用高分一号卫星数据识别多时相山区积雪

发布时间：2018-04-15 11:00

本文选题：高分一号 + 多视图　；参考：《遥感信息》2017年02期

【摘要】：针对目前从高分辨率遥感图像中提取多时相山区积雪信息精度较低、效率不高的问题,基于高分一号卫星数据,提出了一种基于多视图的多时相山区积雪提取方法。该方法基于GF-1卫星3个时相的图像数据,将多时相遥感图像视为多个视图,通过空间约束构建积雪识别多视图。针对山区阴影的巨大影响的问题,提出了将积雪分为非阴影区积雪和阴影区积雪2个类别,分别进行训练样本的选取。仅通过一次样本选取,运用旋转森林算法,基于多视图构建面向多时相的识别模型,实现多时相遥感图像的积雪识别。实验结果表明,3个时相识别结果的F值分别达到0.941、0.951和0.945,精度较高,且具有较高的效率,具有实际应用价值。
[Abstract]:To solve the problem of low accuracy and low efficiency of extracting snow information from high resolution remote sensing images, a multi-view based snow extraction method for multi-temporal mountain areas is proposed based on the high-resolution satellite data.Based on the image data of three phases of GF-1 satellite, the multitemporal remote sensing image is regarded as multiple views, and the snow cover recognition multi-view is constructed by spatial constraints.In order to solve the problem of the great influence of shadow in mountain area, the snow cover is divided into two categories: the non-shadow region and the shadow area, and the training samples are selected respectively.A multi-temporal recognition model based on multi-view is constructed by selecting only one sample and using rotating forest algorithm to realize snow cover recognition in multi-temporal remote sensing images.The experimental results show that the F values of the three phase recognition results are 0.941 / 0.951 and 0.945 respectively, which are of high precision and high efficiency, and have practical application value.
【作者单位】：南京大学地理信息科学系;江苏省地理信息资源开发与利用协同创新中心;莫纳什大学土木工程系;
【基金】：国家高分辨率对地观测系统重大专项(95-Y40B02-9001-13/15-04) 国家自然科学基金(41271353)
【分类号】：TP751

【相似文献】