Abstract
|
|
One webcam located at a coffee shop was used to estimate the times of the day the store has the most customers. The results from the manual and the automated analysis are presented. We concluded that the preferred period of the day to get coffee products is from 10:00 am to 2:00 pm.
|
Introduction
|
|
Knowing the times a store has the most customers represents key information to the store administrators. This project introduces an algorithm to estimate this information using a publicly available webcam. Webcams can gather a large amount of data at relatively low cost.
|
Methods
|
|
One image per minute was downloaded from The Coffee Gallery Website. Only a small region of the image is processed by the algorithm. A customer is defined as a person standing in front of the counter that pays the employee for a product.
|
Results
|
|
The algorithm was implemented in Java. The observation was performed on April 6th, 2006 from 6:54 am to 2:32 pm and from 5:09 pm to 8:00 pm (Hawaiian time). The application was run over 536 images. It identifies a customer in a given image with accuracy of 97.94%; likewise, it determines whether a customer in an image is different from the one on the previous frame with accuracy of 75.13%. Some of the results are shown below.
|
Privacy
|
|
Under a principle of Fair Information Practices, the customers should be given notice that they are being recorded before any data is collected from them. If the store uses an explicit way to make the customers aware that there is a webcam and its video feed is transmitted over the Internet in real time, their privacy is not violated.
Even though the video feed reveals personal information, the proposed algorithm is not based on biometrics or uniquely identifying facts to perform its task.
|
References
|
|
Camera: https://www.roastmaster.com/cameras.htm
Project Paper (PDF)
|
