
[2]游景炎,华北冰雹发生条件的合成对比分析,强对流天气文集,气象出版社,1983
[3]隋东,沈阳地区一次冰雹强对流天气过程形成机制的数值模拟:[硕士学位论文],南京:南京气象学院,2003
[4]Romero R.Observations and fine grid simulation of a convective outbreak in northeastern Spain: Importance of diurnal forcing and convective cold pools. Monthly weather revies,2001,129(9),2157~2182
[5]Fiuley C.A.Numerial simulation of tornadogenesis in a high-precipitaton supercell Part I:Storm evolution and transition into a bow echo. Journal of theatmaspheric sciences,2001,58(13),1597-1679
[6] Auer Jr., A. H., 1994: Hail Recognition through the Combined Use of Radar Reflectivity and Cloud-Top Temperatures, Mon. Wea, Rev,122, 2218-2221
[7]段勇,贺敬安,冰雹云雷达回波自动识别系统,成都信息工程学院学报,2001,16 卷,2 期,110-116
[8]王耀生,人工智能、模式识别在气象领域应用的现状与展望,气象人工智能专辑,1994
[9]中国人民解放军总参谋部气象局,多普勒天气雷达资料分析与应用,北京:解放军出版社,2000.1
[10]张培昌,杜秉玉,戴铁丕,雷达气象学,北京:气象出版社, 2001.1
[11]Rafael C. Gonzalez, Richard E. Woods, 数字图像处理(阮秋琦,阮宇智等译),北京:电子工业出版社,2003
[12]王润生,图像理解,长沙:国防科技大学出版社,1995:184～207
[13] F H Freeman. On the encoding of arbitrary geometric configurations[J]. IRE Trans Electron Comput, 1961: EC-10: 260~268
[14] C T Zahn, R Z Roskies. Fourier descriptors for plane closed curves[J]. IEEE Trans on Computers, 1972: 21: 269~81
[15]王伟智,刘秉瀚,郑智勇,基于形态特征的淋巴瘤病理图像分级分割方法,中国体视学与图像分析,2005,10 卷,1 期,61-65
[16]汪源源,沈嘉琳,王涌,王怡,基于形态特征判别超声图像中乳腺肿瘤的良恶性,光学精密工程,2006,14 卷,2 期,333-340
[17]HAO Bao-qing, YIN Guang-fu, QIN Tian-ying, WANG Yan, Study on erythrocyte morphological features by computerized image processing, Journal of Southeast University of Nationality ? Natural Science Edition, 2005, 31(3), 390-395
[18]孙即祥等,模式识别中的特征提取与计算机视觉不变量,北京:国防工业出版社,2001:115～172
[19]边肇祺,张学工等,模式识别,北京:清华大学出版社,2000.1
[20]Vladimir N. Vapnik,统计学习理论的本质(张学工译),北京:清华大学出版社,2000.9
[21] Chih-Chung Chang and Chih-Jen Lin, LIBSVM : a library for support vector machines, 2001. Software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm
[22] Keerthi, S. S. and C.-J. Lin (2003). Asymptotic behaviors of support vector machineswith Gaussian kernel. Neural Computation 15 (7), 1667–1689.
[23] Lin, H.-T. and C.-J. Lin (2003). A study on sigmoid kernels for SVM and the training of non-PSD kernels by SMO-type methods. Technical report, Department of Computer Science and Information Engineering, National Taiwan University. Available at http://www.csie.ntu.edu.tw/~cjlin/papers/tanh.pdf.
[24] Chih-Wei Hsu, Chih-Chung Chang, and Chih-Jen Lin. A Practical Guide to Support Vector Classification, 2001.
[25] Michie, D., D. J. Spiegelhalter, and C. C. Taylor (1994). Machine Learning, Neural and Statistical Classification. Englewood Cliffs, N.J.: Prentice Hall. Data available at http://www.ncc.up.pt/liacc/ML/statlog/datasets.html.
[26]Fiuley C.A.Numerial simulation of tornadogenesis in a high-precipitaton supercell Part I:Storm evolution and transition into a bow echo. Journal of theatmaspheric sciences,2001,58(13),1597-1679
[27]Anil K. Jain, Robert P. W. Duin, and Jianchang Mao, "Statistical Patern Recognition: A Review", IEEE Transaction on Pattern Analysis and Machine Intelligence, Vol.22, No. 1, 2000.1
[28] Chun-Hung Lin and Ja-Ling Wu, Automatic Facial Feature Extraction by Genetic Algorithms, IEEE Transactions on Image Processing, vol. 8, No. 6, june, 1999, 834-844
[29]李智勇,动态图像分析,北京:国防工业出版社,1999
[30]郑南宁,计算机视觉与模式识别,北京:国防工业出版社,1998
[31]王天珍.计算机视觉研究发展.武汉汽车工业大学学报,1998 年 2 月:56-61
[32]Brooks, H. E. Dowell C. A. The role of midtroposphric winds in the evolution and maintenance of low-level mesocyclones. Mon. Wea.Rev,1994,122,126-136
[33]Liwei Wang, Xiao Wang, and Jufu Feng, On Image Matrix Based Feature Extraction Algorithms, IEEE Transactions on Systems, Man, And Cybernetics—part B: CYBERNETICS, VOL. 36, NO. 1, FEBRUARY 2006,194-197