以光谱信息熵改进的N-FINDR高光谱端元提取算法
作者简介:杨可明(1969-),男,博士,教授,研究方向为高光谱遥感、矿山空间信息及沉陷控制。E-mail: ykm69@163.com
收稿日期: 2014-10-31
要求修回日期: 2015-02-13
网络出版日期: 2015-08-05
基金资助
国家自然科学基金项目“矿区环境重金属污染的高光谱遥感监测与分析方法研究”(41271436)
Improved N-FINDR Algorithm on Hyperspectral Endmember Extraction Based on Spectral Shannon Entropy
Received date: 2014-10-31
Request revised date: 2015-02-13
Online published: 2015-08-05
Copyright
端元提取是高光谱混合像元分解的关键步骤,也是高光谱影像分析的重要前提。N-FINDR算法是一种经典且有效的端元提取算法,但其需遍历所有可能的像元组合,计算量巨大,时间效率不高。本文以光谱信息熵和凸面几何学理论,利用高光谱影像像元,在光谱特征空间形成的单形体顶点附近为相对纯净像元,单形体内部为混合像元的特性,提出了一种结合光谱信息熵的N-FINDR改进算法。该方法根据各波段像元灰度概率计算影像中每个像元的光谱信息熵,将大于光谱信息熵阈值的像元作为混合像元被剔除,在保留的像元组成的单形体上搜索最大体积,并提取最大体积顶点处像元作为端元。最后,使用美国EO-1卫星获取的江西省德兴某铜矿的Hyperion数据,对改进后的算法进行验证。结果表明,改进后的N-FINDR算法在确保较高端元提取精度的同时,大大提高了数据处理的时间效率。
杨可明 , 魏华锋 , 刘飞 , 史钢强 , 孙阳阳 . 以光谱信息熵改进的N-FINDR高光谱端元提取算法[J]. 地球信息科学学报, 2015 , 17(8) : 979 -985 . DOI: 10.3724/SP.J.1047.2015.00979
Endmember extraction is a key step for unmixing hyperspectral mixed pixels and an important prerequisite in the further analysis of hyperspectral imagery. The traditional N-FINDR algorithm is a classical and effective algorithm among various endmember extraction methods. However, the N-FINDR algorithm need to compute all the possible pixel combinations, thus it is time consuming. In order to improve the time efficiency of the N-FINDR algorithm, this paper proposed an improved N-FINDR algorithm on hyperspectral endmember extraction based on spectral Shannon entropy theory and the convex geometry, meanwhile we utilize the characteristic that in spectral feature space, all pixels of the hyperspectral imagery could compose a single shape body, in which the pure pixels are located at the apex and the mixed pixels in the interior or at the surface. The spectral Shannon entropies of all pixels are calculated according to the pixel gray probability, and are used to determine the purity of pixels. The pixel is removed if its spectral Shannon entropy is greater than the threshold value of the spectral Shannon entropy, otherwise it is preserved. Next, the N-FINDR was used to search the largest volume from the single shape body composed by the preserved pixels, and the pixels at the apex of the body with the largest volume would be the endmembers. Finally, we use the Hyperion data of a copper mine in Dexing city from Jiangxi province to testify the improved N-FINDR algorithm. By analyzing the experimental results, the improved algorithm ensured a high accuracy as well as improved the data processing efficiency very greatly in the course of extracting hyperspectral endmembers.
Fig. 1 Single shape body of two-dimensional space图1 二维空间的单形体 |
Fig. 2 Workflow of the improved N-FINDR algorithm图2 改进N-FINDR算法的流程图 |
Fig. 3 Hyperion hyperspectral image of a miningarea in Dexing city图3 德兴矿区Hyperion高光谱遥感影像 |
Fig. 4 Endmember spectra extracted by the improved N-FINDR algorithm图4 改进N-FINDR法提取的端元光谱 |
Fig. 5 Inversion maps of the extracted endmember abundance图5 所提取的端元丰度反演图 |
Tab. 1 Analysis of similarity between the extracted endmembers and USGS-library表1 所提取的端元与USGS波谱库相似度分析 |
端元编号 | Name | 中文名称 | SAM | SFF |
---|---|---|---|---|
1 | Staurol | 十字石 | 0.849 | 0.758 |
2 | Galena | 方铅矿 | 0.781 | 0.836 |
3 | Juniper | 刺柏 | 0.908 | 0.756 |
4 | Stilbite | 辉沸石 | 0.866 | 0.794 |
5 | Epsomite | 泻利盐 | 0.838 | 0.748 |
6 | Augite | 斜辉石 | 0.859 | 0.805 |
7 | Hematita | 赤铁矿 | 0.836 | 0.766 |
8 | Mascagnin | 硫铵石 | 0.769 | 0.728 |
9 | Cookeite | 锂绿泥石 | 0.868 | 0.795 |
Tab. 2 Comparison of calculation time for 100×100 pixel image表2 100×100像元影像的运算时间对比(s) |
实验组别 | 迭代次数10 | 迭代次数50 | |||
---|---|---|---|---|---|
N-FINDR | 改进的N-FINDR | N-FINDR | 改进的N-FINDR | ||
1 | 315.8 | 19.3 | 535.7 | 23.7 | |
2 | 338.1 | 18.1 | 520.6 | 25.4 | |
3 | 319.3 | 18.4 | 529.4 | 23.3 | |
4 | 304.6 | 19.7 | 537.9 | 23.9 | |
5 | 310.4 | 19.0 | 531.6 | 24.2 |
Tab. 3 Comparison of calculation time for 400×250 pixel image表3 400×250像元影像的运算时间对比(s) |
实验组别 | 迭代次数10 | 迭代次数50 | |||
---|---|---|---|---|---|
N-FINDR | 改进的N-FINDR | N-FINDR | 改进的N-FINDR | ||
1 | 952.8 | 43.3 | 1558.4 | 69.4 | |
2 | 979.3 | 47.9 | 1498.4 | 61.7 | |
3 | 971.8 | 50.3 | 1532.7 | 73.3 | |
4 | 980.6 | 55.7 | 1626.3 | 66.7 | |
5 | 968.7 | 49.6 | 1573.6 | 75.8 |
Fig. 6 Endmember spectra extracted by improved N-FINDR, traditional N-FINDR and SMACC algorithms图6 改进的N-FINDR、传统N-FINDR与SMACC算法提取的端元光谱 |
Tab. 4 Comparison of accuracy between traditional N-FINDR, improved N-FINDR and SMACC algorithms for endmember extraction表4 传统N-FINDR、改进的N-FINDR与SMACC算法提取端元的精度比较 |
提取算法 | 端元1 | 端元2 | 端元3 | 端元4 | 端元5 | 端元6 | 端元7 | 端元8 | 端元9 |
---|---|---|---|---|---|---|---|---|---|
N-FINDR | 十字石0.744 | 赤铁矿0.791 | 锂绿泥石0.832 | 刺柏0.793 | 斜辉石0.795 | 泻利盐0.770 | 辉沸石0.845 | 硫铵石0.837 | 未知 |
改进的N-FINDR | 十字石0.785 | 赤铁矿0.787 | 锂绿泥石0.817 | 刺柏0.802 | 斜辉石0.821 | 泻利盐0.775 | 辉沸石0.816 | 硫铵石0.74 | 方铅石0.82 |
SMACC | 十字石0.753 | 赤铁矿0.741 | 锂绿泥石0.615 | 刺柏0.674 | 斜辉石0.579 | 海绿石0.586 | 方铅石 0.683 |
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] |
|
/
〈 |
|
〉 |