2023-08-07
【學術亮點】番茄的光譜數據早期乾旱壓力檢測：一種新型的帶特徵選擇的卷積神經網絡
字體大小
小
中
大
Intelligent Detection: Development of expert diagnosis system for crop cultivation and management【Department of Agronomy / Bo-Jein Kuo / Professor】
智慧檢測：作物栽培管理專家診斷系統開發【農藝系郭寶錚教授】
智慧檢測：作物栽培管理專家診斷系統開發【農藝系郭寶錚教授】
| 論文篇名 | 英文：Early detection of drought stress in tomato from spectroscopic data: A novel convolutional neural network with feature selection 中文：番茄的光譜數據早期乾旱壓力檢測：一種新型的帶特徵選擇的卷積神經網絡 |
| 期刊名稱 | Chemometrics and Intelligent Laboratory Systems |
| 發表年份, 卷數, 起迄頁數 | 2023, 239, 104869 |
| 作者 | [11] Chin-En Kuo, Yuan-Kai Tu, Shih-Lun Fang, Yong-Rong Huang, Han-Wei Chen, Min-Hwi Yao, & Bo-Jein Kuo (郭寶錚)* |
| DOI | 10.1016/j.chemolab.2023.104869 |
| 中文摘要 | 本研究改進了一維光譜圖能量網絡(1D-SP-Net)，以構建一個帶有嵌入式殘差全局上下文(ResGC)塊的一維卷積神經網絡，稱為1D-ResGC-Net。該網絡處理番茄葉片的可見光和近紅外(Vis/NIR)光譜數據，以識別乾旱壓力的早期跡象。在評估實驗中，所提出的1D-ResGC-Net模型優於偏最小二乘判別分析(PLSDA)和隨機森林(RF)模型。利用梯度加權類激活映射、投影中的變量重要性和變量重要性，分別識別1D-ResGC-Net、PLSDA和RF輸出的最重要的特徵波段(即與乾旱壓力最密切相關的波段)。當使用最重要的15個特徵波段時，1D-ResGC-Net模型達到90％的準確率；相比之下，PLSDA和RF模型需要超過90個最重要的特徵波段才能達到90％的準確率。當輸入特徵的數量相似時，1D-SP-Net和1D-ResGC-Net的準確性非常接近。然而，當輸入特徵的數量減少時，1D-SP-Net的準確性遠低於1D-ResGC-Net。總之，1D-ResGC-Net在更低的成本下提供更高的準確性。 |
| 英文摘要 | The yield and quality of tomato (Solanum lycopersicum L.) crops are lower when the plants are exposed to drought stress. Drought stress can be prevented through timely irrigation if it is identified early. Thus, this study modified the one-dimensional spectrogram power net (1D-SP-Net) to formulate a 1D convolutional neural network with an embedded residual global context (ResGC) block; this network, called 1D-ResGC-Net, processes visible and near-infrared (Vis/NIR) spectroscopy data of tomato leaves to identify the early signs of drought stress. In evaluation experiments, the proposed 1D-ResGC-Net model outperformed partial least squares discriminant analysis (PLSDA) and random forest (RF) models. Gradient-weighted class activation mapping, variable importance in projection, and variable importance were used to identify the most important feature bands (i.e., those that were most strongly associated with drought stress) as output by the 1D-ResGC-Net, PLSDA, and RF, respectively. The 1D-ResGC-Net model achieved 90% accuracy when the 15 most important feature bands were used; by contrast, the PLSDA and RF models required more than 90 of the most important feature bands to reach 90% accuracy. When the number of input features is similar, the accuracy of 1D-SP-Net and 1D-ResGC-Net is very close. However, when the number of input features is reduced, the accuracy of 1D-SP-Net will be much lower than 1D-ResGC-Net. In summary, 1D-ResGC-Net offers greater accuracy at a lower cost. |
| 發表成果與AI計畫研究主題相關性 | 應用於溫室作物生產之過程，期建立一套應用於農業設施中的節水化作物生產系統，減少資源的浪費 |
