以葉片溫度檢測作物的水分逆境為新興的非接觸檢測技術,該技術具備大範圍且快速檢測作物缺水狀況。作物的水分逆境及周遭環境會影響其葉片的溫度,相同環境情況下缺水越嚴重之植株,其葉片溫度越高。本研究開發之檢測系統,是由熱影像儀、溫度和照度感測器去預測作物的水分逆境,研究中採用全卷積神經網路建立影像處理演算法,使熱像儀拍攝之熱影像能自動擷取葉片形體,並偵測其溫度。使用葉片水勢(Leaf water potential, LWP)作為作物水分逆境之標準參考,進行相關量測及試驗。為量測有效的牛番茄葉片水勢,也針對其量測方式進行確效。本系統參考作物水分逆境指數(Crop water stress index, CWSI),建立葉片水勢之預測模式,其結果r^2=0.74,SEC=0.17。建立之系統能自動分割葉片熱影像之位置,並清楚區分不同缺水程度之牛番茄植株。
Using leaf temperature to detect crop water stress is an emerging non-contact detection technology. The thermal imaging technology provides the advantages in large-area and rapid detection of crop water stress. In this study, a thermal imaging detection system was developed to predict crop water stress with thermal camera, temperature sensor and illuminance sensor. A fully convolutional neural network (CNN) was used to establish an image processing algorithm, so that the thermal image captured by the thermal camera can automatically cover the shape of the leaf and its temperature can be detected. Leaf water potential (LWP) was used as a standard reference of the crop water stress to correlate the thermal imaging measurements and analyses. Crop water stress index (CWSI) was referred to establish the predict model of leaf water potential and the prediction results of the leaf water potential by the proposed thermal detection method have shown a sufficient accuracy with r^2 = 0.74, SEC = 0.17. The results implied that the thermal imaging detection system could distinguish tomato crops with different levels of water deficit.