本文選題：土壤有機質(zhì) + 可見近紅外光譜　； 參考：《浙江工業(yè)大學》2015年碩士論文

【摘要】：土壤有機質(zhì)是土壤重要的組成部分,對土壤肥力的發(fā)展和植物的生長起著重要作用。我國傳統(tǒng)農(nóng)業(yè)生產(chǎn)中注重精耕細作、大量施用有機肥料,導致勞動生產(chǎn)率較低、對土地環(huán)境的影響較大。因此快速獲取土壤中有機質(zhì)含量信息,對于農(nóng)業(yè)生產(chǎn)的科學管理、定量施肥的推廣和精細農(nóng)業(yè)的發(fā)展具有基礎(chǔ)性的作用和實際意義。傳統(tǒng)的土壤有機質(zhì)含量測定主要是依靠實驗室化學方法測定,盡管該方法測量精度較好,然而其具有對操作人員的要求較高、耗費時間較長以及測量成本較高的缺點。由于土壤有機質(zhì)含量是農(nóng)田養(yǎng)分分級的重要指標,因此亟需一種快速簡便、精度較高的方法和儀器對農(nóng)田土壤有機質(zhì)含量進行分級評估。本研究基于可見近紅外光譜技術(shù)針對浙江省的土壤特點對有機質(zhì)含量進行了光譜建模研究和儀器研制的探索,其中實驗土壤樣本采集自浙江省內(nèi)十個不同地區(qū)共104個;采用美國海洋公司的USB4000光纖光譜儀獲取土壤光譜,范圍為350~1050nm。本文的主要工作和研究內(nèi)容如下:(1)采用可見近紅外光譜技術(shù)建立了基于全譜的土壤有機質(zhì)光譜檢測模型,對土壤有機質(zhì)含量進行定量檢測。根據(jù)采集的土壤光譜特點,分析了異常樣本及其對建模的影響。針對土壤光譜中噪聲較大的問題,采用平滑濾波、多元散射校正、基線校正和小波閾值消噪等預(yù)處理方法對光譜進行預(yù)處理并進行對比,其中小波閾值消噪法在sym6小波函數(shù)7層分解下除噪效果最佳,PLS模型的預(yù)測結(jié)果相比原始光譜決定系數(shù)(R2P)由0.74提高到了0.76,相對分析誤差(RPD)由2.00提高到了2.07。結(jié)果表明適當?shù)念A(yù)處理方法可以提高土壤有機質(zhì)檢測模型精度。(2)研究了適用于儀器設(shè)計的光譜計算模型簡化方法。由于土壤光譜數(shù)據(jù)量大、包含了大量冗余信息,導致基于全譜方式建立的檢測模型復雜度高、計算量大,因此通過提取光譜中的特征波長簡化土壤有機質(zhì)定量分析模型。提出了采用間隔偏最小二乘法、無信息變量消除、連續(xù)投影算法、競爭自適應(yīng)重加權(quán)采樣、遺傳算法、蟻群優(yōu)化算法和粒子群優(yōu)化算法等方法提取光譜中的特征波長建模,結(jié)果顯示粒子群優(yōu)化算法提取的26個波長PLS建模預(yù)測效果最佳,其R2P為0.81、RPD為2.31,通過選擇特征波長有效的簡化了檢測模型并縮短了計算時間。(3)研制了基于光譜特征波長的便攜式土壤有機質(zhì)檢測儀器樣機。采用模塊化的硬件設(shè)備設(shè)計并制作了儀器原型,并根據(jù)設(shè)計的光譜儀硬件特點使用Java編程語言開發(fā)了基于ARM-Linux嵌入式系統(tǒng)的土壤有機質(zhì)快速檢測軟件,該軟件具有良好的交互界面并且功能實現(xiàn)較為完備。采用該儀器對20個樣本實際檢測結(jié)果R2P為0.78,RPD為1.74,根據(jù)《浙江省標準農(nóng)田地力調(diào)查與分等定級技術(shù)規(guī)范》有機質(zhì)等級劃分標準對土壤有機質(zhì)含量分級準確率達到85%。該樣機經(jīng)過進一步完善后可適用于土壤有機質(zhì)的分級和現(xiàn)場使用。
[Abstract]:Soil organic matter is an important part of soil and plays an important role in the development of soil fertility and plant growth. In the traditional agricultural production of our country, we pay more attention to intensive ploughing and a large amount of organic fertilizer, which results in low labor productivity and great influence on the land environment. Therefore, it is of fundamental and practical significance for scientific management of agricultural production, extension of quantitative fertilization and development of fine agriculture to obtain the information of soil organic matter content quickly. The traditional determination of soil organic matter content mainly depends on the laboratory chemical method. Although the precision of this method is good, it has the disadvantages of higher requirement for operators, longer time consumption and higher measuring cost. Because the content of soil organic matter is an important index of farmland nutrient classification, it is urgent to use a rapid, simple and accurate method and instrument to evaluate the soil organic matter content in farmland. Based on the visible near infrared spectroscopy (VNIR), the content of organic matter in Zhejiang Province was studied by spectral modeling and instrument development. The experimental soil samples were collected from 10 different regions of Zhejiang province. The soil spectrum was obtained by USB4000 optical fiber