多源遥感数据时空融合模型应用分析
作者简介:邬明权(1983-),男,湖南株洲人,博士,助理研究员,主要从事多源遥感数据时空融合研究。E-mail:wumq@irsa.ac.cn
收稿日期: 2013-11-15
要求修回日期: 2013-12-26
网络出版日期: 2014-09-04
基金资助
国家自然科学基金项目(41301390)
Assessing the Accuracy of Spatial and Temporal Image Fusion Model of Complex area in South China
Received date: 2013-11-15
Request revised date: 2013-12-26
Online published: 2014-09-04
Copyright
多源遥感数据时空融合模型是解决目前遥感数据获取能力不足问题的重要方法之一,当前主要融合方法的研究,集中于平原区域,缺乏复杂条件下的多源遥感数据融合技术的应用研究。针对我国南方复杂条件,本文对比研究了多源遥感数据时空融合模型在我国南方复杂条件下的应用能力。针对LORENZO模型、LIU模型、统计回归模型、STARFM和ESTARFM 5种主流多源遥感数据时空融合模型,采用Landsat-ETM+和MODIS数据,以江苏省南京市的小块区域为实验区,利用5种模型生产融合影像,以真实Landsat-ETM+数据为模板,定性和定量评价融合效果的好坏。结果表明:除LORENZO模型外,其余4种模型获得的融合影像与真实影像之间都具有较高的相关性,相关系数均高于0.6,其中,ESTARFM模型的融合影像与真实影像间的相关性最高,融合效果最好,其次为STARFM模型,再次为LIU模型和统计模型法。在融合过程中采用距离、时间和光谱等信息越多,融合效果越好,在复杂地区的适用能力越强,融合影像更能反映地物的细节特征。
邬明权 , 牛铮 , 王长耀 . 多源遥感数据时空融合模型应用分析[J]. 地球信息科学学报, 2014 , 16(5) : 776 -783 . DOI: 10.3724/SP.J.1047.2014.00776
Due to cloud coverage and obstruction, it is difficult to obtain useful images during the critical periods of monitoring vegetation using medium resolution spatial satellites such as Landsat and Satellite Pour l'Observation de la Terre (SPOT), especially in pluvial regions. A solution for fine-scale vegetation research is to blend the data from both high temporal resolution sensors (e.g., MODIS) and moderate ground resolution satellites (e.g., Landsat) to generate synthetic observations with characteristics of both. In recent decades, several approaches have been proposed to enhance the temporal frequency of high-resolution spatial satellite observations. However, there is a lack of application research of those methods, especially in South China where the climate is complex and the region is scattered with broken terrain. In order to evaluate the application ability of spatial and temporal image fusion models in South China, five spatial and temporal image fusion models were assessed in this paper. The five models are LORENZO model, LIU model, statistical model, STARFM and ESTARFM. Using the Landsat-ETM+ and MODIS data, the five methods were tested in an area near the Nanjing city of Jiangsu Province. QualitativeE:\app:ds:qualitativeevaluationE:\app:ds:evaluation and quantitativeE:\app:ds:quantitativeevaluationE:\app:ds:evaluation methods were used to evaluate the similarity between the simulated images and the real Landsat ETM+ images. Results showed that except Lorenzo model, the other models were able to produce synthetic images very similar to the actual observed images with a correlation coefficients r of higher than 0.6. The more information, such as distance, temporal and spectral information, isused in the image fusion, the synthetic fusion image could better reflect the detailed features of land surface.
Tab.1 Input parameters of models表1 各模型输入参数 |
模型 | 输入数据 | 验证数据 | |||||
---|---|---|---|---|---|---|---|
MODIS数据 | Landsat数据 | 分类影像 | Landsat数据 | ||||
波段 | 数据日期 | 波段 | 数据日期 | 波段 | 数据日期 | ||
LORENZO模型 | MOD02 | 2002-10-08 2002-10-24 | ETM+4 | 2002-10-08 | 30 m分类图 | ETM+4 | 2002-10-24 |
LIU模型 | MOD02 | 2002-10-08 | 15 m分类图 | ETM+4 | 2002-10-08 | ||
统计模型法 | MOD02 | 2002-10-08 | 15 m分类图 | ETM+4 | 2002-10-08 | ||
STARFM | MOD02 | 2002-10-08 2002-10-24 | ETM+4 | 2002-10-08 | ETM+4 | 2002-10-24 | |
ESTARFM | MOD02 | 2002-10-08 2002-10-24 | ETM+4 | 2002-10-08 2002-11-09 | ETM+4 | 2002-10-24 |
Fig.1 outputs of different models and real images图1 各方法融合结果与真实影像 |
Tab.2 Comparison of different model results表2 各模型融合效果比较 |
模型 | R2 | var(a) | MAD | bias | RMSE |
---|---|---|---|---|---|
LORENZO模型 | 0.198 | 0.081 | 0.236 | 0.273 | 0.395 |
LIU模型 | 0.635 | 0.026 | 0.119 | 0.169 | -0.049 |
统计模型法 | 0.641 | 0.026 | 0.121 | 0.165 | -0.034 |
STARFM | 0.752 | 0.019 | 0.105 | -0.050 | 0.146 |
ESTARFM | 0.887 | 0.009 | 0.069 | -0.034 | 0.095 |
Fig.2 Scatter diagram between the fusion image and the real image图2 融合影像与真实影像间的散点图 |
The authors have declared that no competing interests exist.
[1] |
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
|
[9] |
|
[10] |
|
[11] |
|
[12] |
|
[13] |
|
[14] |
|
[15] |
|
[16] |
|
[17] |
|
[18] |
|
[19] |
|
[20] |
|
/
〈 | 〉 |