【作者】 庄立； 于凯；

【Author】 ZHUANG Li;YU Kai;College of Water Sciences, Beijing Normal University;School of Geography, Beijing Normal University;

【机构】 北京师范大学水科学研究院； 北京师范大学地理学与遥感科学学院；

【摘要】 MODIS(Moderate-Resolution Imaging Spectroradiometer)数据以其时间分辨率高、空间覆盖区域广、光谱范围宽、全球免费接收使用等特点成为了应用领域最广、使用量最大的遥感数据。其陆地标准产品中的MOD09反射率产品更是成为了全球变化研究必不可少的数据源。MOD09数据包含1天、8天两种不同时间分辨率的产品,虽然它们经过了一定的处理,但是却仍无法避免受到云、云影、雪、及其他异常气候条件的影响。针对这一问题,本文提出一种新的基于地物光谱随时间变化连续这一特点的MODIS数据时空滤波方法。该方法首先根据时间序列平均值对空间数据进行分类,确定各像元类别;然后在时间维上分段处理,获得各类地物在相应时间段内最连续的光谱变化趋势(本文用线性趋势进行处理),再利用各类地物的变化趋势模版对该时间段内相应的数据进行滤波处理。处理后的数据基本消除了云、云影、雪等异常因素的影响,具有时间上连续平滑变化、空间上自然清晰的特点。应用该方法对中国、蒙古和美国等地区的多景MOD09 8天间隔数据和1天间隔数据进行处理,通过与原始数据的比较证实了该方法的有效性。更多还原

【Abstract】 MODIS data is one of the most popular remote sensing data source for its high temporal resolution, broad spatial scale,wide spectral range and freely supply. Its land standard product MOD09 has become the indispensable data source for global change researches. It contains two kinds of products whose temporal resolution are one day and eight days. Although being properly processed, the cloudy, shadow, snow or other weather conditions could badly affect its quality. To solve this problem, we present a new temporal spatial filter algorithm for MOD09 sequence data based on the transformation continuity of target spectrum. First,classify the MOD09 data by its temporal mean and decide the class of each pixel; Then segment on temporal dimension to obtain the most continuous spectral variation trend of each class(in this paper we use liner trend); Finally, filter each pixel by the spectral variation trend of the corresponding class in temporal segment. The result removes the influence of cloudy, shadow, snow, and becomes continuous smooth in temporal dimension and clear natural in spatial dimension. The comparison between filter result and original data of China, Mongolia and American MOD09 eight days and one day temporal sequences validates the effect of this algorithm.更多还原