GMS卫星红外云图强对流云团的识别与追踪

利用ＧＭＳ静止气象卫星探测所得到的红外云图资料，用计算机图像学技术来研究经对流云图的识别和追踪，以求能够准确，及时地发现并追踪象强对流云团这种能够造成强烈灾害的天气系统。对红我云图的图象处理，提出区域平滑滤波和阈值剔除相结合的强对流云团算法，对于过滤出的经对流云团，应用图象处轮廓编码法的Ｔ算法和ＩＰ算法提取出它的边界，边界初始点。     

图像特征检测与匹配方法研究综述

几十年来,图像特征检测与匹配一直是图像处理的最核心领域之一,是计算机视觉的基石.没有特征检测与匹配就没有SLAM、Sfm、AR、通用图像检索、图像配准、全景图像等视觉任务.本文在回顾几十年来的经典检测算法的基础上,阐述了引用最新的以深度学习为首的机器学习算法后,在本领域取得的最新进展,包括特征点、局部特征子、全局特征子、匹配及优化、端到端框架等所有关键点,展示了算法各自的优缺点.总而言之,面对工业界的宽基线、实时、低算力检测的要求,图像特征检测和匹配仍然是一项未能完整攻克的任务,融合特征点、局部特征子、全局特征子、匹配及优化的多任务全局框架成为未来发展的趋势.     

多普勒天气雷达下击暴流图像识别

以下击暴流具有的低层显著辐散特征为基础,引入图像识别中的连通区识别技术,开发出了应用于多普勒天气雷达的下击暴流图像识别算法。首先通过设置的速度阈值将径向速度进行二值化处理,然后运用8邻域法寻找速度大值区,并采用距离、夹角和正、负速度差值等条件进行约束对正、负速度大值区进行配对,最后对未成功配对的速度大值区进行邻近区域的二次匹配,从而识别出下击暴流区域。利用多个下击暴流个例实测的多普勒雷达数据对该算法进行了测试,结果表明该算法对一些较小尺度的下击暴流,尤其对受环境风场影响而具有不对称辐散特征的下击暴流有良好的识别效果。     

A Test Pattern Identification Algorithm and Its Application to CINRAD/SA(B) Data简

A variety of faulty radar echoes may cause serious problems with radar data applications,especially radar data assimilation and quantitative precipitation estimates.In this study,＂test pattern＂ caused by test signal or radar hardware failures in CINRAD （China New Generation Weather Radar） SA and SB radar operational observations are investigated.In order to distinguish the test pattern from other types of radar echoes,such as precipitation,clear air and other non-meteorological echoes,five feature parameters including the effective reflectivity data percentage （Rz）,velocity RF （range folding） data percentage （RRF）,missing velocity data percentage （RM）,averaged along-azimuth reflectivity fluctuation （RNr,z） and averaged along-beam reflectivity fluctuation （RNa,z） are proposed.Based on the fuzzy logic method,a test pattern identification algorithm is developed,and the statistical results from all the different kinds of radar echoes indicate the performance of the algorithm.Analysis of two typical cases with heavy precipitation echoes located inside the test pattern are performed.The statistical results show that the test pattern identification algorithm performs well,since the test pattern is recognized in most cases.Besides,the algorithm can effectively remove the test pattern signal and retain strong precipitation echoes in heavy rainfall events.     

基于判别式字典的正则化稀疏表示人脸识别算法

为了克服非约束性(光照、表情变化)条件下会大大降低人脸识别率的缺陷,提出一种基于Fisher判别准则的正则化稀疏表示人脸识别算法.首先将人脸图像经过Gabor滤波器滤波得到Gabor幅值图像,提取其统一化的局部二进制直方图,然后利用Fisher判别准则学习得到新的字典,最后通过正则化的稀疏表示判断测试图像所属类.利用AR数据库的数据进行实验的结果表明,与SRC、FDDL、RSC识别算法相比,本文算法在非约束性条件下具有最佳的识别率.     

