非破壞果實品質測定儀測定不同品種酪梨之乾物質含量初報

Nondestructive Evaluation of Avocado Fruit Dry Matter Content Using Near Infrared Spectroscopy

劉惠菱(Huei-ling Liu)；徐錦木(Ching-Mu Hsu)；李婕瑜(Chieh-Yu Li)

酪梨 ； 非破壞果實品質測定儀 ； 乾物質含量 ； 近紅外線光譜 ； avocado ； nondestructive evaluation ； dry matter content ； near infrared spectroscopy

臺中區農業改良場研究彙報

161期（2023 / 12 / 01）

73 - 89

繁體中文;英文

在臺灣酪梨為新興果樹，其果實品質與成熟度息息相關，以往破壞性測量曠日廢時，近年以近紅外線光譜進行非破壞測量技術十分先進，可快速檢測可溶性固形物含量、乾物質含量等與果實品質相關之性狀。本試驗以近紅外線光譜之非破壞果實品質測定儀，測量早、中、晚熟品種酪梨之乾物質含量，並建立預測模型，期能提供酪梨採收成熟度判斷依據。預測模型準確度的判斷上，‘露拉’（‘Lula’）不論在殘值預測偏差（Residual prediction deviation, RPD）、預測殘值平方和（Predicted residual sum of squares, PRESS）、均方根誤差（Root mean square error, RMSE）及交叉驗證的均方根誤差（Root mean square error of cross validation, RMSECV）值在2021年、2022年及上述2年綜合數據結果皆較佳，因此綜合考量6個試驗品種中‘露拉’（‘Lula’）以非破壞果實品質測定儀（F-750 Produce Quality Meter）測量所得之乾物質含量預測模型較其他參試品種準確；然此預測模型以殘值預測偏差（Residual prediction deviation, RPD）值判斷預測模型之可行性卻為不可信任預測模型，加上大部分果實測量易有光線洩漏影響吸收數值，進而影響預測模性之建立，因此試驗所使用之非破壞果實品質測定儀（F-750 Produce Quality Meter）實際應用於酪梨採收成熟度判斷上仍待精準化。

Avocado was a new fruit industry in Taiwan, that fruit quality was closely related to maturity. In the past, destructive measurement was spent too much time. In recent years, nondestructive measurement technology used near-infrared spectroscopy(NIF) was very advanced, which could quickly detect traits related to fruit quality, such as soluble solids content or matter content, etc. In this experiment, the dry matter content of different avocado varieties was analyzed with a nondestructive evaluation by near infrared spectroscopy, and establishes a prediction model which could provide a basis for prediction the avocado maturity of harvest. In terms of determine the accuracy of the prediction model, 'Lula' has better Residual prediction deviation (RPD), Predicted residual sum of squares (PRESS), Root mean square error (RMSE) and Root mean square error of cross validation (RMSECV) values in 2021, 2022 and the comprehensive data results of the above two years. Therefore, 'Lula' was comprehensively considered among the 6 test varieties. The dry matter content prediction model measured by F-750 produce quality meter. is more accurate; however, the feasibility of the prediction model was determined by the RPD value as an untrustworthy prediction model. In addition, most fruit measurements are prone to light leakage that affects the determine of absorption values, thereby affecting the establishment of predictive models. This result indicates the F-750 produce quality meter used in avocado harvest maturity was still needs to be accurate.