高光谱的矿区植物异常信息提取

doi:10.3964/j.issn.1000-0593(2019)01-0241-09

Abstract
References
Related Citation (15)

Download: PDF (5142 KB)
Export: BibTeX | EndNote (RIS)

Abstract Seriphidium terrae-albae is a kind of plant widely distributed in various mining areas of Fuyun County, Xinjiang, China. The traditional exploration methods are difficult to play a role due to the existence of plant information and other obstacles, and some new methods and new ideas are urgently needed. The remote sensing plant geochemistry method is a kind of natural information source that smartly utilizes plants, transforming the plant from the barrier information to the useful information. Help people quickly and economically obtain the useful information about minerals under plant barriers. Because of its large area, being fast and non-destructive and other advantages, it has attracted more and more attention of scholars, and has become the current research hotspot. In recent years, although some scholars have synthetically considered “absorption coefficient” and “contrast coefficient” to prove that Seriphidium terrae-albae can be used as a good indicator for the exploration of concealed deposits. The plants in the upper part of the deposit can absorb the ore-forming elements in the soil better, but at the same time they form geochemical anomalies in their bodies, and the information is more visible than other plant anomalies. However, no one has studied whether the geochemical anomalies in Seriphidium terrae-albae can be found from the spectral point of view, then providing some references for the exploration of concealed deposits. Therefore, our study first tries to look for the feature bands or eigenvalues closely related to geochemical anomalies, and then construct the prediction model of hidden deposit based on plant spectrum. First, the method adopted was to measure the reflectance spectra of plants growing in the upper part of deposit and background area by ASD FieldSpec3 spectrometer respectively. Then the spectra of plants growing in these two regions were analyzed and compared from five aspects, including the original spectrum, the first derivative spectrum, the second derivative spectrometry, the first derivative fractal dimension and the second derivative fractal dimension. Finally the 10 characteristic bands that were notably different were selected including R′₈₂₄, R′₈₃₄, R′_{1 533}, R′_{1 573}, R′_{1 633}, R′_{1 643}, R″_{1 284}, R″_{1 703}, the first derivative fractal dimension and the second derivative fractal dimension. These characteristic bands can be used as botanogeochemistry marks for seeking exploration of concealed deposits. Taking these ten optimized bands as input parameters, random forest (RF) and partial least squares support vector machine (PLS-SVM) were used to construct a prediction model that seeks the position of hidden deposits based on abnormal spectrum of plant. The results showed that these two models can obtain satisfactory results, but compared with the random forest model, the partial least squares support vector machine model has a better robustness and stronger generalization ability. The results also indicated that it has great potential in looking for hidden deposit using extraordinary spectrum of plants, due to the advantages of being simple and quick. Our team has built a “very low altitude detection platform” using dynamic delta wing and HySpex imaging hyperspectral sensor, which can realize the observation of “sub-meter”. But the problems will be our next research focus as follows, how to effectively solve the problem of “spatial scale” and “spectral scale”? How to better apply the model established on the ground test ground to the very low altitude detection platform, and how to extract the plant anomaly information in a large area and quickly in the research area?

Key words：Hyperspectral; Hidden deposits; Random forest (RF); Partial least squares support vector machine (PLS-SVM); Botanogeochemistry marks

Received: 2017-12-11 Accepted: 2018-05-05

ZTFLH:

P627

Corresponding Authors: ZHOU Ke-fa E-mail: zhoukf@ms.xjb.ac.cn

Cite this article:

URL:

https://www.gpxygpfx.com/EN/10.3964/j.issn.1000-0593(2019)01-0241-09 OR https://www.gpxygpfx.com/EN/Y2019/V39/I01/241

[1] Rathod P H, Brackhage C, Meer F D V D, et al. European Journal of Remote Sensing, 2015, 48(3): 283.
[2] Sridhar B B, Vincent R K, Roberts S J, et al. International Journal of Applied Earth Observation & Geoinformation, 2011, 13(4): 676.
[3] Hede A N H, Kashiwaya K, Koike Katsuaki, et al. Remote Sensing of Environment, 2015, 171: 83.
[4] Zhang C, Ren H, Qin Q, et al. Remote Sensing Letters, 2017, 8(6): 576.
[5] Zhang B, Wu D, Zhang L, et al. Environmental Earth Sciences, 2012, 65(3): 649.
[6] Shi T, Liu H, Chen Y, et al. Journal of Hazardous Materials, 2016, 308: 243.
[7] Liu M L, Liu X N, Wu M X, et al. Computers & Geosciences, 2011, 37: 1642.
[8] Wang J J, Wang T J, Shi T Z, et al. Remote Sensing, 2015, 7: 15340.
[9] WANG Hui, ZENG Lu-sheng, SUN Yong-hong, et al(王慧, 曾路生, 孙永红, 等). Transactions of the Chinese Society of Agriculture Engineering(农业工程学报), 2017, 33(2): 171.
[10] ZHU Ye-qing, QU Yong-hua, LIU Su-hong, et al(朱叶青, 屈永华, 刘素红, 等). Journal of Remote Sensing(遥感学报), 2014, 18(2): 335.
[11] Chi G Y, Shi Y, Chen X, et al. Advanced Materials Research, 2012, 347: 2735.
[12] Asmaryan S, Warner T A, Muradyan V, et al. Remote Sensing Letters, 2013, 4(2): 200.
[13] SONG Ci-an, SONG Wei, DING Ru-fu, et al(宋慈安, 宋玮, 丁汝福, 等). Geotectonica et Metallongenia(大地构造与成矿学), 2017, 41(1): 122.
[14] YOU Jun, HONG Tao, WU Chu, et al(游军, 洪涛, 吴楚, 等). Acta Petrologica Sinica(岩石学报), 2016, 32(5): 1262.
[15] QIN Ke-zhang, TIAN Ye, YAO Zhuo-sen, et al(秦克章, 田野, 姚卓森, 等). Geology in China(中国地质), 2014, 41(3): 912.
[16] Ge W, Jiao Y Q, Sun B L, et al. South African Journal of Botany, 2012, 83(4): 98.
[17] Caldelas C, Bort J, Febrero A. Cell Biology and Toxicology, 2012, 28(1): 57.
[18] Filippidis A, Papastergios G, Kantiranis N et al. Chemie der Erde-Geochemistry, 2012, 72(1): 71.