利用微型機器學習與微控制器即時檢測室內地磚空心缺陷

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：94

、訪客IP：18.226.251.72

姓名

莊庭翰(Ting-Han Chuang) 查詢紙本館藏

畢業系所

土木工程學系

論文名稱

利用微型機器學習與微控制器即時檢測室內地磚空心缺陷
(Detection of Indoor Tile Hollow Defects Using Tiny Machine Learning and Microcontrollers)

相關論文

★ 應用智慧標籤及數據驅動方法於水接觸結構物之結構評估	★ 基於低功耗嵌入式系統及高精度MEMS感測器的智慧鋼索監測系統研發
★ Sensor Code-based Smart Tag Embedded in Concrete for Seepage Sensing Caused by Cracks	★ 智慧型居家機器人用於地震後自動巡查及應變處置之研究
★ 利用UAV整合LoRa與磁導喚醒技術的物聯網架構研發	★ 基於磁吸附與全向輪技術的鋼結構攀爬機器人開發與驗證
★ 基於微型機器學習的智能避障系統在外牆檢測自主移動機器人中的應用	★ 基於ROS的遠端自動多螺栓檢測機器人系統開發
★ 基於BERT語意分析模型的智慧型BIM資訊搜尋問答系統之研究	★ 基於BIM與無線喚醒物聯網裝置之智慧化結構檢測系統開發
★ 結合智慧感測標籤與支持向量機快速判定混凝土裂縫位置	★ 應用於鋼結構檢測之高機動型蚇蠖攀爬機器人設計分析及實作驗證
★ 混凝土缺陷自動修補機器人之研發	★ 研發具邊緣運算能力之無線振動量測裝置應用於橋梁鋼索特徵頻率偵測
★ 結合智慧感測標籤與機器學習方法判別混凝土內部鋼筋鏽蝕可能性之研究	★ 應用於攀爬檢測機器人之輕量級即時多目標螺栓缺陷影像檢測系統之研究

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

至系統瀏覽論文 (2027-6-30以後開放)

摘要(中)

現今社會在買房交屋前會先進行驗屋，檢測室內地磚是否敲擊有空心的聲音並且是否符合標準。傳統驗屋方式為使用打診棒敲擊地磚五個點(左上、左下、右上、右下、中點)，如果超過三點（含三點）空心就要更換重貼，現今驗屋方式更加嚴格謹慎，將地磚分成九宮格進行敲擊，一樣超過三點（含三點）空心就要換掉重貼。地磚檢測過度依賴有經驗的技術人員判斷，並且會產生檢測標準過度主觀且無法紀錄的問題。因此本論文以機器學習領域中的深度學習做為解決問題的核心，將地磚空心辨識功能嵌入在Arduino Nano 33 BLE Sense微控制器上，以便於攜帶，使民眾或技術人員能以更加簡單且更有效率的方式進行地磚檢測，達到民眾可自行檢測且可以以數值的方式記錄，讓檢測結果更加有可信度，也能及早發現地磚空心問題並得以立即進行補強或更換。本論文先是設計並利用3D列印技術製作出空心磚模型，使用打診棒敲擊地磚並以Arduino Nano 33 BLE Sense內建的麥克風採集音訊資料，將所有音訊資料分成數個類別並透過Edge Impulse進行音訊前置處理和採用頻譜圖(Spectrogram)、梅爾濾波器組能量(Mel-filter bank energy,MFE)和梅爾頻率倒普係數(Mel-Frequency Cepstral Coefficients,MFCC)三種方法提取特徵，使用卷積神經網路(Convolutional Neural Networks ,CNN)訓練模型，之後藉由驗證和測試，比較Spectrogram、MFE和MFCC三種方法之衡量指標，最後發現Spectrogram模型且辨識門檻為0.75準確率最高，驗證集準確率為97.7%，測試集準確率為92.48%，實際敲擊地磚準確率最高可達81.25%，因此將此模型部署至Arduino Nano 33 BLE Sense上並安裝在自行設計打診棒上以便直接利用。一般民眾或檢測人員可自行在家或驗屋時利用智慧打診棒檢測地磚，透過簡易的敲擊便能即時辨識出地磚是否空心和是否需要更換或補強，以降低過度依賴主觀判斷和無法記錄的問題。

摘要(英)

