鳳梨是台灣的重要作物之一。農夫將鳳梨分成兩類，一類是鼓聲鳳梨，另一類是肉聲鳳梨。由於肉聲鳳梨含有較多的水分，在銷售時，比起鼓聲鳳梨而言，比較容易腐壞，更不容易保存。因此，農夫需要將肉聲鳳梨去除。最簡易分辨鼓聲鳳梨和肉聲鳳梨的方式是藉由聲音, 農夫會用尺去拍打鳳梨，鼓聲鳳梨會發出類似於打鼓的聲音，而肉聲鳳梨則會發出類似於拍打手臂的聲音。
然而，這是一個耗時耗力的工作。因此，本篇論文提出了自動化分類鳳梨的方式。利用樹莓派、伺服馬達以及超聲波感測器組成的敲擊儀器，配合輸送帶來達成自動化鳳梨分類。在辨識鳳梨上本篇論文使用了兩種方式，第一種是藉由聲音的資料，另外一種則是使用近紅外線感測器。在蒐集聲音資料時，使用了前面所提出敲擊儀器，並且搜集了多組的資料，每組資料都包含了一些實驗設定上的差異，其中，也包含了來自農場的噪音。藉由這些不同的設定差異，我們測試了所建立模型的適應力並提出相關分析。在近紅外線上，我們偵測了鳳梨不同的部位。除了利用近紅外線資料和將鳳梨鳳梨分類成鼓聲鳳梨及肉聲鳳梨，我們也利用電容值量化鳳梨種的含水量，並且配合近紅外線資料進行迴歸分析。

The pineapple is one of the essential fruits in Taiwan. Farmers separate pineapples into two types by the difference percentage of water in the pineapple. One is the drum sound pineapple, and the other is the meat sound pineapple. Because there is more water in the meat sound pineapple, the meat sound pineapple is more easily rotted and more challenging to store than the drum sound pineapple. Thus, farmers need to filter the meat sound pineapple out so that they can sell pineapples overseas. The most straightforward way to distinguish a drum-sound pineapple from meat-sound pineapples is to compare the sounds generated upon striking the fruit with any rigid objects, in this case, plastic rules, commonly used by the farmers due to the neglectable low cost and easy to obtain. The sound is like hitting the drum when hitting the drum sound pineapple, and like hitting on the arm when hitting the meat sound pineapple.
However, it is a time-consuming job, so we proposed a method to automatically classify pineapples in this thesis. Using Raspberry Pi, servo, and ultrasonic sensor, we built a hitting machine and combined it with a conveyor to automatically separate pineapples. To classify pineapples, we proposed two methods related to a spectrogram. One uses acoustic data to generate a spectrogram and the other uses near infrared spectroscopy. In the acoustic data collection step, we used the hitting machine mentioned before and collected many groups of data with different factors, and some groups also included the noise in the farm. With these differences, we can test our model performance. For near infrared, we detected near infrared on different parts of a pineapple. Besides using near infrared data to classify pineapples directly, we also used capacitance to quantify water in pineapples and built a regression model to predict capacitance from near infrared data.

Acknowledgements
Abstract i
摘要 ii
1 Introduction 1
2 Related Work 7
2.1 Durians Classification . . . . . . . . . 7
2.2 Tomatoes Classification . . . . . . . . 9
2.3 Pear Contents . . . . . . . . . . . . . 10
2.4 Pineapple Translucency . . . . . . . . 13
2.5 Pineapple Maturity . . . . . . . . . . 15
2.6 Pineapple Resistance . . . . . . . . . 18
2.7 Comparison to Previous Works . . . . . 19
3 Design and Implementation 20
3.1 Acoustic Classification . . . . . . . . 20
3.1.1 System Architecture . . . . . . . . . 20
3.1.2 Data Collection . . . . . . . . . . . 23
3.1.3 Data Preprocessing . . . . . . . . . 24
3.1.4 Data Augmentation . . . . . . . . . . 24
3.1.4.1 Adding noise . . . . . . . . . . . 25
3.1.4.2 Frequency masking . . . . . . . . . 27
3.1.4.3 Time masking . . . . . . . . . . . 29
3.1.4.4 Shifting . . . . . . . . . . . . . 31
3.1.5 Short-Time Fourier Transform . . . . 33
3.1.6 Model Architecture . . . . . . . . . 35
3.2 Near-Infrared Spectroscopy . . . . . . 36
3.2.1 Motivation . . . . . . . . . . . . . 36
3.2.2 Theory . . . . . . . . . . . . . . . 36
3.2.3 NIR Sensing Equipment . . . . . . . . 37
3.2.4 Data Collection . . . . . . . . . . . 40
3.2.5 Data Labeling . . . . . . . . . . . . 40
3.2.6 Model Architecture . . . . . . . . . 41
4 Experiment and Results 43
4.1 Acoustic Classification . . . . . . . . 43
4.1.1 Data Collection . . . . . . . . . . . 43
4.1.2 Results . . . . . . . . . . . . . . . 45
4.2 Near-Infrared Spectroscopy . . . . . . 58
4.2.1 Classification . . . . . . . . . . . 58
4.2.2 Quantification . . . . . . . . . . . 68
5 Conclusions and Future Work 70
References 73

[1] 陳潮宗, 蘇義傑, and 賴永祥, “夏天就吃鳳梨吧！助消化、抗發炎好處多，但這 3 種人少吃.” https://health.gvm.com.tw/article/66387. Accessed: 2021-10-13.
