科技蓬勃發展下，自動化工廠的出現，圖像與感測器信號是常見的辨識來源，當前機器視覺仍以二維圖像做辨識與處理為主，但相似形貌的物件容易產生誤判，為了解決此問題，常見的方法是通過學習大量數據以改進辨識能力。隨著硬體設備的升級，使得影像資訊添加了深度資訊，透過三維影像資訊可以更好地表現出物體的真實情形。
近年來，隨著圖像辨識領域的發展，K-mer頻率編碼是其中的演算法之一，而通過多項研究顯示，K-mer頻率編碼在二維圖像有良好的辨識能力。因此本研究期望使用K-mer頻率編碼應用至多維度數據，分別為信號辨識與立體影像辨識，在一維信號辨識方面，數據集由PVDF-石墨烯所構成的曲形、螺旋形感測器所收集，通過子序列對其採樣，以此建立K-mer信號編碼，並以三種機器學習將其分類，兩種數據集的最佳辨識結果分別為98.36%、96.7%。在立體辨識方面，使用3D深度攝影機收集七種幾何物體的影像資訊，在建立編碼的過程中，涵蓋多種參數與方法，通過多種實驗以此定義出良好的K-mer立體編碼，在使用全連接層與卷積神經網路作為分類器時，最佳辨識結果分別達到94.65%、94.28%，此表現高於LeNet的76.69%、AlexNet的79.66%與PointNet的90.91%。本論文所提出的K-mer頻率編碼方法，不受限於單一數據集，並且不同分類器皆具有一定的辨識能力，相信此方法能應用在各維度資料處理當中。

With the flourishing development of technology and the emergence of automated factories, images and signals are common sources of identification. At present, machine vision is still mainly based on two-dimensional images for identification and processing, but objects with similar appearance are prone to misjudgment. In order to solve this problem that learn a large amount of data to improve the recognition ability. With the upgrade of hardware equipment, depth information can be added to the image information, allowing objects to be represented through the three-dimensional image information.
In recent years, K-mer frequencies have been adopted as is one of the algorithms for image recognition, and many studies have shown that K-mer frequencies have good recognition ability in two-dimensional images. Therefore, this study expects to apply frequency of K-mer to multi-dimensional data, namely signal recognition and stereo image recognition. In terms of one-dimensional signal recognition, the data set of curved sensors and spiral sensors are composed of PVDF-graphene. Sampled it by subsequence, and established signal encoder of K-mer, and classified it with three kinds of machine learning. The best identification results of the two datasets were 98.36% and 96.7% respectively. Whereas in aspect of stereo recognition, a 3D depth camera us employed to collect the visual data of seven geometric objects. A wide range of variables and techniques are addressed during the code creation process, and numerous experiments are used to design an effective stereo K-mer encoder which is used to process data in a variety of dimensions. When the Fully-Connected Neural Network and Convolutional Neural Network are used as classifiers, the best recognition results reach 94.65% and 94.28% respectively, which are higher than LeNet's 76.69%, AlexNet's 79.66% and PointNet's 90.91%. The frequency of K-mer method proposed in this paper is not restricted to a single data set, and different classifiers have certain identification capabilities. It is believed that this method can be applied to data processing of various dimensions.

M. A. Kaleem, M. Khan, Significance of additive manufacturingfor industry 4.0 with introduction of artificial intelligence in additive manufacturing regimes, 2020 17th International Bhurban Conference on Applied Sciences and Technology (IBCAST), 152-156, 2020.
[2] J. Xie, S. Wan, P. Jin, Fast and Effective Object Classification for Big Image Data, 2020 IEEE International Conference on Big Data (Big Data), 5852-5854, 2020.
[3] F. Wu, T. Wu, M. R. Yuce, Design and implementation of a wearable sensor network system for IoT-connected safety and health applications, 2019 IEEE 5th World Forum on Internet of Things (WF-IoT), 87-90, 2019.
[4] A. Gams, S. Reberšek, B. Nemec, J. Škrabar, J. Skvarč, A. Ude, Robotic learning for increased productivity: autonomously improving speed of robotic visual quality inspection, 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE), 1275-1281, 2019.
[5] A. Krizhevsky, I. Sutskever, G. E. Hinton, Imagenet classification with deep convolutional neural networks, 25, 2012.
[6] J. Mrazek, S. Karlin, Detecting Alien Genes in Bacterial Genomes a, 870(1), 314-329, 1999.
[7] S. Karlin, Z.-Y. Zhu, K. D. Karlin, The extended environment of mononuclear metal centers in protein structures, 94(26), 14225-14230, 1997.
[8] S. Kariin, C. Burge, Dinucleotide relative abundance extremes: a genomic signature, 11(7), 283-290, 1995.
[9] F. Zhou, V. Olman, Y. Xu, Barcodes for genomes and applications, 9(1), 1-11, 2008.
[10] Y.-W. Wu, Y.-H. Tang, S. G. Tringe, B. A. Simmons, S. W. Singer, MaxBin: an automated binning method to recover individual genomes from metagenomes using an expectation-maximization algorithm, 2(1), 1-18, 2014.
[11] Y.-W. Wu, B. A. Simmons, S. W. Singer, MaxBin 2.0: an automated binning algorithm to recover genomes from multiple metagenomic datasets, 32(4), 605-607, 2016.
