【摘要】：水稻是世界三大糧食作物之一,其生產(chǎn)狀況與整個世界的糧食安全、社會穩(wěn)定息息相關。為此,各國政府、農(nóng)業(yè)管理部門以及農(nóng)業(yè)從業(yè)者迫切希望對水稻農(nóng)情信息實現(xiàn)及時、有效地監(jiān)控。近年來,遙感技術以其覆蓋廣、重訪周期短等特點引起了相關從業(yè)人員和管理部門的濃厚興趣。由于水稻生長周期內往往長時間被云雨覆蓋,不能確保實時獲取清晰可用的光學遙感數(shù)據(jù),因此全天時、全天候的雷達遙感成為水稻監(jiān)測和估產(chǎn)的有效觀測手段。隨著雷達技術從單極化、多極化向全極化不斷發(fā)展,以及緊致極化雷達等新型極化雷達數(shù)據(jù)的出現(xiàn),極化雷達遙感成為人們的關注熱點。它能獲得水稻冠層在不同極化方式下的雷達響應特征,包含散射強度和相位信息,更好地反映水稻冠層含水量、形態(tài)結構和長勢等各方面信息,為水稻農(nóng)情監(jiān)測提供豐富的數(shù)據(jù)支撐。本文以多時相全/緊致極化雷達遙感影像為主要數(shù)據(jù)源,輔以近同步的多光譜光學數(shù)據(jù),以高精度水稻物候期識別和參數(shù)反演為目標,在極化信息挖掘、特征提取算法、反演模型以及結果對比分析等方面進行了深入研究,嘗試改進或解決現(xiàn)有水稻物候期識別和參數(shù)反演研究中存在的重要問題,在考慮研究科學價值的同時注重研究方法和結果在實際應用中的可推廣性,為區(qū)域/田塊尺度高精度水稻物候期識別和參數(shù)反演提供可靠的理論依據(jù)和可行的實踐方法。本文的主要研究內容包括:1)基于緊致極化SAR數(shù)據(jù),不僅實現(xiàn)插秧秈稻田和撒播粳稻田的高精度分類(精度高于85%),而且較好地識別兩類水稻田的7個物候期。2)協(xié)同利用光學植被指數(shù)和雷達特征參數(shù),發(fā)展了基于蒙特卡洛隨機抽樣和相關性抑制的特征選擇算法(MCCL),構建最優(yōu)特征子集,實現(xiàn)水稻物候期自動識別。最終水稻8個物候期區(qū)間的識別總體精度為86.59%。3)深入、定量地討論了水稻物候期識別過程中的關鍵問題。首先,討論基于多時相極化SAR和多光譜數(shù)據(jù)的水稻物候期識別最優(yōu)方案。接著,發(fā)現(xiàn)考慮水稻種類和種植方式的差異對物候監(jiān)測至關重要。在考慮插秧秈稻田和撒播粳稻田的差異后,物候識別精度提高了至少16%。此外,僅使用雷達或光學數(shù)據(jù)對水稻物候期識別的精度較低(80%),光學植被指數(shù)和極化SAR特征參數(shù)分別對分蘗后期-蠟熟早期和水稻生長后期的水稻冠層生長變化不敏感。4)本文改進的極化分解方法可以有效地降低體散射過高估計和負像元的現(xiàn)象。水稻作為時變目標,在不同物候期內呈現(xiàn)出不同的散射機制。在改進的極化分解方法中,廣義體散射模型通過考慮HH和VV極化方式下后向散射系數(shù)的比值,可以更好地刻畫水稻在不同物候期內的體散射機制,為改善水稻參數(shù)反演的精度提供模型支撐。5)考慮水稻冠層水平方向含水量的差異,將二次散射引入水云模型,提出改進的水云模型,建立改進水云模型與改進極化分解的耦合架構,實現(xiàn)水稻全生育期參數(shù)反演。反演結果優(yōu)于傳統(tǒng)的水云模型,尤其是在水稻營養(yǎng)生長階段(幼苗期-孕穗期)。本文的創(chuàng)新性貢獻如下:1)首次基于緊致極化SAR數(shù)據(jù),考慮水稻品種、種植方式差異,實現(xiàn)兩類水稻田7個物候期的識別,論證了緊致極化SAR數(shù)據(jù)在水稻監(jiān)測中的潛力,拓展了其在農(nóng)情監(jiān)測中的應用,為新一代對地觀測SAR系統(tǒng)中緊致極化SAR衛(wèi)星的先期驗證提供可靠依據(jù)。2)協(xié)同利用極化SAR和多光譜數(shù)據(jù),發(fā)展了基于蒙特卡洛隨機抽樣和相關性抑制的最優(yōu)特征選擇算法,構建水稻物候期自動識別框架,首次構建水稻8個物候期識別的最優(yōu)特征矩陣,并給出基于多時相極化SAR和多光譜數(shù)據(jù)的水稻物候期識別最優(yōu)方案。3)考慮了水稻冠層的水平異質性,首次將二次散射機制引入傳統(tǒng)水云模型,發(fā)展了一種考慮物候期信息的改進水云模型。結合改進的極化分解方法,首次建立改進的水云模型與極化分解信息耦合架構,發(fā)展了一種自適應的水稻全生育期參數(shù)反演高精度反演方法。未來研究方向主要包括:第一,獲取更多不同頻段、不同角度的數(shù)據(jù),嘗試提出一種針對水稻物候期識別和水稻參數(shù)反演的最優(yōu)成像模式。第二,本文提出的改進水云模型主要針對插秧水稻田,對撒播水稻田的改進建模需要進一步思考。
[Abstract]:Rice is one of the three major food crops in the world, and its production status is closely related to food security and social stability of the whole world. To that end, governments, agricultural management and agricultural practitioners are eager to monitor the information of rice farmers in a timely and effective manner. In recent years, the remote sensing technology has aroused the strong interest of the relevant practitioners and the management departments with the features of wide coverage and short period of repeated visits. As long as the rice growth period is covered by the cloud rain for a long time, it is not possible to ensure that the clear and available optical remote sensing data is acquired in real time, and therefore, all-weather radar remote sensing becomes an effective observation means for rice monitoring and estimation all day. With the development of the radar technology from the single polarization, the multi-polarization to the full polarization, and the emergence of new polarized radar data such as the compact-induced polarization radar, the polarization radar remote sensing has become the focus of