【作者】 卢鹏；

【导师】 李志军；

【作者基本信息】 大连理工大学 ， 港口、海岸和近海工程， 2007， 博士

【摘要】 本文以中国第二次北极科学考察和第十九次、第二十一次南极科学考察中获取的海冰现场调查图像资料为基础，结合实验室物理模拟试验，利用图像分析技术，对与海冰动力学模型相关的图像采集与处理、几何特征参数分析及参数化方法等进行深入探讨。文中首先阐述了海冰对全球气候变化和极区、亚极区海域海洋工程问题的重要性，并指出作为海冰研究重要手段之一的海冰数值模拟发展到目前，其准确性更加受到海冰参数而非模式本身的限制，因此以目前快速发展的数字图像技术为工具对海冰动力学过程中的重要参数进行研究，以提供数值模式所需要的高精度和准确性的数据，是非常有必要的。同时，文中还对各重要参数及参数化的研究现状和发展趋势进行了总结。第二章即重点讨论了本文中所用到的数字图像采集技术和图像分析技术。分别介绍了海冰现场观测和实验室物理模拟实验中常用的数字图像采集技术，并应本文后续研究内容的需要对海冰航拍图像的处理进行了重点研究，比较了不同的图像分割算法的优缺点和各自的适用范围，并提出了一种基于图像形态学的具有较高自动化程度和准确性的海冰分割算法，提高了在图像中识别单个海冰的效率。随后的章节则分别对海冰动力学的各重要几何参数进行讨论。第三章重点对海冰厚度和密集度参数进行讨论，主要以中国第二次北极科学考察中利用船侧CCD获取的走航海冰图像和海冰航拍图像为基础，研究从中提取海冰厚度和密集度的方法，通过数据分析得到了北冰洋夏季海冰厚度、雪厚度及密集度均随纬度增大而逐渐增大的结论。同时，将不同尺度上(包括卫星遥感、海冰航拍和船舶走航观测)获取的海冰密集度数据进行对比，发现卫星遥感的海冰密集度数据在中等密集度时明显偏高。最后针对利用双CCD设备在船侧同时进行海冰厚度和密集度观测的方法，给出了修正倾斜拍摄时图像变形的算法，从而可以得到更准确的海冰密集度。论文第四章主要以海冰尺寸分布特征为研究对象。通过对中国第二次北极科学考察期间获取的北冰洋夏季不同纬度的海冰航拍图像的分析，发现了海冰形态随纬度的显著变化；利用幂函数对海冰尺寸分布进行拟合发现分布维数基本保持在1.05～1.27之间，并且随着纬度增加有变大的趋势。另外，通过对中国第二十一次南极科学考察期间获取的Prydz湾夏季海冰边缘区航拍图像的分析也发现海冰的尺寸和形状随着与开阔水距离增大也发生有规律的变化，而这种变化的原因就是边缘区内海冰与波浪的相互作用。在进行海冰尺寸分布分析时发现由于热力学作用和采样区域限制的影响，累积频率分布曲线与理想的幂函数分布有明显的偏离，而改用一种改进的幂函数分布以及Weibull分布来对累积频率进行拟合时则都能获得很好的拟合效果，拟合结果中各参数的变化情况很好的反映了海冰边缘区内破碎海冰的分布变化。同时，通过对中国第十九次南极科学考察在南大洋航行期间利用船舶雷达获取的冰山尺寸和运动数据的分析发现，不同尺寸的冰山由于其所经历的不同热、动力学过程原因而具有不同的尺寸分布特征，Weibull分布只适合于较小尺寸的冰山，而较大尺寸的冰山分布更接近与幂函数分布。通过对冰山漂流速度和当地的风、流速度的回归分析发现，风对冰山运动的贡献是海流贡献的1／2左右。论文第五章主要对冰-水拖曳系数的参数化进行了研究。在回顾已有拖曳系数获取方法(涡动法、剖面法和动量法)的基础上，说明了其局限性和需要改进之处，并提出利用参数化方法建立冰．水拖曳系数同海冰几何形态参数的关系才是提高拖曳系数适应性和准确性的最佳途径。然后以实验室物理模拟试验为基础并利用一种参数优化辨识算法，建立了孤立冰块形拖曳系数、摩拖曳系数和冰块几何参数之间的关系。通过将原型条件下的冰-水总拖曳力考虑成表面摩拖曳力、冰侧形拖曳力和冰脊形拖曳力三部分，建立了冰-水总拖曳系数同海冰和冰脊几何形态参数之间的定量关系，并由计算分析发现现场观测到的拖曳系数随密集度的非单调变化就是冰侧形拖曳力分量作用的重要体现。更多还原