利用云导风矢量的台风中心自动定位

文中提出了一种利用云导风矢量图进行台风中心自动定位的方法。云导风矢量通过对具有一定时间间隔的两幅相关卫星云图进行模板匹配而得出。模板匹配就是根据已知模式的图像在一幅陌生图像中寻找对应于该模式的子图像的过程。根据气象知识 ,台风云系运动的特点是中心平移量大而自旋很弱 ,且台风中心与台风云系整体的移动方向一致 ,因此求出云导风矢量图中与台风整体移动矢量大小和方向一致的矢量集合的最密集区 ,经过数学形态学处理后便得到台风中心     

The role of raindrop coalescence and breakup in rainfall modeling

In a one dimensional model, the microphysical properties of rainwater content (M), rainfall rate (R) and radar reflectivity factor (Z) are examined in five numerical experiments. In general, neglecting both coalescence and breakup produces less than 10% of the errors in M and R within the top 2 km of a rainshaft (below cloud base). This is true regardless of the rain intensity and downdraft strength for Mashall-Palmer distributions at the top boundary. However, as the distance from the top boundary increases, the errors of M, R and Z calculated without including the collisional processes become larger. These errors tend to increase as the rain intensity increases or the downdraft decreases. The errors due to neglecting breakup are generally larger than those resulting from the exclusion of both coalescence and breakup. At lower levels, the errors for Z are usually very large unless the rainfall is very light (less than 10 mm h⁻¹ ). For heavy rainfall with strong downdraft, coalescence and breakup can be neglected for fall distances greater than 2 km without appreciable errors. On the other hand, for heavy rainfall in a weak downdraft, coalescence and breakup must be included to ensure accuracy.     

Tropical cyclone track predictability

Summary In this study an attempt is made to estimate theinherent limits to tropical cyclone mean absolute track position errors out to 72 hours ahead and to compare these estimates with the position errors currently being obtained inpractice at weather centres around the world. A knowledge of the magnitude of the difference between the lower limit to predictability and that being achieved with state-of-the-art numerical weather prediction (NWP) models is of vital importance. A small difference would indicate that there is little further need for continued initiatives in the prediction of tropical cyclone tracks. On the other hand, a large difference would imply either that the problem requires continued emphasis or if there has been no significant trend towards reducing the forecast track errors, that present research and development techniques need to be extended or new procedures developed.It was found that the difference between the inherent and practical limits of tropical cyclone track position errors is presently about 35 to 40 per cent for advanced baroclinic NWP systems, a moderate to large difference, and one that is almost invariant between tropical cyclone basins. For simpler models, such as barotropic models, the difference is closer to 45 per cent but is again almost invariant. As far as the authors are aware, these are the first estimates of the lower bounds of tropical cyclone track predictability. Finally, very recent research studies with emerging range of high quality data, high density data sources, improved models and new data assimilation techniques suggest that the difference possibly is now down to about 30 to 35 per cent. This value is encouragingly small but still large enough to continue active research programs in improving tropical cyclone motion prediction. Much of the forecast track errors now come from major forecast errors associated with tropical cyclones that follow erratic tracks.With 3 Figures     

基于支持向量机的雷达地物回波识别研究

多普勒天气雷达探测过程中的非气象因子会显著影响雷达资料的定量化应用,在雷达基数据的应用前需对雷达资料进行抑制地物杂波、去距离折叠和退速度模糊等质量控制。本文在现有的自动识别地物回波方法的基础上,提出了基于支持向量机（Support Vector Machine,SVM）识别雷达地物杂波的方法,2013年6-8月对安庆和常州两地的CINRAD/SA雷达观测资料进行雷达地物回波识别,并将其与运用人工神经网络（Artificial Neural Networks,ANNs）识别的结果进行对比,结果表明支持向量机方法能够取得更好的效果。在地物、降水回波总样本识别和地物回波识别方面更为有效;降水回波的误判方面,神经网络略优于支持向量机,但两者差异不大,都将降水回波的误判率控制在了一个较小的范围内;另外支持向量机方法较之神经网络方法对于训练样本数目的依赖性较小,在训练样本较少时,支持向量机方法仍能保持有效的识别效果。     

变分法和卫星云图模式识别在强降水面雨量预报中的应用

假设数值预报对雨量的落区预报是准确的，用卫星云图模式识别预报的流域雨量极值，应用变分法对数值预报雨量极值进行订正，使得订正预报与未订正场之间的泛函达最小，从而求得预报场。试验证明，该方法对强降水面雨量预报的准确率较数值预报有一定提高。     

ENERGETIC DIAGNOSIS FOR TWO KINDS OF LOW LEVEL JETS

On the basis of the budget equations for K_R and K_D, this paper presents the horizontal pattern of budget terms for two kinds of low level jets (LLJ) with and without heavy rain. The results show that the mechanisms for generating and maintaining LLJ are different, and especially, the direction of energy conversion is opposite. A positive conversion from K_D to K_R appears to be a necessary but not sufficient condition in the lower troposphere near the heavy rain area. The intensity and direction of energy conversion depends not only on the relative position of vorticity and divergence field, but also on the vertical profile of the jets directly.     

