臺灣博碩士論文加值系統

隨著智慧營運的到來，愈來愈多的商家使用科技了解顧客的消費取向，以適當地推銷商品。就實體商店而言，最直接的方式是利用既有的監視設備，從中分析消費者在實體店內的消費行為，一方面可望取得有別於結帳清單的喜好資訊，另一方面則可匯整消費者於店內的瀏覽資訊，使其獲得更符合自身需求的消費體驗。
有鑑於此，我們提出兩種計算方式：(1)觸碰關聯－利用深度學習追蹤消費者手部位置，另輔以前景分析提取觸碰時的商品資訊，提升在高密度商品擺設時的分辨率，以分析消費者的觸碰商品行為；(2)瀏覽關聯－利用上述觸碰行為發生時，消費者通常會注視這商品的特性，作為初始消費者頭部角度資訊，利用疊代最近點演算法分析頭部的平移與旋轉情形，掌握消費者可能注視的商品區域。實驗結果顯示，通過模擬商場的方式，結合(1)與(2)的技術分析，可以有效即時分析消費者的購物行為並了解其喜好意向。

Nowadays, more and more merchant stores employ various kinds of technologies to probe the potential interest of customers for better promotion. For a physical store, the most direct way to do this is leveraging the existent surveillance equipment. By analyzing the footages taken from the surveillance equipment's, it is possible to obtain shopping preference that is not revealed in invoices. Also, traces can be properly summarized and presented to customers for better shopping experience.
In this regard, we propose methods to analyze the footage. We first employ neural network to track the hands of a customer and subsequently filter out the non-hand area within a tracking box. Then, the time during which the hand image overlaps with that of a merchant is calculated to reveal the potential of a customer touching the merchant. Along the way, we further utilize the position of head during touch calculation to conjecture the time when a customer only browses but not touches. Experiments with a simulated shop shelf show that the proposed methods are able to analyze and summarize the touch and browse behaviors of customers during shopping.

誌謝 i
摘要 ii
Abstract iii
目錄 iv
圖目錄 vi
表目錄 viii
第一章 緒論 1
第二章 相關研究 4
2.1 以深度學習實現目標偵測 4
2.1.1 神經網路基本單元 4
2.1.2 卷積神經網絡 6
2.1.3 深度學習模型 9
2.1.4 YOLO 11
2.1.5 YOLOv2 13
2.2 圖像特徵比對 17
第三章 反應消費意願之現場行為判別 21
3.1 目標追蹤之光流輔助改善 22
3.2 觸碰關聯(Touch Correlation) 23
3.2.1 前景偵測 25
3.2.2商品拿取判斷 26
3.3瀏覽關聯(Watch Correlation) 28
3.3.1 頭部資訊提取 30
3.3.2 頭部角度分析 31
3.3.2.1 移除非目標之位移向量 31
3.3.2.2 疊代最近點演算法(iterative closest point) 33
第四章 實驗結果 37
4.1 實驗設置 37
4.2消費者感興趣關聯度分析 38
4.3結果評估 41
第五章 結論與未來展望 46
參考文獻 48

[1] 王稚慧(2017)。美國知名零售陸續傳出關閉 －實體通路危機逐漸顯現？。檢自http://www.credit.com.tw/creditonline/Epaper/ThemeContent.aspx?sn=12&unit=425 (July.18, 2018)
[2] J. Peck and T. L. Childers, "Individual differences in haptic information processing: The “need for touch” scale," Journal of Consumer Research, vol. 30, no. 3, pp. 430-442, 2003.
[3] H. N. Schifferstein and M. P. Cleiren, "Capturing product experiences: a split-modality approach," Acta psychologica, vol. 118, no. 3, pp. 293-318, 2005.
[4] M. Balaji, S. Raghavan, and S. Jha, "Role of tactile and visual inputs in product evaluation: a multisensory perspective," Asia Pacific Journal of Marketing and Logistics, vol. 23, no. 4, pp. 513-530, 2011.
[5] J. Redmon and A. Farhadi, "YOLO9000: Better, Faster, Stronger," 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6517-6525, 21-26 July 2017 2017.
[6] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, "You only look once: Unified, real-time object detection," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779-788.
[7] H. Bay, T. Tuytelaars, and L. Van Gool, "Surf: Speeded up robust features," in European conference on computer vision, 2006, pp. 404-417: Springer.
[8] D. G. Lowe, "Distinctive image features from scale-invariant keypoints," International journal of computer vision, vol. 60, no. 2, pp. 91-110, 2004.
[9] R. Girshick, J. Donahue, T. Darrell, and J. Malik, "Rich feature hierarchies for accurate object detection and semantic segmentation," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2014, pp. 580-587.
[10] R. Girshick, "Fast r-cnn," in Proceedings of the IEEE international conference on computer vision, 2015, pp. 1440-1448.
[11] Q. D. Vu and S.-T. Chung, "Real-time robust human tracking based on lucas-kanade optical flow and deep detection for embedded surveillance," in Information and Communication Technology for Embedded Systems (IC-ICTES), 2017 8th International Conference of, 2017, pp. 1-6: IEEE.
[12] O. M. M. Mohamed and M. Jaïdane-Saïdane, "Generalized Gaussian mixture model," in Signal Processing Conference, 2009 17th European, 2009, pp. 2273-2277: IEEE.
[13] https://en.wikipedia.org/wiki/Evaluation_measures_(information_retrieval)#Mean_average_precision
[14] https://zh.wikipedia.org/wiki/ROC%E6%9B%B2%E7%BA%BF
[15] P. J. Besl and N. D. McKay, "Method for registration of 3-D shapes," in Sensor Fusion IV: Control Paradigms and Data Structures, 1992, vol. 1611, pp. 586-607: International Society for Optics and Photonics.
[16] B. D. Lucas and T. Kanade, "An iterative image registration technique with an application to stereo vision," 1981.