基于无人机多光谱影像特征的最佳波段组合研究

发布时间：2018-01-01 05:20

本文关键词：基于无人机多光谱影像特征的最佳波段组合研究　出处：《农业机械学报》2016年03期 　论文类型：期刊论文

【摘要】：针对卫星遥感影像分辨率低、时间周期长、波段冗余信息多等问题,利用无人机多光谱数据获取便捷、成本低、周期短的优势,以玛纳斯河畔为研究区,使用固定翼无人机搭载Micro MCA12 Snap多光谱传感器获取高分辨率多光谱影像。通过对多光谱影像数据标准差及相关性进行分析排序,结合OIF方法得到原始波段最佳组合,使用多种植被及水体指数、主成分分析、灰度共生矩阵确定信息量最大的光谱特征与纹理特征波段,提出将光谱特征、纹理特征信息与最佳波段指数结合的方法来确定地物分类最佳波段组合。实验结果表明,针对Micro MCA12 Snap多光谱传感器,可选择波段1、6、11、NDVI、NDWI以及灰度共生矩阵中的Mean参量作为其地物分类的最佳波段组合。感兴趣区域内非监督IsoData分类精度从83.57%提升到89.80%,监督的SVM分类精度从95.58%提升到99.76%。研究结果可为无人机多光谱遥感最佳波段组合选择提供借鉴和参考。
[Abstract]:Aiming at the problems of low resolution, long time period and redundant information in the band of satellite remote sensing image, using UAV multispectral data to obtain convenient, low cost and short period, the Manas River is used as the research area. Micro MCA12 Snap multispectral sensor was used to obtain high resolution multispectral images. The standard deviation and correlation of multispectral images were analyzed and sorted. Combined with the OIF method, the best combination of the original bands was obtained. The spectral and texture bands with the largest amount of information were determined by using various vegetation and water body indices, principal component analysis and gray level co-occurrence matrix, and the spectral features were proposed. The method of combining texture feature information with the best band index is used to determine the best band combination of ground objects classification. The experimental results show that the multi-spectral sensor is aimed at Micro MCA12 Snap. The NDVI can be selected in the frequency band of 1 ~ (6) C ~ (11). NDWI and Mean parameters in gray level co-occurrence matrix are the best band combination for ground object classification. The accuracy of unsupervised IsoData classification in region of interest is improved from 83.57% to 89.80. %. The supervised SVM classification accuracy is improved from 95.58% to 99.76.The results can be used for reference and reference for the selection of optimal band combination of UAV multispectral remote sensing.
【作者单位】：石河子大学信息科学与技术学院;兵团空间信息工程技术研究中心;
【基金】：国家自然科学基金项目(31260291) 新疆生产建设兵团科技计划项目(2015BA006)
【分类号】：TP751;S251
【正文快照】： 引言随着无人机(Unmanned aerial vehicle,UAV)技术和轻型多光谱相机的发展,基于无人机的低空多光谱遥感越来越被人们所关注。以无人机为飞行平台的遥感系统[1]具有起降灵活、低空飞行、低成本、易维护[2-3]的特点,可获得高空间分辨率、高光谱分辨率及高时间分辨率的多光谱影

【共引文献】