用双通道动态阈值对GMS-5图像进行自动云检测

该文用双通道动态阈值法对GMS-5图像进行自动的云检测。在红外通道和可见光通道，分别对每32×32个像元组成的像元阵进行直方图统计，求出区别云和地物值域的阈值，然后对每个像元阵内的逐个像元进行云判识。讨论了在进行直方图分析以求取阈值的过程中，像元阵大小的选取和动态平滑间距的选取对云检测结果的影响。分析结果表明:像元阵大小取32×32时，像元阵所占的地域空间尺度足够小，像元阵内的观测像元样本数足够多，保证了在直方图聚类分析时，每一类含有足够多的样本；GMS-5观测图像的红外通道对原始直方图进行二次平滑时小平滑间距取1.6 K，可见光通道小平滑间距取1.2%，使得确定动态阈值时步长相对小，保证了分析的精度。用目视图像对分析结果进行真实性检验，在中低纬度地区，可见光和红外两个通道都有资料时，该算法的云判识精度较好。在高纬度地区，由于地表温度低，积雪覆盖多，太阳光照角低，该算法的云判识精度较差。     

不同类型降水回波的自动识别方法

讨论了不同类型降水回波的自动识别方法。根据不同类型降水回波的特征,选择了４个识别因子——面积复杂度、强度梯度、强中心强度和面积因子,形成一个可以自动识别３种类型降水回波的３层判别系统。     

一种基于模糊判别的指纹识别算法

利用模糊判别技术提出了一种指纹识别方法，改进了Anil Jain等提出的细节匹配算法，有效地解决了图像质量很差时经常检测到很多伪细结点而丢掉很多真细结点、细节匹配难有大作为的情况。该算法有较强的抵抗噪声与非线性形变的能力。     

Development and validation of an evaporation duct model. Part II: Evaluation and improvement of stability functions

This study aims to validate and improve the universal evaporation duct (UED) model through a further analysis of the stability function (ψ). A large number of hydrometeorological observations obtained from a tower platform near Xisha Island of the South China Sea are employed, together with the latest variations in ψ function. Applicability of different ψ functions for specific sea areas and stratification conditions is investigated based on three objective criteria. The results show that, under unstable conditions, ψ function of Fairall et al. (1996) (i.e., Fairall96, similar for abbreviations of other function names) in general offers the best performance. However, strictly speaking, this holds true only for the stability (represented by bulk Richardson number R _iB) range ?2.6 ? R _iB < ?0.1; when conditions become weakly unstable (?0.1 ? R _iB < ?0.01), Fairall96 offers the second best performance after Hu and Zhang (1992) (HYQ92). Conversely, for near-neutral but slightly unstable conditions (?0.01 ? R _iB < 0.0), the effects of Edson04, Fairall03, Grachev00, and Fairall96 are similar, with Edson04 being the best function but offering only a weak advantage. Under stable conditions, HYQ92 is the optimal and offers a pronounced advantage, followed by the newly introduced SHEBA07 (by Grachev et al., 2007) function. Accordingly, the most favorable functions, i.e., Fairall96 and HYQ92, are incorporated into the UED model to obtain an improved version of the model. With the new functions, the mean root-mean-square (rms) errors of the modified refractivity (M), 0–5-m M slope, 5–40-m M slope, and the rms errors of evaporation duct height (EDH) are reduced by 21.65%, 9.12%, 38.79%, and 59.06%, respectively, compared to the classical Naval Postgraduate School model.

     

Q矢量的改进与完善

在完全考虑非绝热加热项潜热作用的前提下，从原始方程出发，推导出改进后的湿Q矢量(以下记为Q^M)的表达式以及用其散度作强迫项的ω方程，并结合一次梅雨锋暴雨天气过程将改进前、后的湿Q矢量诊断能力进行比较，结果发现：(1)改进后的湿Q矢量对同时刻的地面降水的反映能力较原湿Q矢量(Q^*)有显著的提高。(2)在整个梅雨锋暴雨过程中，600hPa高度上的改进后的湿Q矢量散度辐合场的辐合强度及其辐合中心的位置对同时刻的地面实际降水的强度及雨区位置都有非常好的指示作用。     

