開發可見/近紅外光譜測定單粒稻穀水分含量系統

Development of a System for Determination of Moisture Content of Single-Kernel Rough Rice with Visible/NIR Spectroscopy

何德查(Adcha Heman)

指導教授 ： 謝清祿 ； 邱亞伯

國立屏東科技大學/國際學院/熱帶農業暨國際合作系/博士(2020年)

https://doi.org/10.6346/NPUST202000008

摘要

稻穀的水分含量(含水率)，直接影響儲存與加工的稻米品質；一般稻穀含水率約在濕基12¬~13%，以確保較為安全的儲藏。然而許多測量稻穀含水率的方式，皆相當費時，且無法提供單粒稻穀的資訊；所以，本研究的研究目的為1)探討並比較前人在稻穀含水率的直接測量與間接測量方法；2)利用光譜儀反射光譜與多重線性分析(MLR、Multiple linear regression)及部分最小平方法(PLSR、Partial least square regression)模式，預測稻穀的含水率；3)比較三種稻穀樣本型態:單粒(SK、Single kernel)、多粒(MK、Multiple kernel)、碾碎多粒(CMK、Craked multiple kernel)，及兩種光譜儀，測定稻穀含水率；4)設計並開發一台可以利用光譜儀自動測量單粒稻穀含水率的機器(自動化系統)；5)利用反應曲面法(RSM、Response surface method)，探討該自動化系統的最佳作業條件。文獻探討中顯示，非破壞稻穀水分檢測方法通常比較簡單且快速，而且可以一次分析，獲得多項成分數值；這些間接式非破壞的方法，包括近紅外線光譜、電容(Capatance)、無線電波(RF， Radio frequency)、及電感(Condutance)等。在試驗稻穀樣本，台南11號稉稻，被可見光(400-700 nm)與近紅外光(700-1050 nm)光譜儀，掃描反射光譜，並利用MLR與PLSR模式，預測單粒稻穀的含水率，七個試驗模式中，PLSR模式一次微分21點距，可以具有訓練組相關係數(rc)=0.98、誤差平方和(SEC)=1.1%，測試組相關係數(rp)=0.96，誤差平方和(SEP)=1.9%，400-1050 nm波段光譜，可獲得預測含水率準確度平均98.4%。本研究所開發的自動化測量系統，包括震動供料盤、步進馬達、光譜儀及電腦，以Box-Behnken 3x3試驗設計，探討RSM的最佳化，結果顯示供料盤頻率450Hz，每16秒一次迴轉供料，及光譜儀掃描一次，能獲得最低的含水率預測誤差0.0001，此時每分鐘可處理60粒稻穀

關鍵字

自動化 ； 含水率 ； 近紅外線光譜 ； 稻穀 ； 單粒

並列摘要

The moisture content (MC) of rice directly affects the quality of stored and processed rice; generally, the moisture content of rice is about 12 to 13% on wet basis to ensure safer storage. However, many methods for measuring the moisture content of rice are quite time-consuming and cannot provide information on single rough rice. In this study aims 1) to review the literature for determining the moisture content of rough rice: 2) to calibrate spectra with multiple linear regression (MLR) and partial least square regression (PLSR) method along with various spectral pre-treatments so as to test the performance of different treatments: 3) to determine the moisture content of rough rice by record visible and NIR spectra of three sample types for rough rice: single kernel (SK), multi kernels (MK), and cracked multi kernels (CMK) with two different spectrometers: 4) to design and develop an auto machine for measuring moisture content of single rough rice by spectral meter: 5) to optimize an auto machine performance of the system by response surface method (RSM). Literature research shows that non-destructive rice moisture detection methods are usually simple and fast, and can analyze multiple components at one time. These indirect non-destructive methods include near-infrared spectrum, capacitance, and radio wave (RF, Radio frequency), and inductance (Condutance). In the test rice sample, Tainan No.11 japonica rice was scanned by visible light (400-700 nm) and near-infrared light (700-1050 nm) spectrometers, and the reflection spectrum was scanned. The MLR and PLSR modes were used to predict the moisture content of single rough rice. In each of the test modes, the PLSR mode is differentiated by 21 points at a time, and it can have the training group correlation coefficient (rc) = 0.98, the sum of square error (SEC) = 1.1%, the test group correlation coefficient (rp) = 0.96, and the sum of square error (SEP ) = 1.9%, 400-1050 nm band spectrum, the average accuracy of predicted moisture content can be obtained 98.4%. The automated measurement system developed by this institute includes a vibrating feed tray, stepper motor, spectrometer, and computer. The Box-Behnken 3x3 test design is used to explore the optimization of RSM. The results show that the feed tray frequency is 450 Hz, once every 16 seconds. Rotary feeding and scanning by the spectrometer can get the lowest prediction error of 0.0001. At this time, 60 grains of rice can be processed per minute.

並列關鍵字

automation ； moisture content ； near-infrared spectroscopy ； rice ； single grain

參考文獻

Akowuah, J. O., Addo, A., and Bart-Plange, A. 2012. Influence of drying temperature and storage duration on fissuring and milling quality of Jasmine 85 rice variety. Journal of Science and Technology, Vol. 32, No. 2 (2012), pp 26-33.

ASAE, 1982. "Moisture measurement grain and seeds", ASAE Standard (29th Edn.), S3521St. Joseph, MI, USA.

Association of Official Analytical Chemists, 1980. "Official Methods of Analysis", 13th Edition, Washington.

American Society of Agricultural and Biological Engineers, 1988. Moisture Measurement- Unground Grain and Seeds ASABE S352.2APR.

ASABE, American Society of Agricultural and Biological Engineers, 2008. Standard: Moisture Measurement-Unground Grain and Seeds ASABE S352.2APR.