[2] “農委會統計資料.” https://www.coa.gov.tw/ws.php?id=2512263. Accessed: 2021-10-13.
[3] 直 接 跟 農 夫 買, “鳳 梨 『鼓 聲 果』 與 『肉 聲 果』 的 辨 別 方 式.” https://www.buydirectlyfromfarmers.tw/catalogue/pineapple-sound_1013/. Online. Last accessed: 13 October 2021
[4] “農業生產自動化發展與應用.” https://www.coa.gov.tw/ws.php?id=9390. Online. Last accessed: 13 October 2021
[5] A. Kamilaris, F. X, and Prenafeta-Boldú, “Deep learning in agriculture: A survey,” Computers and Electronics in Agriculture, vol. 147, pp. 70–90, April 2018.
[6] Y.-A. Chen, W.-H. Hsieh, Y.-S. Ko, and N.-F. Huang, “An ensemble learning model for agricultural irrigation prediction,” in 2021 International Conference on Information Networking (ICOIN), (Jeju Island, Korea (South)), pp. 311–316, 13-16 January 2021.
[7] N.-F. Huang, D.-L. Chou, and C.-A. Lee, “Real-time classification of green coffee beans by using a convolutional neural network,” in 2019 3rd International Conference on Imaging, Signal Processing and Communication (ICISPC), (Singapore), pp. 107–111, 27-29 July 2019.
[8] W. Kharamat, M. Wongsaisuwan, and N. Wattanamongkhol, “Durian ripeness classification from the knocking sounds using convolutional neural network,” in 2020 8th International Electrical Engineering Congress (iEECON), (Chiang Mai, Thailand), pp. 1–4, 4-6 March 2020.
[9] “Mel-frequency Cepstrum.” https://en.wikipedia.org/wiki/Mel-frequency_cepstrum. Accessed: 2021-10-13.
[10] A. Baltazar, J. I. Aranda, and G. Gonzalez-Aguilar, “Bayesian classification of ripening stages of tomato fruit using acoustic impact and colorimeter sensor data,” Computers and Electronics in Agriculture, vol. 60, pp. 113–121, March 2008.
[11] D. I. Barnea and H. F. Silverman, “A class of algorithms for fast digital image registration,” IEEE Transactions on Computers, vol. C-21, pp. 179–186, February 1972.
[12] “Feature Scaling.” https://en.wikipedia.org/wiki/Feature_scaling. Online. Last accessed: 13 October 2021
[13] “Naive Bayes Classifier.” https://en.wikipedia.org/wiki/Naive_Bayes_classifier. Online. Last accessed: 13 October 2021
[14] P. Mishra, E. Woltering, B. Brouwer, and E. H. van Echtelt, “Improving moisture and soluble solids content prediction in pear fruit using near-infrared spectroscopy with variable selection and model updating approach,” Postharvest Biology and Technology, vol. 171, p. 111348, January 2021.
[15] J. W. Palmer, F. R. Harker, D. S. Tustin, and J. Johnston, “Fruit dry matter concentration: a new qualitymetric for apples,” Postharvest Biology and Technology, vol. 90, 15, pp. 2586–94, December 2010.
[16] “Partial Least Squares Regression.” https://en.wikipedia.org/wiki/Partial_least_squares_regression. Online. Last accessed: 13 October 2021
[17] J.M.Roge, B.Palagos, D.Bertrand, and E.Fernandez-Ahumada, “Covsel: Variable selection for highly multivariate and multi-response calibration: Application to ir spectroscopy,” Postharvest Biology and Technology, vol. 106, pp. 216–223, April 2011.