双线偏振多普勒雷达资料质量分析

以中国气象科学研究院灾害天气国家重点实验室的车载C波段双线偏振多普勒雷达（CPDRW）2011年完成升级改造后观测的一次大范围层状云降水过程资料为例,讨论了该设备改造后探测资料质量的改进情况及其对降水粒子的探测性能,同时利用模糊逻辑法结合地物的常规特性和偏振特性,建立了一种运用双线偏振雷达观测资料识别地物及地物与降水混合回波的方法,并利用该雷达的观测资料对识别结果进行了分析检验,最后统计分析了降水回波的差分反射率因子(Z_DR)、零滞后互相关系数(ρ_HV（0）)与信噪比的关系,以及系统初始相位的稳定性及降水资料的一致性。分析结果表明,在常规参量的基础上添加偏振参量并采用模糊逻辑法构建的综合识别模式和常规识别模式都能较好地识别地物,但综合识别模式明显改善了常规识别模式0速度区过度识别的问题,且在地物与降水回波混合的情况下叠加混合回波识别模式具有更加明显的优势。雷达改造前,Z_DR及ρ_HV（0）在信噪比小于25 dB时受噪声影响很大,测量值存在较大误差,资料变得不可信;通过这次的升级改造,其信噪比探测阈值由25 dB降低到15 dB。系统初始相位比改造前更加稳定,随仰角和时间的波动性更小。利用垂直扫描数据进行系统误差订正后的Z_DR资料能较好地反映降水粒子的实际情况,Z_DR随信噪比的变化较之改造前更加平稳,而对差传播相移Φ_DP进行分类处理后拟合的差传播相移率K_DP资料也与实际情况一致。     

雷达数据外推与特征识别的下击暴流预警方法

现今利用普通气象资料对下击暴流进行识别预警的可行性很低,因此提出一种基于多普勒天气雷达资料的下击暴流识别追踪和预警方法。首先,采用光流法对反射率因子垂直剖面的光流场进行分析,得出风暴核心随时间演变的规律。然后,采用拉格朗日力学模型对风暴核心顶高的下降过程进行函数拟合。再采用直方图和巴氏系数统计分析法,对风暴核心中层径向速度场中的"正负速度对"图像进行匹配识别,综合一系列阈值的判定,最终实现下击暴流的智能预警。提出基于多层迭代的局部约束光流算法,有效改善了传统光流法对回波这类非刚体移动目标的不适用性。提出直方图和巴氏系数统计分析法,解决了因径向"正负速度对"图像的非对称结构而造成的图像匹配难题。个例检验结果表明,该方法可在风暴核心抬升和急速下沉的前期识别出潜在的下击暴流,并对风暴核心下沉速率和到达近地面的时间做出估测,从而实现下击暴流的智能识别、预警。     

Drag and heat transfer relations for the planetary boundary layer

A theory is offered for the drag and heat transfer relations in the statistically steady, horizontally homogeneous, diabatic, barotropic planetary boundary layer. The boundary layer is divided into three regionsR ₁,R ₂, andR ₃, in which the heights are of the order of magnitude ofz ₀,L, andh, respectively, wherez ₀ is the roughness length for either momentum or temperature,L is the Obukhov length, andh is the height of the planetary boundary layer. A matching procedure is used in the overlap zones of regionsR ₁ andR ₂ and of regionsR ₂ andR ₃, assuming thatz ₀ L h. The analysis yields the three similarity functionsA(),B(), andC() of the stability parameter, = u _*/fL, where is von Kármán's constant,u _* is the friction velocity at the ground andf is the Coriolis parameter. The results are in agreement with those previously found by Zilitinkevich (1975) for the unstable case, and differ from his results only by the addition of a universal constant for the stable case. Some recent data from atmospheric measurements lend support to the theory and permit the approximate evaluation of universal constants.     

Computational parameter estimation for a maize crop

During a whole growing season, the evolution of the displacement height, d, and roughness length, z ₀, of a maize crop has been estimated by a measurement programme. The results have been used to check different types of existing models to calculate these parameters from canopy characteristics only; a simple geometric model and two matching models have been investigated. A geometric model is based on geometric features of the surface only. After a simple modification, the geometric model gives good results for the displacement height as well as for the roughness length.A matching model, based on gradient-diffusion theory, yields good results for the displacement height. The roughness parameter, however, is overestimated by 17%. By a simple modification, the model results could be improved considerably.A matching model, based on a second-order closure procedure, yields excellent results for the displacement height and good results for the roughness length. But it appears that, when applying this model, the plant density index and plant area density distribution as a function of height must be well known.