奇异值分解与神经网络结合的卫星云图云团移动预测

云团运动和发展演变的预测是暴雨等灾害性天气监测预报的重点和难点问题,针对当前云团预测中缺乏有效的非线性、非平稳预测手段,提出了奇异值分解SVD(singular value decomposition)与径向基网络相结合的云团预测途径.首先用SVD对云图进行分解,提取主要的云团结构特征,然后用提取出的云图奇异特征值和左右奇异向量作为模式识别因子,选择特定区域和季节的云图时滞序列采样,并用前后时段样本云图的奇异值和奇异矢量作为云图预测模型的输入、输出,通过对径向基网络的学习训练和误差迭代收敛,建立了云团运动的非线性预测模型.试验结果表明,该方法能合理地描述云团运动的基本特征和演变趋势.

Cloud cluster movement forecast technique of satellite cloud pictures based on singular value decomposition and artificial neural networks

The forecast of cloud cluster movement and evo lut ion is impor tant , w hich is dif f icul t for inspecting and predicting such disaster weather as rainsto rm. In view of lacking ef fect ive non-linear and non-stable cloud cluster mov ement fo recast technique, a cloud cluster fo recast new technique based on singular value decomposit ion ( Singular Value Decomposit ion, SVD) and radial basis neur al netwo rks w as presented in this paper. First , the satell ite cloud picture w as decomposed by SVD, and it s chief character s were pickedup, then the picked singular characterist ic value and it s rig ht / lef t sing ular character ist ic vector s w er e taken as pat tern identif icat io n factors, the satellite cloud picture time series w ere sampled in specif ied region and season, and the fore-/ back-period samples singular values and singular vectors w ere taken as the input / output o f the forecast mo del. By t raining the AN N mo del, a non-linear for ecast model of cloud cluster mov ement w as establ ished. The result s show that the model ing technique can reasonably describe the basic characters and the evo lvement t rend of cloud cluster .