気象衛星画像の赤外画像と水蒸気画像から計算する高層のオプティカルフロー

概要

論文の詳細を見る
Estimating upper air current velocity is important to fix the initial values in numerical weather forecasting. We tried to estimate optical flow of upper air current from infrared images and water vapor images in 'meteorological satellite (GMS-5) image sequences. An originally developed gradient-based method is adopted to evaluate optical flow under non-uniform and non-stationary illumination. The method combines a spatiotemporal local optimaization and a stabilization method using reliability indices of optical flow. Through comparison between optical flows estimated from water vapor images and from infrared images, we confirmed that vapor images well-reflected upper air current. Upper air current is-estimated from a processed sequence of infrared images by eliminating gray values at respective pixels having a lower gray-level (higher temperature) than a threshold. We confirmed a high correlation between optical flows excepting for winter season. We found that a periodic temperature variation from day to night brought an artifact in optical flow evaluation, especially on a continent. Compensation of the periodic temperature variation brought a higher correlation between optical flows. A possible method is discussed to separate lower air current from upper air one using infrared satellite image sequences.
徳山工業高等専門学校の論文
2004-12-01