基于SURF和顾及维数的高光谱影像端元提取新算法

发布时间：2018-08-21 13:59

【摘要】：端元提取,丰度反演是高光谱遥感技术的重要内容,其中端元提取是关键的步骤。首次将特征提取算法speed-up robust features(SURF)引入到高光谱影像端元提取中。兼顾高光谱影像丰富的光谱信息改进了SURF算法,提出了在多维尺度空间内寻找极值点作为端元的高光谱影像端元提取新算法,即多维SURF(multi-dimensional speed-up robust features,MDSURF)算法;将其应用于美国EO—1卫星获取的云南中甸普朗地区的Hyperion高光谱影像,并成功提取了影像端元。为了进一步验证结果的可靠性,设计两组对比实验,分别利用N-FINDR和连续最大角凸锥(sequential maximum angle convex cone,SMACC)算法在同等条件下提取实验影像的端元,然后对三种方法的结果进行综合评价和分析,得出MD-SURF算法提取端元的观感效果较好、精度最高、质量最好。提出了一种新的高光谱影像端元提取算法,实验结果表明新方法具有精度高、鲁棒性好等特点,证明了基于新物理机理的MD-SURF算法是一种可行的高光谱端元提取算法。
[Abstract]:End-component extraction and abundance inversion are important contents of hyperspectral remote sensing technology. End-component extraction is a key step. For the first time, feature extraction algorithm (speed-up robust features (SURF) is introduced into the End-component extraction of hyperspectral images. Taking into account the rich spectral information of hyperspectral images, the SURF algorithm is improved, and a new algorithm, multi-dimensional speed-up robust features MDSURF (multi-dimensional speed-up robust features MDSURF) algorithm, is proposed to extract hyperspectral images with extreme points as endpoints in multidimensional scale space. It has been applied to the Hyperion hyperspectral images obtained from the EO-1 satellite of Yunnan Zhongdian Plang area and the image endpoints have been extracted successfully. In order to further verify the reliability of the experimental results, two groups of comparative experiments were designed to extract the endpoints of experimental images under the same conditions using N-FINDR and (sequential maximum angle convex conical SMACC algorithm respectively. Then the results of the three methods are evaluated and analyzed synthetically. It is concluded that the MD-SURF algorithm has the best quality and the highest precision. A new End-component extraction algorithm for hyperspectral images is proposed. The experimental results show that the new method has the advantages of high precision and good robustness. It is proved that the MD-SURF algorithm based on the new physical mechanism is a feasible algorithm for extracting hyperspectral endelements.
【作者单位】：中国矿业大学(北京)地球科学与测绘工程学院;
【基金】：国家自然科学基金项目(41271436) 中央高校基本科研业务费专项资金(2009QD02)资助
【分类号】：TP751

【相似文献】