【Abstract】 Based on the sea ice in-situ investigation images obtained during the 2nd Chinese National Arctic Research Expedition and the 19th, 21st Chinese National Antarctic Research Expedition, combing with laboratory physical modelling experiments, image analysis technologies are used in the present dissertation to provide a deep discussion on image acquisition, image processing, parameter analysis and parameterization method buildup related with sea ice dynamic processes.As the beginning of this thesis, the important meanings of sea ice to global climate change and ocean engineering in polar and subpolar sea regions are explained firstly. Then, it is pointed out that sea ice numerical modelling, as a powerful tool of sea ice research, is limited by the precision of parameters within models instead of itself so far. It is meaningful to study the important parameters of sea ice dynamics by means of the rapidly developing digital imaging techniques so as to provide better input parameters needed in numerical models. Moreover, the research conditions and developing trends of the sea ice parameters and parameterization are also summarized there.Techniques of image acquisition and image processing used in this thesis are discussed in the second chapter. The image acquisition techniques used in sea ice in-situ investigation and laboratory physical model experiment are firstly introduced, and then the image processing techniques for sea ice aerial photographing that needed in later chapters are studied. Different image segmentation algorithms are compared and their advantages and disadvantages are pointed out as well as their own applicabilities. Furthermore, a kind of floe separation algorithm based on mathematical morphology is proposed to improve the efficiency of floe identification in sea ice image analysis, and such method is proved to be more automatic and exact through many applications.Some important parameters of sea ice dynamics are then discussed in the follow parts. The third chapter is focus on sea ice thickness and concentration. Based on sea ice images captured by ship-based CCD camera and aerial photographing during the 2nd Chinese National Arctic Research Expedition, the methods of obtaining ice thickness and concentration from such images are firstly introduced, and then it is found that ice/snow thickness and ice concentration increase with the increasing latitude in summer Arctic after data analysis. Moreover, ice concentration data obtained at different scales are compared and the result from satellite remote sensing is found to be obviously higher than others at middle values. Besides, to a systemic method of obtaining ice thickness and concentration simultaneously by using two ship-based CCD cameras, the corresponding algorithm modifing the image distortion is also provided to obtain more veracious ice concentrations.The objective of the fourth chapter is floe size distribution. Through analyzing aerial images of sea ice at different latitudes in the 2nd Chinese National Arctic Research Expedition, an obvious variation of morphology of Arctic ice floes with increasing latitude is obtained, and after mean caliper diameter of ice floe is used as characteristic size to analyze floe size distribution, it is found that cumulative probabilities of floe size agree with a power-law function, and distribution dimension is generally in the range of 1.05～1.27 and slightly increases with increasing latitude. While in the analysis of sea ice images captured in the marginal ice zone of summer Prydz Bay during 21st Chinese National Antarctic Research Expedition, obvious variations of floe size and shape parameters with the distance into the marginal ice zone are also observed, and the mainly reason of which is the wave-ice interaction there. But in floe size distribution analysis, large deviations of cumulative distribution of floe size to the ideal power law induced by both the truncation effect of sampling and the considerable thermodynamic effect on ice floes in summer Prydz Bay occur, and so that an upper-trtmcated power-law function and a Weibull function are used in curve fitting and four calculated parameters of alternative functions are confirmed to be important descriptors of the evolution of floe size distribution in the marginal ice zone. Moreover, the iceberg size and drift data obtained during the 19th Chinese National Antarctic Research Expedition is analyzed, and the result shows that icebergs of different size have different size distribution characteristics owing to their different dynamic and thermodynamic experiences. The Weibull function is suitable only for smaller iceberg but size distribution of larger iceberg agrees better with the power-law function. Combining with synchronously collected wind, current velocity and iceberg drift velocity, a kinematic relationship is developed as a multiple linear regression function and the result shows that the contribution of wind to iceberg drift is just half of that of current.In the last chapter, ice-water drag coefficient and corresponding parameterization method is studied. After a general review of measuring methods of drag coefficient (including eddy correlation method, profile method, momentum method), the limitations of such methods are explained and a parameterization method relating ice-water drag coefficient with ice morphological parameters is proposed as a good solution. Then a model to distinguish skin drag coefficient and form drag coefficient is established based on a floe drag experiment by using a wave-current tank in laboratory, and the quantitative relationship between skin/form drag coefficient and floe geometric parameters is established as a result. Afterwards, considering the total ice-water drag comprise of skin friction, form drag induced by ice edge and form drag induced by ice ridge, a quantitative relationship between ice-water drag coefficient and. ice/ridge geometric parameters is established, and from computational analysis it is found that the unmonotonous variations of drag coefficient with ice concentration that frequently observed in sea ice field investigation mostly result from the influence of form drag from ice edge.更多还原