Before buying a house, an inspection is carried out to check whether the indoor floor tiles have a hollow sound when knocked and whether they meet the standards. The traditional home inspection method uses a medical stick to hit five points on the floor tiles (upper left, lower left, upper right, lower right, and center). Floor tile inspection relies too much on the judgment of experienced technicians, and the inspection standard is too subjective and cannot be recorded. Therefore, this paper takes deep learning in the field of machine learning as the core of solving the problem and embeds the hollow recognition function of floor tiles on the Arduino Nano 33 BLE Sense microcontroller so that it is easy to carry so that people or technicians can use it more simply and conveniently. The efficient way to test the floor tiles enables the public to test by themselves and record them numerically so that the test results are more credible. The hollow problem of floor tiles can be found early and can be reinforced or replaced immediately. This thesis first designs and uses 3D printing technology to make a hollow brick model, taps the floor brick with a medical stick and collects audio data with the built-in microphone of Arduino Nano 33 BLE Sense, divides all audio data into several categories, and conducts audio through Edge Impulse Preprocessing and extracting features using three methods: Spectrogram, Mel-filter bank energy (MFE) and MelFrequency Cepstral Coefficients (MFCC), using convolution Neural network (Convolutional Neural Networks, CNN) training model, then through verification and testing, compare the measurement indicators of the three methods of Spectrogram, MFE, and MFCC, and finally found that the Spectrogram model and the identification six thresholds of 0.75 have the highest accuracy, and the accuracy of the validation set is The accuracy of the test set is 97.7%, the accuracy of the test set is 92.48%, and the accuracy of the actual hitting the floor tile is up to 81.25%. Therefore, this model is deployed on the Arduino Nano 33 BLE Sense and installed on the self-designed diagnosis stick for direct use. The general public or inspectors can use the intelligent diagnostic stick to detect floor tiles at home or during house inspections. Through a simple tap, they can instantly identify whether the floor tiles are hollow and whether they need to be replaced or reinforced to reduce excessive reliance on subjective judgments and problems that cannot be recorded.

關鍵字(中)

★ 地磚敲擊
★ 空心檢測
★ 聲音辨識
★ Spectrogram
★ MFE
★ MFCC

關鍵字(英)

★ Floor Tile Percussion
★ Hollow Detection
★ Sound Recognition
★ Spectrogram
★ MFE
★ MFCC

論文目次

摘要 I
ABSTRACT II
致謝 IV
目錄 V
圖目錄 VIII
表目錄 XII
一、緒論 1
1-1 研究動機 2
1-2 研究目的 2
1-3 論文架構 3
二、文獻回顧 4
三、研究方法 11
3-1 聲音取樣數位化流程 13
3-2 音訊資料採樣流程 13
3-3 音訊資料前置處理 16
3-4 特徵提取方法 22
3-4-1 頻譜圖(Spectrogram) 23
3-4-2 梅爾濾波器組能量(MFE) 25
3-4-3 梅爾頻率倒譜係數(MFCC) 28
3-5 頻譜轉換參數設定 29
3-5-1 特徵提取方法參數設定 29
3-5-2 建立CNN 31
3-6 演算法之虛擬碼 35
四、實驗規劃與設計 37
4-1 打診棒手把設計製作 37
4-1-1 微控制器介紹 41
4-1-2 電池、開關和升壓模組 43
4-1-4 地磚介紹 44
4-1-5 3D列印介紹 45
4-2 實驗設計製作 46
4-2-1 空心磚模型設計 47
4-2-2 灌製水泥砂漿試體 48
4-2-3 訓練和測試的地磚樣本 53
五、實驗結果與討論 55
5-1 實驗說明 55
5-1-1 訓練成果 55
5-1-2 混淆矩陣 59
5-2實驗結果 59
5-2-1 驗證集測試 60
5-2-2 測試集測試 62
5-2-3 Floor tile Ⅰ、Floor tile Ⅱ、Floor tile Ⅲ測試 63
5-2-4 異材質地磚測試 91
5-3 量化部署 94
5-4 數值化紀錄 96
5-5 實驗討論 97
5-5-1 三種模型和辨識門檻之比較 97
5-5-2 敲擊力道對地磚聲音辨識的影響 100
5-5-3 檢測環境之影響 106
5-5-4 水泥砂漿乾縮之問題 106
5-5-5 智慧打診棒優缺點之比較 107
六、結論與未來展望 109
6-1 結論 109
6-2 未來展望 109
七、參考文獻 111