spectrometer of American Ocean Company, with a range of 350 ~ 1050nm. The main work and research contents of this paper are as follows: (1) A spectral detection model of soil organic matter based on full spectrum was established by using visible near infrared spectroscopy (VNIR) to quantitatively detect the content of soil organic matter. According to the characteristics of soil spectrum collected, the abnormal samples and their effects on modeling were analyzed. Aiming at the problem of high noise in soil spectrum, the spectral pretreatment methods such as smoothing filtering, multivariate scattering correction, baseline correction and wavelet threshold de-noising are used to pre-process and compare the spectra. The prediction result of wavelet threshold de-noising method under the sym6 wavelet function 7-layer decomposition is improved from 0.74 to 0.76, and the relative analysis error increases from 2.00 to 2.07 compared with the original spectral determinant coefficient (R _ 2P). The results show that proper pretreatment method can improve the precision of soil organic matter detection model. Because of the large amount of soil spectral data and a large amount of redundant information, the detection model based on full spectrum method has high complexity and large amount of computation. Therefore, the quantitative analysis model of soil organic matter is simplified by extracting characteristic wavelengths from the spectrum. In this paper, the methods of interval partial least square method, information free variable elimination, continuous projection algorithm, competitive adaptive re-weighted sampling, genetic algorithm, ant colony optimization algorithm and particle swarm optimization algorithm are proposed to extract the characteristic wavelength modeling in the spectrum. The results show that 26 wavelength PLS models extracted by particle swarm optimization algorithm have the best prediction effect. The R2P of R2P is 0.81g RPD is 2.31. By selecting characteristic wavelength, the detection model is simplified and the calculation time is shortened. A portable instrument for detecting soil organic matter based on spectral characteristic wavelength is developed. The prototype of the instrument is designed and made with modularized hardware equipment. According to the hardware characteristics of the spectrometer, the rapid detection software of soil organic matter based on ARM-Linux embedded system is developed with Java programming language. The software has a good interactive interface and complete function. The actual R2P of 20 samples detected by this instrument was 0.78 and RPD was 1.74. According to the Technical Specification for soil fertility investigation and grading in Zhejiang Province, the classification accuracy of soil organic matter content was 85% according to the standard of organic matter classification. After further improvement, the prototype can be applied to soil organic matter classification and field use.
【學位授予單位】：浙江工業(yè)大學
【學位級別】：碩士
【學位授予年份】：2015
【分類號】：S151.95

