基于全极化SAR与多光谱的喀斯特山区农村林地提取

发布时间：2018-03-21 13:08

本文选题：全极化　切入点：Radarsat-　出处：《中国农业资源与区划》2017年07期 　论文类型：期刊论文

【摘要】：[目的]为加快推动贵州省"互联网+"林业建设,打破贵州喀斯特高原山区遥感监测瓶颈,选取了空间分辨率8m的Radarsat-2全极化SAR数据与空间分辨率6m的SPOT 6多光谱影像作为数据源,探究微波与光学遥感结合在喀斯特山区农村地区的林地提取技术。[方法]首先采用ENVI SARscape与NEST软件对SAR图像预处理。将Radarsat-2全极化数据与SPOT 6标准假彩色影像进行HSV融合。计算融合图像的平均梯度、信息熵、标准差与均值,评价出最优融合效果的极化方式。基于K均值(K-means)与最大期望(EM聚类)聚类算法分割图像,选择合适的算法,基于聚类分割的阈值进行面向对象的林地分类。最后,基于像素的混淆矩阵精度评价,结合贵州省林业厅调查数据、野外样方和航拍图,建立参考样本评价分类结果。[结果](1)融合之后,目视解译出林地边缘明显但较粗糙;对于在林地中小面积建筑物、农田中的较分散的林地小图斑能够识别,但边缘粗糙;有林地和灌木林地在色调上区分明显;在融合后的明度图中的灌木林地有明度较大的像元,此类像元为石旮旯地。(2)通过定量分析,融合之后的影像较SAR和光学数据信息量大。同极化平均梯度大于交叉极化,HH极化方式下各指标最大。图像EM聚类分割比K-means聚类更加细化。EM聚类图像的特征区分明显;(3)研究分类出了有林地、灌木林地和其他林地。面向对象的林地分类总体分类精度达到85.71%。[结论]研究将微波与光学遥感结合,为喀斯特山区中农村地区的林地提取提供新思路,与传统的林地监测相比,数据获取快捷,提高工作效率,精度准确。有助于通过遥感的手段解决地块破碎区域的林地提取问题,为提高多源遥感技术在喀斯特农村地区中的林地智能监测的能力提供借鉴。
[Abstract]:[objective] in order to speed up the construction of "Internet" forestry in Guizhou Province and break the bottleneck of remote sensing monitoring in Guizhou karst plateau mountain area, Radarsat-2 fully polarized SAR data with spatial resolution of 8 m and multispectral image of SPOT 6 with spatial resolution 6 m are selected as data sources. To explore the extraction technology of woodland in rural areas of karst mountain area by microwave and optical remote sensing. [methods] first, the SAR image was preprocessed by ENVI SARscape and NEST software. Radarsat-2 full polarization data and SPOT 6 standard pseudocolor image were used for HSV. Fusion. Calculate the average gradient of the fused image, Information entropy, standard deviation and mean value are used to evaluate the polarization mode of optimal fusion effect. Based on K-Means clustering algorithm and maximum expectation EM clustering algorithm, the image is segmented and the appropriate algorithm is selected. Finally, based on the accuracy evaluation of pixel confusion matrix, combining with the survey data of Guizhou Forestry Bureau, field sample and aerial map, the classification of forest land is carried out based on the threshold of clustering segmentation. Establishing reference samples to evaluate the classification results. [results] 1) after fusion, the forest land edge is obvious but rough, for the small and medium-sized buildings in the forest land, the scattered forest land small map spot in the farmland can be recognized, but the edge is rough; Forest land and shrub land are clearly differentiated in tone; in the fused brightness map, there are larger brightness pixels in the shrubbery area, which is called "Cooki.") quantitative analysis is made by means of quantitative analysis. Compared with SAR and optical data, the fused image has more information. The average gradient of the same polarization is larger than that of the cross polarization / HH polarization. The EM clustering segmentation of images is more detailed than K-means clustering. The feature areas of EM clustering images are clear. The study classifies the woodland. The overall classification accuracy of shrubbery land and other woodland is 85.71.Conclusion Microwave and optical remote sensing are combined to provide a new idea for woodland extraction in rural areas of karst mountain area, and compared with traditional forest monitoring, the classification accuracy of shrub land and other woodland classification is 85.71.Conclusion the study combines microwave with optical remote sensing. The data acquisition is quick, the work efficiency is improved, the precision is accurate. It is helpful to solve the woodland extraction problem in the broken area by remote sensing. It provides reference for improving the ability of intelligent monitoring of woodland in karst rural areas by multi-source remote sensing technology.
【作者单位】：贵州师范大学喀斯特研究院;国家喀斯特石漠化防治工程技术研究中心;
【基金】：国家自然科学基金地区项目“喀斯特石漠化地区生态资产与区域贫困耦合机制研究”(41661088) 贵州省高层次创新型人才培养计划——“百”层次人才(黔科合平台人才[2016]5674) 贵州省科技计划“基于北斗卫星的山地高效农业产业园区智能管理系统开发与应用”(黔科合GY字[2015]3001) 国家遥感中心贵州分部平台建设(黔科合计Z字[2012]4003)(黔科合计Z字[2013]003)
【分类号】：S757;TP751

【相似文献】