[12] 陸韋豪, 林書平, 林柏廷, 吳育瑋, 人工智慧影像辨識系統之開發及應用, 第19屆非破壞檢測技術研討會, 台北, 台灣, 2018.9.27-28.
[13] 姚佑達, 基於二階段多保真最佳化之智慧影像辨識方法, 國立臺灣科技大學, 2020.
[14] 林新翔, 基於K-mer深度學習於旋轉圖像之影像辨識方法, 國立臺灣科技大學, 2021.
[15] 張皓崴, 基於K-mer圖像特徵生成影像資料擴增, 國立臺灣科技大學, 2021.
[16] K. Muralidharan, A. Ramesh, G. Rithvik, S. Prem, A. Reghunaath, M. Gopinath, 1D Convolution approach to human activity recognition using sensor data and comparison with machine learning algorithms, 2, 130-143, 2021.
[17] J. X. Goh, K. M. Lim, C. P. Lee, 1D Convolutional Neural Network with Long Short-Term Memory for Human Activity Recognition, 2021 IEEE International Conference on Artificial Intelligence in Engineering and Technology (IICAIET), 1-6, 2021.
[18] F. Li, M. Liu, Y. Zhao, L. Kong, L. Dong, X. Liu, M. Hui, Feature extraction and classification of heart sound using 1D convolutional neural networks, 2019(1), 1-11, 2019.
[19] T. Cover, P. Hart, Nearest neighbor pattern classification, 13(1), 21-27, 1967.
[20] C.-H. Yeh, S. Subburaj, W.-S. Hung, C.-Y. Chang, P.-T. Lin, Classification of Piezoelectric Signals from PVDF/Graphene Membrane Sensors Using K-mer-based Sensing Recognition (KSR) 45th National Conference on Theoretical and Applied Mechanics (CTAM 2021) New Taipei, Taiwan, 0256, Nov. 18-19, 2021.
[21] TE CONNECTIVITY (TE), "PIEZO FILM LAB PRE-AMPLIFIER," URL: https://www.te.com/usa-en/product-CAT-PFS0015.html.
[22] National Instruments, "NI 9234 Datasheet," URL: https://www.ni.com/docs/zh-CN/bundle/ni-9234-specs/page/specs.html.
[23] S. Yahia, Y. B. Salem, M. N. Abdelkrim, 3D face recognition using local binary pattern and grey level co-occurrence matrix, 2016 17th international conference on sciences and techniques of automatic control and computer engineering (STA), 328-338, 2016.
[24] G. Zhao, M. Pietikainen, Dynamic texture recognition using local binary patterns with an application to facial expressions, 29(6), 915-928, 2007.
[25] A. Ortiz, A. A. Palacio, J. M. Górriz, J. Ramírez, D. Salas-González, Segmentation of brain MRI using SOM-FCM-based method and 3D statistical descriptors, 2013, 2013.
[26] B. Shi, S. Bai, Z. Zhou, X. Bai, Deeppano: Deep panoramic representation for 3-d shape recognition, 22(12), 2339-2343, 2015.
[27] Q. Zheng, J. Sun, L. Zhang, W. Chen, H. Fan, An improved 3D shape recognition method based on panoramic view, 2018, 2018.
[28] I. Birri, B. S. B. Dewantara, D. Pramadihanto, 3D Object Detection and Recognition based on RGBD Images for Healthcare Robot, 2021 International Electronics Symposium (IES), 173-178, 2021.
[29] J. Han, C. Moraga, The influence of the sigmoid function parameters on the speed of backpropagation learning, International workshop on artificial neural networks, 195-201, 1995.
[30] B. Karlik, A. V. Olgac, Performance analysis of various activation functions in generalized MLP architectures of neural networks, 1(4), 111-122, 2011.
[31] R. M. Neal, Connectionist learning of belief networks, 56(1), 71-113, 1992.
[32] A. F. Agarap, Deep learning using rectified linear units (relu), 2018.
[33] D. H. Hubel, T. N. Wiesel, Receptive fields of single neurones in the cat's striate cortex, 148(3), 574, 1959.
[34] K. Fukushima, S. Miyake, Neocognitron: A self-organizing neural network model for a mechanism of visual pattern recognition, Competition and cooperation in neural nets, 267-285, Springer, 1982.
[35] Y. LeCun, L. Bottou, Y. Bengio, P. Haffner, Gradient-based learning applied to document recognition, 86(11), 2278-2324, 1998.
[36] K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, 2014.
[37] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, A. Rabinovich, Going deeper with convolutions, Proceedings of the IEEE conference on computer vision and pattern recognition, 1-9, 2015.
[38] K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, Proceedings of the IEEE conference on computer vision and pattern recognition, 770-778, 2016.
[39] G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, R. R. Salakhutdinov, Improving neural networks by preventing co-adaptation of feature detectors, 2012.
[40] Intel REALSENSE, "Depth Camera D435, URL: https://www.intelrealsense.com/zh-hans/depth-camera-d435/.
[41] S. Suzuki, Topological structural analysis of digitized binary images by border following, 30(1), 32-46, 1985.
[42] Point Cloud Library, URL: https://pointclouds.org/.
[43] M. A. Fischler, R. C. Bolles, Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography, 24(6), 381-395, 1981.
[44] C. R. Qi, H. Su, K. Mo, L. J. Guibas, Pointnet: Deep learning on point sets for 3d classification and segmentation, Proceedings of the IEEE conference on computer vision and pattern recognition, 652-660, 2017.