於每日行車紀錄探尋新開設店鋪招牌

Discovering New Shop Signboards from Daily Driving Video Data

蘇銘源(Vitor Sou Chiang)

指導教授 ： 林守德

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2021年)

https://doi.org/10.6342/NTU202103692

摘要

移動定位服務，例如導航服務是現在人們最常用的服務之一。一個好的移動定位服務必定伴隨著一個擁有最新商家資訊的資料庫去提供好的服務。這些昂貴的商家資訊，移動定位服務提供業者通常透過製圖機構、黃頁和位置資訊提供業者等渠道購買。在本研究中，我們提出了一個多模組系統架構，以宅急便貨車過去和最近的行車紀錄加上對應的全球定位系統坐標紀錄作為輸入資料，透過比較最近的行車紀錄和過去的行車紀錄探尋最近的行車紀錄中出現的新商家招牌，讓移動定位服務提供業者能利用找出的新商家招牌更新商家資訊資料庫。在實驗結果中，我們提出的系統在一筆2020年11月的行車紀錄和十三筆2020年4月的行車紀錄的進行比較, 並成功找出了31個新商家招牌。除此之外，我們提出了一個商家招牌資料探勘的策略，藉由提高所探勘的商家招牌資料集品質，提升深度學習的特徵提取器的表現。實驗結果展示出在資料集品質提升後所訓練出的特徵提取器的表現有所提升。

關鍵字

全球定位系統 ； 物件偵測 ； 多目標物件追綜 ； 特徵嵌入 ； 資料探勘 ； 深度學習

並列摘要

Location-based services become commonly used in human's daily life. An up to date shop location database is important to be a good location-based service provider. However, service providers usually update shop and frequently accessed public location information by requesting data from mapping agencies or from business and industry partners such as Yellow Pages and location data providers with high cost. In this thesis, we propose a multi-component system architecture to discover new shop signboards for updating shop location database by comparing between a recent driving record and old driving records in the database, which are recorded from dashcams with GPS trackers on delivery box trucks. According to our experimental results, our proposed system successfully discovered 31 various shop signboards between a query driving record in 2020 November and 13 driving records in a 2020 April database from a box truck driving in a specific area in Taiwan. Additionally, we presented a data mining strategy to enhance data quality of a deep feature extractor. The experimental results demonstrated the mined data with enhanced quality greatly improved the deep feature extractor performance.

並列關鍵字

Global Positioning System (GPS) ； Object Detection ； Multiple Object Tracking (MOT) ； Feature Extraction ； Data Mining ； Deep Learning

參考文獻

A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, “Simple online and realtime tracking," in 2016 IEEE international conference on image processing (ICIP), pp. 3464-3468. IEEE, 2016.

A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, “Yolov4: Optimal speed and accuracy of object detection," arXiv preprint arXiv:2004.10934, 2020.

J. Bromley, J. W. Bentz, L. Bottou, I. Guyon, Y. LeCun, C. Moore, E. Sackinger, and R. Shah, “Signature verification using a “siamese" time delay neural network," International Journal of Pattern Recognition and Artificial Intelligence, vol. 7, no. 04, pp. 669-688, 1993.

G. Cai, L. Chen, and J. Li, “Billboard advertising detection in sport tv," in Seventh International Symposium on Signal Processing and Its Applications, 2003. Proceedings., vol. 1, pp. 537-540. IEEE, 2003.

A. Casado-Garca and J. Heras, “Ensemble methods for object detection," inECAI 2020. IOS Press, 2020, pp. 2688-2695.