基于Landsat 8 OLI遥感影像的沈阳市水稻种植面积提取方法

发布时间：2018-11-19 18:56

【摘要】：为了深入研究遥感数据及提取方法对估算水稻种植面积的可行性,以Landsat 8 OLI影像为数据源,运用ENVI5.1的软件平台,对沈阳市2015年6—9月水稻长势进行监测,并最终提取其种植面积。根据实地调查样本,通过分析各地物的光谱特性曲线、归一化植被指数均值特征及遥感影像成像特点,确定了以波段6、波段5、波段2对图像进行伪彩色合成。对合成后的图像,分时段设计了三组不同样本点数量的对比实验,样本数量分别为100、150、200个,采用混合像元的方式,确定了水稻的采集样本点,用变换分散度和J-M距离对各个样本间的可分离性进行检验,采用支持向量机的分类方法对各样本进行分类,最后以Majority/Minority分析方法对提取的结果进行分类后处理,建立了不同的水稻面积提取模型。结果显示,6月、7月、9月中200个样本点的实验提取结果均较为准确,提取面积分别为1 032.044 8、1 201.125 9和1 180.685 5 km2,参考《沈阳统计年鉴2015》对提取结果进行评价,精度分别为94.73%、89.75%和91.62%。试验表明,Landsat 8OLI遥感数据可准确提取沈阳市水稻种植面积,为综合多源数据对水稻进行种植监测奠定基础。
[Abstract]:In order to study the feasibility of remote sensing data and extraction methods in estimating rice planting area, the rice growth in Shenyang from June to September 2015 was monitored by using Landsat 8 OLI image as data source and ENVI5.1 software platform. Finally, the planting area was extracted. According to the field survey samples, by analyzing the spectral characteristic curve of the local objects, normalized vegetation index mean value characteristics and the imaging characteristics of remote sensing images, the pseudo-color synthesis of the images was determined by using band 6, band 5 and band 2. For the synthesized images, three groups of comparative experiments of the number of different sample points were designed in different periods. The number of samples was 100150200, and the sampling points of rice were determined by using mixed pixel method. The separability of each sample is tested by transform dispersion and J-M distance. Support vector machine (SVM) is used to classify each sample. Finally, Majority/Minority analysis method is used to classify and post-process the extracted results. Different models of rice area extraction were established. The results showed that the extraction results of 200 samples were all accurate in June, July and September. The extraction areas were 1 032.044 8 / 1 201.125 9 and 1 180.685 5 km2, respectively. The results were evaluated with reference to Shenyang Statistical Yearbook 2015. The accuracy is 94.7389. 75% and 92. 62% respectively. The experiment shows that the Landsat 8OLI remote sensing data can accurately extract the planting area of rice in Shenyang and lay a foundation for the comprehensive monitoring of rice planting with multi-source data.
【作者单位】：沈阳农业大学信息与电气工程学院;辽宁省农业信息化工程技术中心;
【基金】：国家重点研发计划(2016YFD020060307)
【分类号】：S127;S511

【相似文献】