本文選題：土壤二向反射 + Hapke模型�。� 參考：《華中農(nóng)業(yè)大學(xué)》2015年碩士論文

【摘要】：自然狀態(tài)下,凹凸起伏的地表高度變化會(huì)造成高光譜圖像的像元尺度傾角不一,而二向反射研究能夠?qū)D像上的像元訂正到相同的觀測角度,從而提高定量反演的精度。同時(shí),土壤作為植被的下墊面,研究土壤二向反射特性能夠?yàn)檠芯恐脖还趯庸庾V提供背景參考;土壤反射率的方向性分布還潛在攜帶有土壤濕度、有機(jī)質(zhì)含量、礦物含量等屬性信息。因此,土壤的二向反射特性研究對土壤和植被定量遙感有著重要的理論意義和研究價(jià)值。目前,可見光-近紅外光譜分析技術(shù)已成功應(yīng)用于土壤關(guān)鍵屬性預(yù)測,局部區(qū)域的土壤關(guān)鍵屬性高光譜反演模型已經(jīng)非常成熟;但由于各區(qū)域氣候條件、成土母質(zhì)的差異,局部模型難以適用于其他區(qū)域的土壤樣本。隨著大尺度土壤標(biāo)準(zhǔn)光譜庫的出現(xiàn),充分利用和挖掘大樣本土壤光譜庫中的有效信息,建立基于土壤光譜庫的土壤關(guān)鍵屬性高光譜預(yù)測模型,為解決以上問題提供了可能。本研究采集了三種典型土壤的二向反射率數(shù)據(jù),分析了他們的二向反射特性和差異;進(jìn)一步利用Hapke反演了平均單次散射反照率、粗糙度等土壤參數(shù),分析了各參數(shù)與土粒組成的關(guān)系,在此基礎(chǔ)上模擬了二向反射率分布。另外,本文針對全球土壤標(biāo)準(zhǔn)光譜庫,采用模糊C均值聚類結(jié)合偏最小二乘回歸的方法,提取了與研究區(qū)供試樣本相似的光譜子樣本集,進(jìn)一步利用子樣本集建立了土壤關(guān)鍵屬性高光譜預(yù)測模型,并進(jìn)行了模型不確定性分析。上述研究得出結(jié)論如下:1.三種典型土壤的二向反射率隨觀測角度的變化規(guī)律一致,反射率均隨著觀測天頂角的增加而增大,在前向散射方向達(dá)到最小,后向散射方向達(dá)到最大。原因是隨觀測角度的變化,土壤顆粒之間形成的陰影所占的比例會(huì)發(fā)生變化,導(dǎo)致探測器接收的光照組分有差異。2.Hapke模型各個(gè)參數(shù)對初始值的敏感性有差異。土粒組成相似的土壤單次散射反照率曲線形狀相似。隨著土壤粗顆粒(0.9mm以上)含量的增加,粗糙度參數(shù)增大,平均單次散射反照率反而減小。另外,Hapke模型能夠很好地進(jìn)行土壤二向反射率的模擬,但三種典型土壤之間的模擬精度有差別。3.通過模糊C均值聚類與偏最小二乘回歸(PLSR)相結(jié)合的方法,可挖掘土壤標(biāo)準(zhǔn)光譜庫中與研究區(qū)供試樣本相似的有效光譜信息,建立的有機(jī)碳含量估算模型可用于研究區(qū)供試樣本有機(jī)碳含量的粗略估算。本研究中,有機(jī)碳含量高光譜估算模型的預(yù)測能力主要與土壤樣本的剖面層次有關(guān),模型對下層樣本的預(yù)測能力更好。
[Abstract]:In the natural state, the variation of the surface height caused by the ups and downs of the concave and convex surface will result in different pixel dips of hyperspectral images, while the bidirectional reflection study can correct the pixels on the image to the same observation angle, thus improving the accuracy of quantitative inversion. At the same time, as the underlying surface of vegetation, the study of soil bidirectional reflectance can provide background reference for the study of vegetation canopy spectrum, and the directional distribution of soil reflectance also potentially carries soil moisture and organic matter content. Mineral content and other attribute information. Therefore, the study of the bidirectional reflectance of soil has important theoretical significance and research value for quantitative remote sensing of soil and vegetation. At present, visible light near infrared spectroscopy (VNIR) has been successfully applied to the prediction of soil key attributes. The hyperspectral inversion model of soil key attributes in local areas is very mature, but due to the climate conditions in different regions, the soil-forming parent material is different. Local models are difficult to apply to soil samples in other regions. With the emergence of large scale soil standard spectral database, the effective information of large sample soil spectral database is fully utilized and the hyperspectral prediction model of soil key attributes based on soil spectral database is established, which provides the possibility to solve the above problems. In this study, the bidirectional reflectivity data of three typical soils were collected, and their bidirectional reflectivity characteristics and differences were analyzed, and soil parameters such as average single scattering albedo, roughness and other soil parameters were further retrieved by Hapke. Based on the analysis of the relationship between the parameters and the composition of soil particles, the bidirectional reflectivity distribution is simulated. In addition, aiming at the global soil standard spectral database, the method of fuzzy C-means clustering combined with partial least square regression is used to extract the spectral subsample set similar to the sample in the study area. Furthermore, the hyperspectral prediction model of soil key attributes was established by using the subsample set, and the uncertainty of the model was analyzed. The findings of the study are as follows: 1: 1. The bidirectional reflectivity of the three typical soils is consistent with the observed angle. The reflectivity increases with the increase of the zenith angle and reaches the minimum in the direction of forward scattering and the maximum in the direction of backscattering. The reason is that with the change of observation angle, the proportion of shadow formed between soil particles will change, which leads to the difference of light components received by detector. 2. The sensitivity of each parameter of Hapke model to initial value is different. The shape of soil single scattering albedo curve with similar soil particle composition is similar. With the increase of soil coarse particle content above 0.9mm), the roughness parameter increased and the average single scattering albedo decreased. In addition, the Hapke model can well simulate the bidirectional reflectivity of soil, but the simulation accuracy of the three typical soils is different. By combining fuzzy C-means clustering with partial least square regression (PLSRs), the effective spectral information similar to the sample in the soil standard spectrum library can be mined. The organic carbon content estimation model can be used to estimate the organic carbon content of the sample in the study area. In this study, the prediction ability of the hyperspectral estimation model of organic carbon content is mainly related to the profile level of soil samples, and the prediction ability of the model to the lower samples is better.
【學(xué)位授予單位】：華中農(nóng)業(yè)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：S127;S151.9

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 高榮強(qiáng),范世福,嚴(yán)衍祿,趙麗麗;近紅外光譜的數(shù)據(jù)預(yù)處理研究[J];光譜學(xué)與光譜分析;2004年12期

2 朱連奇;朱小立;李秀霞;;土壤有機(jī)碳研究進(jìn)展[J];河南大學(xué)學(xué)報(bào)(自然科學(xué)版);2006年03期

相關(guān)碩士學(xué)位論文 前1條

1 張美琴;基于成像光譜技術(shù)土壤反射特性及剖面有機(jī)質(zhì)分布估計(jì)[D];華中農(nóng)業(yè)大學(xué);2012年

，

本文編號(hào)：2018999