【相似文獻】

相關(guān)期刊論文 前10條

1 李煥洲;陳婧婧;鐘明全;唐彰國;;基于行為特征庫的木馬檢測模型設(shè)計[J];四川師范大學學報(自然科學版);2011年01期

2 彭凌西;曾金全;;基于免疫的網(wǎng)絡(luò)動態(tài)實時異常檢測模型[J];廣州大學學報(自然科學版);2012年06期

3 李月,楊寶俊,郭華;微弱方波信號檢測模型混沌解的判定[J];地球物理學進展;2002年01期

4 畢殿杰;;多線索多層次的視頻語義事件檢測模型[J];科技信息;2009年30期

5 范國渠;;一種關(guān)聯(lián)云計算的高校網(wǎng)絡(luò)惡意攻擊檢測模型[J];科技通報;2012年08期

6 徐慶禹;;一種非線性的無監(jiān)督對象檢測模型[J];中國西部科技;2013年09期

7 黃玉文;;一種新的緩沖區(qū)溢出檢測模型[J];科技信息;2010年25期

8 蔡靜;;考慮檢測影響的貯存-檢測模型及其參數(shù)估計[J];湖北民族學院學報(自然科學版);2012年02期

9 陳姍;;基于可擴張馬爾可夫的快速精準網(wǎng)絡(luò)攻擊檢測模型[J];科技通報;2013年02期

10 高子渝,張新家,慕德俊;基于多Agent系統(tǒng)的計算機病毒免疫檢測模型[J];西北大學學報(自然科學版);2005年01期

相關(guān)會議論文 前7條

1 劉俊榮;王文槿;劉寶旭;;一種基于網(wǎng)絡(luò)行為分析的木馬檢測模型[A];第十六屆全國核電子學與核探測技術(shù)學術(shù)年會論文集（下冊）[C];2012年

2 馬文忠;郭江艷;陳科成;楊珊;王艷麗;;基于神經(jīng)網(wǎng)絡(luò)的供熱燃燒系統(tǒng)檢測模型的研究[A];2011中國電工技術(shù)學會學術(shù)年會論文集[C];2011年

3 張廣軍;賀俊吉;;基于圓結(jié)構(gòu)光的內(nèi)表面三維視覺檢測模型[A];中國儀器儀表學會學術(shù)論文集[C];2004年

4 王建平;張自立;魏華;;戰(zhàn)術(shù)空域沖突檢測模型研究[A];Proceedings of 14th Chinese Conference on System Simulation Technology & Application(CCSSTA’2012)[C];2012年

5 武照東;劉英凱;劉春;吳秀峰;;Overlay網(wǎng)絡(luò)的鏈路故障檢測模型[A];2008通信理論與技術(shù)新發(fā)展——第十三屆全國青年通信學術(shù)會議論文集（下）[C];2008年

6 李京鵬;楊林;劉世棟;;防火墻狀態(tài)檢測模型研究[A];第十八次全國計算機安全學術(shù)交流會論文集[C];2003年

7 周雙娥;熊國平;;基于Petri網(wǎng)的故障檢測模型的設(shè)計與分析[A];第六屆中國測試學術(shù)會議論文集[C];2010年

相關(guān)博士學位論文 前4條

1 趙靜;網(wǎng)絡(luò)協(xié)議異常檢測模型的研究與應(yīng)用[D];北京交通大學;2010年

2 趙斌;基于圖模型的微博數(shù)據(jù)分析與管理[D];華東師范大學;2012年

3 牛清寧;基于信息融合的疲勞駕駛檢測方法研究[D];吉林大學;2014年

4 劉鵬飛;鋁合金點焊質(zhì)量的逆過程檢測方法研究[D];天津大學;2008年

相關(guān)碩士學位論文 前10條

1 朱遠文;前端啟發(fā)式滲透檢測模型研究[D];天津理工大學;2015年

2 劉嬌;基于高光譜技術(shù)的不同品種豬肉品質(zhì)檢測模型傳遞方法研究[D];華中農(nóng)業(yè)大學;2015年

3 軒照光;ITS系統(tǒng)防碰撞技術(shù)研究[D];電子科技大學;2015年

4 祝e，

本文編號：1982797