|Title:||利用高光譜資料建立小果番茄缺水逆境早期預測模型||Other Titles:||Establishment of the Early Prediction Model of Water Deficit Stress for Cherry Tomato Using Hyperspectral Data||Authors:||杜元凱
|Keywords:||高光譜;小果番茄;缺水逆境;淨最小平方判別分析;hyperspectral data;Cherry tomato;water deficits;PLS-DA||Issue Date:||30-Jun-2022||Publisher:||農業試驗所||Journal Volume:||71||Journal Issue:||2||Start page/Pages:||87-99||Source:||台灣農業研究||Abstract:||
高光譜資料收集為非破壞性檢測方法，所得光譜資料可應用於判斷植物是否處於逆境生理狀態及有無病害發生。番茄 (Solanum lycopersicum L.) 為我國重要的鮮食蔬菜，合適的水分管理對於品質與產量具關鍵影響。為建立番茄缺水逆境生理之預測方法，便於早期偵測缺水逆境，及早進行水分管理，本研究以小果番茄「玉女」品種進行缺水逆境處理，利用蒸散速率、氣孔導度和碳同化率等界定其生理狀態，處理期間收集冠層葉片高光譜資料，並以淨最小平方-判別分析 (partial least squares-discriminant analysis; PLS-DA) 建立缺水逆境預測模型。結果顯示番茄植株蒸散速率、氣孔導度和碳同化率分別於缺水處理8、9 和10 d 後顯著低於對照組 (P < 0.05)，外觀形態則在處理11 d 後才出現肉眼可觀察之葉片乾枯、萎縮現象，顯示早期缺水逆境生理反應約落於處理10−11 d 時發生。利用348−1,052 nm 高光譜資料建立的PLS-DA 模型之準確度、敏感度和特異度分別為0.90−0.93、0.87−0.89 和 0.92−0.95，顯示PLS-DA 模型針對早期缺水逆境狀態發生具有不錯的判別能力。此外，透過迴歸係數和變數投影重要性 (variable importance in the projection) 篩選出4 個特徵波段：348−584 nm、638−817 nm、937−950 nm 和1,016−1,052 nm，並進一步使用這些特徵波段建立新的PLSDA模型，該模型之準確度、敏感度和特異度分別為0.87−0.93、0.79−0.92 和0.90−0.92，顯示若僅使用特徵波段建立PLS-DA 模型，將能降低資料蒐集和後續運算的成本。
Hyperspectral data collection is a kind of non-destructive detection method, which can be applied to determine the physiological status and disease occurrence of plants. Tomato (Solanum lycopersicum L.) is an important vegetable in Taiwan, and proper water management during growth process has crucial impacts on quality and yield. In order to establish a method for predicting the early stage of water deficit stress for tomatoes, ‘Rosada’ tomato seedlings were administrated with drought treatment and physiological parameters including the transpiration rate, stomatal conductance and assimilation rate were determined. During the period of experiments, the hyperspectral reflectance data of the leaf canopy were collected as well. The physiological status of tomato was further determined by the results of physiological parameters. Coupling with the hyperspectral data, the partial least squares-discriminant analysis (PLS-DA) was constructed to predict the early stage of stress status induced by drought treatment. The results showed that the transpiration rate, stomatal conductance, and assimilation rate were significantly lower than those of the control group 8, 9 and 10 d after drought treatment (P < 0.05), while the morphology appeared obvious phenotypic variation 11 d after drought treatment. We concluded that the early physiological response to water deficit stress occurred approximately 10−11 d after drought treatment. The prediction accuracy, sensitivity and specificity of the PLS-DA models obtained from 348−1,052 nm were at the range of 0.90−0.93, 0.87−0.89 and 0.92−0.95, respectively. The results indicated that the PLS-DA model possessed acceptable prediction performance. Furthermore, four important characteristic bands, 348−584, 638−817, 937−950 and 1,016−1,052 nm, which were selected from the regression coefficients and variable importance in the projection (VIP) scores. We established additional PLS-DA models by these characteristic bands, and the prediction accuracy, sensitivity and specificity were 0.86−0.93, 0.79−0.92 and 0.90−0.92, respectively. These results showed that PLS-DA models established by the characteristic bands had no significant improvement to the model performance. However, the models could reduce the cost of data collection and subsequent operations.
|Appears in Collections:||1.台灣農業研究(1950～迄今)|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.