參考文獻

1.郭斯傑，江立偉和羅紫萍，「建築物外牆磁磚劣化診斷模式之研究」，物業管理學報，6(2),63-73頁，民國104年。
2.梁智信和林谷陶，「外牆磁磚系統檢測技術探討」，中興工程，(130), 63-73頁，民國105年。
3.Matysik, M., I. Plskova, and Z. Chobola, " Assessment of the impact-echo method for monitoring the long-standing frost resistance of ceramic tiles " ,Materiali in tehnologije, 49(4): p. 639-643, 2015.
4.Cantavella, V., et al. " Use of ultrasound techniques to measure green tile bulk density and optimise the pressing process " ,IX World Congress on Ceramic Tile Quality, 2006.
5.野野瀨響，「建築物外牆磁磚診斷方法之比較研究」，國立臺北科技大學，碩士論文，民國104年。
6.戴佩宜，「以打音法從事建築外牆瓷磚非破壞檢測之研究」，國立高雄大學，碩士論文，民國97年。
7.顏嘉慶，「利用打音法檢測學校建築外牆磁磚狀況之研究」，逢甲大學，碩士論文，民國102年。
8.Ittichaichareon, C., S. Suksri, and T. Yingthawornsuk. " Speech recognition using MFCC. in International conference on computer graphics, simulation and modeling " ,2012.
9.Madikeri, S.R. and H.A. Murthy."Mel filter bank energy-based slope feature and its application to speaker recognition",IEEE national Conference on Communications (NCC),2011
10.Nasersharif, B. and A. Akbari. " A framework for robust MFCC feature extraction using SNR-dependent compression of enhanced Mel filter bank energies ", Ninth International Conference on Spoken Language Processing, 2006.
11.Hermansky, H., " Perceptual linear predictive (PLP) analysis of speech " ,the Journal of the Acoustical Society of America, 87(4): p. 1738-1752, 1990.
12.Dave, N.," Feature extraction methods LPC, PLP and MFCC in speech recognition ",International journal for advance research in engineering and technology, 1(6): p. 1-4, 2013.
13.Todisco, M., H. Delgado, and N.W. Evans. " Articulation Rate Filtering of CQCC Features for Automatic Speaker Verification ",Interspeech, 2016.
14.Dörfler, M., R. Bammer, and T. Grill. " Inside the spectrogram: Convolutional Neural Networks in audio processing " ,IEEE international conference on sampling theory and applications (SampTA),2017.
15.Yenigalla, P., et al. " Speech Emotion Recognition Using Spectrogram & Phoneme Embedding ", Interspeech, 2018.
16.Panyavaraporn, J., P. Limsupreeyarat, and P. Horkaew, " DWT/MFCC Feature Extraction for Tile Tapping Sound Classification " ,International Journal of Integrated Engineering, 12(3): p. 122-130, 2020.
17.PK, F.A., " What is Artificial Intelligence? Success is no accident. It is hard work, perseverance, learning, studying, sacrifice and most of all, love of what you are doing or learning to do ”, p. 65, 1984.
18.Turing, A.M., " Computing machinery and intelligence, in Parsing the turing test ", Springer, p. 23-65, 2009.
19.Li, G., et al. " Understanding error propagation in deep learning neural network (DNN) accelerators and applications ",Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysi, 2017.
20.Sherstinsky, A., " Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network ", Physica D: Nonlinear Phenomena, 404: p. 132306, 2020.
21.Graves, A., A.-r. Mohamed, and G. Hinton. " Speech recognition with deep recurrent neural networks. " IEEE international conference on acoustics, speech and signal processing, 2013.
22.李曼華，「利用類神經網路和梅爾頻率倒頻譜係數(MFCC)做語音情緒辨識」，臺北醫學大學，碩士論文，民國110年。
23.林冠良，「一個基於MFCC的語者識別系統」，東海大學，碩士論文，民國106年。
24.李佳菁，「應用深度學習之新生兒哭聲辨識」，國立雲林科技大學，碩士論文，民國105年。
25.柯惠裕，「以梅爾倒頻譜係數為基礎之特徵擷取於鳥聲辨識之應用」，中華大學，碩士論文，民國97年。
26.蘇建升，「基於MFCC智慧辨識系統用於風扇品質檢測」，國立雲林科技大學，碩士論文，民國109年。
27.蔣亞恬，「基於梅爾倒頻譜係數之非侵入式膝韌帶音訊檢測系統」，國立臺北科技大學，碩士論文，民國103年。
28.施博翔，「基於卷積神經網路的高樓外牆磁磚檢測」，國立雲林科技大學，碩士論文，民國107年。
29.Shin, H.-C., et al., " Deep convolutional neural networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning ", IEEE transactions on medical imaging, 35(5): p. 1285-1298, 2016.
30.Kattenborn, T., et al., " Review on Convolutional Neural Networks (CNN) in vegetation remote sensing " ISPRS Journal of Photogrammetry and Remote Sensing, 173: p. 24-49, 2021.
