稻株含氮量遙測系統之開發

陳世銘; 吳德輝; 楊智凱; 蔡養正; 黃政偉; 繆八龍; Suming Chen; Te-Hui Wu; Chih-Kai Yang; Yang-Zenq Tsai; Cheng-Wei Huang; Pa-Lung Miao

稻株含氮量遙測系統之開發

Other Title

Development of a Remote Sensing System for Rice Nitrogen Concentration Determination

Journal

應用於水稻精準農業體系之知識與技術

Part Of

農業試驗所特刊第101號

Pages

113-128

ISBN

957-01-1173-9

Date Issued

2002-06

Date

2002-06

Author(s)

陳世銘

吳德輝

楊智凱

蔡養正

黃政偉

繆八龍

Suming Chen

Te-Hui Wu

Chih-Kai Yang

Yang-Zenq Tsai

Cheng-Wei Huang

Pa-Lung Miao

URI

https://scholars.tari.gov.tw/handle/123456789/7794

Abstract

本研究進行稻株含氮量遙測系統之開發研究，以可攜式輻射光譜儀(GER-2600)定期於五種不同追肥及雜草管理之田區進行水稻植被之反射光譜遙測，並使用多光譜影像裝置遙測不同波段的稻株影像，以建立水稻特徵光譜影像之非破壞性檢測模式。另一方面並開發稻株含氮量遙測系統，以供即時監測田間水稻生育過程氮素的狀況，以達到依稻株體內氮素而施肥的精準農業管理模式。試驗結果發現，一次微分光譜479、623、632、715與1611 nm 五個特微波長組合之Stepwise MLR檢測模式最佳（r=0.89, SEC=0.29, SEP=0.46）。另一方面，亦發現一次微分光譜中藍光段之單一特徵吸收波長479 nm 與氮素呈現較高之正相關，其Stepwise MLR之單一波長檢測模式之結果為r=0.79, SEC=0.40, SEP=0.42。在MPLSR模式方面，則以模擬SPOT 4 四個感測器波段遙測監控水稻生長期含氮濃度的二次微分模式結果最佳（r=0.91, SEC=0.26, SEP=0.39）。在PLS-BPN模式方面則以一次微分數學處理的結果最佳，r=0.96, SEC=0.18, SEP=0.52。探討多光譜影像方面，已使用RGB三個波段建立室內多光譜影像偵測植株氮素含量之檢量線（r=0.83, SEC 0.20；使用660、680、945與960 nm 四個波段瀘鏡之多光譜影像遙測稻株氮素，其MIR 分析之最佳結果為r=0.88, SEC 0.27，本研究並據以開發稻株含氮量遙測系統。 This study was aimed to develop a remote sensing system for the determination of nitrogen concentration in rice. Both portable spectroradiometer and multi-spectral imager were used to measure the reflectance characteristics of rice crops with different treatments of fertilization. The calibration models were developed and evaluated; and the remote sensing system was designed with the goal of nondestructive measurement of the nitrogen content in rice stalk. The results showed that the best model of stepwise multiple linear regression (IVILR) was the first derivative of spectra with 5 wavelengths of 479, 623, 632, 715 and 1611 nm (r=0.89, SEC=0.29, SEP=0.46). The wavelength at 479 nm in blue band was found in best MLR correlation with nitrogen concentration (r=0.79, SEC=0.40, SEP=0.42 ). The analysis by MPLSR (modified partial least square regression) with the second derivative spectra provided the best results in simulating the SPOT 4 satellite spectra data (r=0.91, SEC=0.26, SEP=0.39 ). The PLS-BPN (PLS back-propagation neural network) model with the first derivative spectra yielded the results of r=0.96, SEC=0.18, SEP=0.52. In evaluating the multi-spectral images of rice stalk, the calibration model established with RGB bands gave the results with r=0.83 and SEC=0.20, while the calibration model with multi-spectra at 660, 680, 945 and 960 nm yielded the results with r=0.88 and SEC=0.27. A remote sensing system for determining the nitrogen concentration in rice stalk was developed by using the reflectance properties of rice canopy obtained in this study.

Event(s)

水稻精準農業體系之研究計畫階段性成果研討會

Subjects

遙測

多光譜影像

水稻

氮素

精準農業

Remote sensing

Multi-spectral image

Rice

Nitrogen

Precision agriculture

Publisher

農業試驗所

Type

conference paper

File(s)

Name

publication_no101_12.pdf

Size

810.83 KB

Format

Adobe PDF

Checksum

(MD5):984ab6df6d9e1c711ddf5471677c092a