[18] R. Haff, D. Slaughter, Y. Sarig, and A. Kader, “X-ray assessment of translucency in pineapple,” Postharvest Biology and Technology, vol. 30, no. 5, pp. 527–533, 2006.
[19] “Logistic Regression.” https://en.wikipedia.org/wiki/Logistic_regression. Online. Last accessed: 13 October 2021
[20] S. Pathaveerat, A. Terdwongworakul, and A. Phaungsombut, “Multivariate data analysis for classification of pineapple maturity,” Journal of Food Engineering, vol. 89, pp. 112–118, November 2008.
[21] 游 昇 俯, “鳳 梨 出 現 肉 聲 果 不 是 水 分 多 農 試 所 電 阻 判 讀 技 術 助 選 別.” https://www.agriharvest.tw/archives/68133?fbclid=IwAR32cJEedW77U7qFQk-B17NMdAcG5G_l5l4kGCwulazP6tRPIp_kLNVGE8k.Online. Last accessed: 13 October 2021
[22] “Resistance Product.” https://www.tari.gov.tw/news/index-1.asp?Parser=9\%2C4\%2C28\%2C\%2C\%2C\%2C4013&fbclid=IwAR0sKpMOBC46p5MyFx6-haWqm6GorpvtGyclguqoucP7OHTQTeGie6AxcYw. Online. Last accessed: 13 October 2021
[23] “Convolutional Neural Network.” https://en.wikipedia.org/wiki/Convolutional_neural_network. Online. Last accessed: 13 October 2021
[24] “Multilayer Perceptron.” https://en.wikipedia.org/wiki/Multilayer_perceptron.Online. Last accessed: 13 October 2021
[25] “Librosa Onset Detection.” https://librosa.org/doc/main/generated/librosa.onset.onset_strength.html. Online. Last accessed: 13 October 2021
[26] “Onset Detection.” https://musicinformationretrieval.com/onset_detection.html. Online. Last accessed: 13 October 2021
[27] D. S. Park, W. Chan, Y. Zhang, C.-C. Chiu, B. Zoph, E. D. Cubuk, and Q. V. Le, “Specaugment: A simple data augmentation method for automatic speech recognition,” Proc. Interspeech, pp. 2613–2617, December 2019.
[28] “Additive White Gaussian Noise.” https://en.wikipedia.org/wiki/Additive_white_Gaussian_noise.Online. Last accessed: 13 October 2021
[29] “Wav Format.” https://www.mathworks.com/help/matlab/ref/audioread.html. Online. Last accessed: 13 October 2021
[30] “Adding Noise.” https://stackoverflow.com/questions/14058340/adding-noise-to-a-signal-in-python. Online. Last accessed: 13 October 2021
[31] “Short-time Fourier transform.” https://en.wikipedia.org/wiki/Short-time_Fourier_transform. Online. Last accessed: 13 October 2021
[32] “Matlab: Short-time Fourier transform.” https://www.mathworks.com/help/signal/ref/stft.html. Online. Last accessed: 13 October 2021
[33] S. Dieleman and B. Schrauwen, “End-to-end learning for music audio,” in 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), (Florence, Italy), 4-9 May 2014.
[34] “Near Infrared.” https://en.wikipedia.org/wiki/Near-infrared_spectroscopy. Online. Last accessed: 13 October 2021
[35] “Senorics.” https://www.senorics.com/. Online. Last accessed: 13 October 2021
[36] “16 Points Represent Spectrogram.” https://www.senorics.com/faq#Senorics-technology. Online. Last accessed: 13 October 2021
[37] A.A.Bhosale and K.K.Sundaram, “Firmness prediction of the apple using capacitance measurement,” Procedia Technology, vol. 12, pp. 163–167, 2014.
[38] “Softmax.” https://en.wikipedia.org/wiki/Softmax_function. Online. Last accessed: 13 October 2021
[39] “Classification v.s. Qualification.” https://www.quora.com/What-is-the-difference-between-classification-and-qualification. Online. Last accessed: 13 October 2021
[40] “R Square.” https://en.wikipedia.org/wiki/Coefficient_of_determination. Online. Last accessed: 13 October 2021
[41] “Scikit Learn R Square.” https://scikit-learn.org/stable/modules/generated/sklearn.metrics.r2_score.html. Online. Last accessed: 13 October 2021