attention. The radar response characteristics of the rice crown layer in different polarization modes can be obtained, the scattering intensity and the phase information are included, and the information of the water content, the morphological structure and the long potential of the rice crown layer is better reflected, and the rice crown layer is provided with abundant data support for the rice agricultural condition monitoring. In this paper, a multi-time-phase full/ compact polarization radar remote sensing image is used as the main data source, and the near-synchronous multi-spectral optical data is used as the target for high-precision time-period identification and parameter inversion of the rice, In order to improve or solve the important problems existing in the research of the identification of the phenological period and the inversion of the parameters of the existing rice, the paper tries to improve or solve the important problems in the research of the identification of the phenological period and the inversion of the parameters of the existing rice, and to pay more attention to the research methods and the replicability of the results in the practical application while considering the scientific value. and provides a reliable theoretical basis and a feasible practical method for the identification and the parameter inversion of the region/ field block scale high-precision rice phenological period identification and parameter inversion. The main research contents of this paper are as follows: 1) Based on the compact-induced polarization SAR data, not only the high-precision classification of the rice-planting rice field and the seed-sowing japonica rice field is realized (the precision is higher than 85%), A feature selection algorithm (MCCL) based on Monte Carlo random sampling and correlation suppression is developed to realize the automatic identification of the waiting period of rice. The key problems in the identification process of rice phenological period were discussed in depth and quantificationally. First, the optimal scheme for identifying the waiting period of rice based on multi-time-phase polarization SAR and multi-spectral data is discussed. Next, it was found that the difference in the type of rice and the planting pattern was of great importance to the monitoring of phenology. After considering the difference of rice field and sowing rice field, the accuracy of phenological recognition is increased by at least 16%. In addition, the accuracy of the identification of the rice phenological period using only radar or optical data is low (80%), The parameters of the optical vegetation index and the polarization SAR are insensitive to the change of the growth of the crown of the rice in the later stage of the stage-wax and the later stage of the growth of the rice. As a time-varying target, rice presents different scattering mechanisms for different phenological periods. in the improved polarization decomposition method, the generalized body scattering model can better characterize the body scattering mechanism of the rice in different phenological periods by taking into account the ratio of the back scattering coefficient in the HH and the VV polarization mode, in order to improve the precision of the rice parameter inversion, a model support is provided. 