31.Boureau, Y.-L., J. Ponce, and Y. LeCun. " A theoretical analysis of feature pooling in visual recognition ", Proceedings of the 27th international conference on machine learning (ICML-10), 2010.
32.Graham, B., " Fractional max-pooling ", arXiv preprint arXiv:1412.6071, 2014.
33.Liu, W., et al., " Large-margin softmax loss for convolutional neural networks "arXiv preprint arXiv:1612.02295, 2016.
34.Disabato, S. and M. Roveri. " Incremental on-device tiny machine learning ", Proceedings of the 2nd International Workshop on Challenges in Artificial Intelligence and Machine Learning for Internet of Things, 2020.
35.Disabato, S. and M. Roveri, " Tiny machine learning for concept drift ", arXiv preprint arXiv:2107.14759, 2021.
36.Lee, K.-C. and B.H. Cho, " Design and analysis of automotive high intensity discharge lamp ballast using micro controller unit ",IEEE Transactions on Power Electronics, 18(6): p. 1356-1364, 2003.
37.Won, T.-H. and S.-J. Park, " Design and implementation of an omni-directional underwater acoustic micro-modem based on a low-power micro-controller unit " ,Sensors,12(2): p. 2309-2323, 2012.
38.Vergin, R. and D. O′Shaughnessy. " Pre-emphasis and speech recognition ", IEEE proceedings 1995 Canadian Conference on Electrical and Computer Engineering, 1995.
39.Rapuano, S. and F.J. Harris, " An introduction to FFT and time domain windows "IEEE instrumentation & measurement magazine, 10(6): p. 32-44, 2007.
40.Wyse, L.,"Audio spectrogram representations for processing with convolutional neural networks", arXiv preprint arXiv:1706.09559, 2017.
41.Xu, H., X. Zhang, and L. Jia. " The extraction and simulation of Mel frequency cepstrum speech parameters " ,IEEE international Conference on Systems and Informatics (ICSAI2012), 2012.
42.Al-Mimi, H., et al., " A Study on New Arduino NANO Board for WSN and IoT Applications ",International Journal of Advanced Science and Technology, 29(4): p. 10223-10230, 2020.
43.Tong, F., et al., " Evaluation of tile–wall bonding integrity based on impact acoustics and support vector machine ",Sensors and Actuators A: Physical, 144(1): p. 97-104,2008.
44.Luk, B.L., et al., "Impact-acoustics inspection of tile-wall bonding integrity via wavelet transform and hidden Markov models ", Journal of sound and vibration, 329(10): p. 1954-1967, 2010.
45.Xie, L., et al., " Inspection of magnetic tile internal cracks based on impact acoustics "Nondestructive Testing and Evaluation, 30(2): p. 147-164, 2015.
46.Huang, Q., Y. Yin, and G. Yin, " Automatic classification of magnetic tiles internal defects based on acoustic resonance analysis " Mechanical Systems and Signal Processing, 60: p. 45-58, 2015
47.Cunha, R., et al. " Applying non-destructive testing and machine learning to ceramic tile quality control " IEEE VIII Brazilian Symposium on Computing Systems Engineering (SBESC), 2018.
48.Liu, L., L. Jiang, and L. Qiao. " Nondestructive Detection of Ceramic Products Based on Tapping Sound Signal Feature Recogition. in Journal of Physics: Conference Series ", IOP Publishing, 2021.
49.Wang, K., et al. " Surface Material Classification Based On Tapping Sound Characteristics ", IEEE 7th International Conference on Computer and Communications (ICCC), 2021.
50.Lu, H., et al., " Multimodal Fusion Convolutional Neural Network with Cross-attention Mechanism for Internal Defect Detection of Magnetic Tile " ,IEEE Access, 2022.
51.Fukumura, T., et al., " Improvement of Sound Classification Method on Smartphone for Hammering Test Using 5G Network ", International Journal of Networking and Computing, 12(2): p. 359-371, 2022.
52.李琳，洪青陽，語音識別原理與應用，電子工業出版社，2020
53.柳若邊, 最專業的語音辨識全書：使用深度學習實作, 2019
54.Pete Warden，Daniel Situnayake , TinyML：TensorFlow Lite機器學習, 2020

指導教授

林子軒(Tzu-Hsuan Lin)

審核日期

2022-9-1

推文