In-store customer traffic and path monitoring in small-scale supermarket using UWB-based localization and SSD-based detection

Nowadays, retailers are embracing the Internet of Things as the latest technology to drive superior customer experience. Leverage data sources from sensors, beacons and mobile devices to identify and analyze in-store customer shopping behavior. With this motivation, this study implemented an in-store customer traffic and path monitoring system for supermarket using image processing and object detection. the system utilized the ultra-wideband indoor positioning technique to monitor the customer shopping path and the single shot multibox detection technique to monitor the real-time customer traffic. The customer monitoring system was implemented and evaluated in an actual small-scale supermarket. Results showed that the detection model prediction score and the traffic counting both obtained an accuracy score of 99%. In addition, the localization system achieved the minimum error difference of 9.73% for x coordinate and 3.86% for y coordinate between pre-determined positions and the actual anchor position readings. Furthermore, the system successfully generated the most frequent path and the total customer traffic of the day. In the future, this work can aid retail owners make better choices, run businesses more efficiently, and deliver improved customer service.