5) considering the difference of the water content in the horizontal direction of the rice crown layer, the secondary scattering is introduced into the water cloud model, an improved water cloud model is proposed, and a coupling structure for improving the water cloud model and the improved polarization decomposition is established, and the full growth period parameter inversion of the rice is realized. The result of the inversion is better than that of the traditional water cloud model, especially in the vegetative growth stage of rice (seedling stage-booting stage). The innovative contribution of the paper is as follows: 1) The first time based on the compact-induced polarization SAR data, considering the difference of the rice variety and the planting mode, the identification of the seven phenological periods of the two kinds of paddy fields is realized, the potential of the compact-induced polarization SAR data in the rice monitoring is demonstrated, the application of the compact-induced polarization SAR data in the monitoring of the agriculture is expanded, in order to provide a reliable basis for the early verification of the compact-induced polarization SAR satellite in the ground-observation SAR system of the next generation, the optimal feature selection algorithm based on the Monte Carlo random sampling and the correlation suppression is developed in cooperation with the polarized SAR and the multispectral data, and the automatic identification frame for the waiting period of the rice is constructed, In this paper, the optimal characteristic matrix for the first time-period identification of rice is constructed, and the optimal scheme for the waiting period of rice based on multi-time-phase polarization SAR and multi-spectral data is given. 3) The horizontal heterogeneity of the rice crown layer is considered, and the secondary scattering mechanism is first introduced into the traditional water cloud model. An improved water cloud model considering the phenological information was developed. In combination with the improved polarization decomposition method, the improved water cloud model and the polarization decomposition information coupling architecture are established for the first time, and a high-precision inversion method for the adaptive rice full growth period parameter inversion is developed. The future research direction mainly includes: firstly, acquiring more data of different frequency bands and different angles, and attempting to propose an optimal imaging mode aiming at the identification of the waiting period of the rice and the inversion of the rice parameters. Secondly, the improved water cloud model proposed in this paper is mainly aimed at the rice field of rice transplanting, and the improved modeling of the rice field is needed to be further thought.
【學位授予單位】：中國科學院大學(中國科學院遙感與數(shù)字地球研究所)
【學位級別】：博士
【學位授予年份】：2017
【分類號】：S511;S127

【相似文獻】

相關期刊論文 前10條

1 張建華,李迎春,余行杰;作物物候規(guī)律的模擬研究[J];作物學報;2000年05期

2 溫小玲;吳希從;;淺談物候資料在園藝中的應用[J];現(xiàn)代園藝;2007年08期

3 傅玉瑚;;鴨梨花芽分化物候期觀察[J];山西果樹;1980年02期

4 蔡興元,虞錦星;李花芽分化物候期的觀察[J];山東果樹;1982年03期

5 趙錫如;;防護林對果樹物候期的影響[J];河北農(nóng)業(yè)大學學報;1983年01期

6 蔡興元;虞錦星;褚孟Z，

本文編號：2378505