超市情境中以智慧型購物車為基礎的排隊行為分析

Queuing Behavior Analysis using Smart Shopping Carts in Supermarkets

穆哈曼(Muhammad, Alfiansyah)

指導教授 ： 曾煜棋

國立交通大學/電機學院/電機資訊國際學位學程/碩士(2016年)

摘要

排隊行為是人類日常生活中一種群體的行為活動。這種行為經常出現在超市中。檢測和顯示每個收銀台前的隊列長度和預期等待時間，可以讓客戶和企業主得到很多便利的好處。目前現有的排隊檢測方法都不是針對超市中的排隊行為。在超市的排隊行為中，每個收銀台的預期等待時間高度依賴與當前隊列中的屬性，如排隊人數和商品購買數量。使用基於智能手機的偵測方法是不適合的，因為用戶的移動性高，需要將手機擺放在固定位置，以及需要強制讓用戶安裝專屬的APP。另外，在超市的隊列中，客戶之間的距離很近，每個人移動的行為模式也很相似，因此很難將他們進行分類。為了解決所有的上述問題，在這篇論文中，我們提出了使用智能購物車（縮寫SSC）進行排隊檢測的新型框架。該框架包括四個主要貢獻：（1）基於SSC的動作識別方案，（2）基於SSC的排隊識別方案，（3）基於SSC的隊列分區方案，以及（4）基於SSC的隊列屬性估計方案。實驗表明，我們的系統與目前最先進的方法相比：在隊列長度偵測方面，僅有80％的CDF誤差（相當於小於1個智能購物車的長度）；在等待時間預測方面，僅有80％的CDF誤差（低於111秒）。我們進行真實的超市排隊情景實驗，並且開發了智能購物車的雛形。該框架不僅可以幫助客戶減少排隊時的等待時間，提高客戶滿意度，也有利於企業主制定有效的經營策略以增加收入。

關鍵字

自動感測 ； 排隊行為分析 ； 隊列屬性估計 ； 智能購物車

並列摘要

Queuing behavior is very common in our daily life. Supermarkets are one of the places where queue often formed. Detecting and presenting each cashier line queue length and expected waiting times may benefit both customers and business owners. Existing detection methods are not suitable for supermarkets. A queuer’s waiting time is highly dependent of queuer’s properties, such as the number of people and the number of items purchased. Smartphone based approaches are not practical due to mandatory crowd app usage, high mobility of users, and fixed position limitation. Moreover, the location and uniformity of each cahier line towards each other’s poses unique queue partition problem. To address the aforementioned issues, in this thesis, we propose a novel framework for queuing detection using smart shopping carts (abbr. SSC). The proposed framework consists of four major contributions: (1) SSC-based movement recognition scheme, (2) SSC-based queuing recognition scheme, (3) SSC-based queue partition scheme, and (4) SSC-based queue properties estimation scheme. Experiments shows 80% CDF error in queue length are less equal to 1 smart shopping carts and 80% CDF waiting time estimation error are below 111 Seconds. We develop SSC’s prototypes and conducted extensive experiments in a real supermarket queuing scenario. The experiment results demonstrate that the proposed framework not only reduces customers’ queueing waiting time, improves customer satisfaction, but also benefits business owners to develop effective business strategies to increase revenues.

並列關鍵字

automatic sensing ； queuing behavior analysis ； queue properties estimation ； smart shopping carts

參考文獻

[2] W. Fan and A. Bifet, “Mining Big Data: Current Status, and Forecast to the Future,” SIGKDD Explorations, 14(2), pp. 1-5, 2012.

[3] A. B. Chan, Z. S. J. Liang, N. Vasconcelos “Privacy Preserving Crowd Monitoring: Counting People without People Models or Tracking,” IEEE Conf. on Computer Vision and Pattern Recognition, Anchorage, 2008.

[5] M. F. Bulut, Y. S. Yilmaz and M. Demirbas “LineKing: Crowdsourced Line Wait-Time Estimation using Smartphones,” University at Buffalo, SUNY, Buffalo, NY 14260, USA.

[8] D. Bauer, M. Ray, and S. Seer. Simple sensors used for measuring service times and counting pedestrians. Transportation Research Record: Journal of the Transportation Research Board, 2214(1):77–84, 2011.

[9] D. M. Bullock and et.al. Automated measurement of wait times at airport security. Transportation Research Record, 2010.