The Prototype of In-Store Visitor and People Passing Counters using Single Shot Detector Performed by OpenCV
Andes Herviana, Dodi Wisaksono Sudiharto, Fazmah Arif Yulianto · 2020
Information related to the power hours of a mall or store is important. By typically knowing it, the manager of the store or the mall can wisely determine the staff planning decision. Without the right decision, it potentially decreases customer satisfaction. The decision can be defined by utilizing in-store visitors and people passing traffic patterns. The other problem also arises when the calculation of in-store visitors and people passing are executed manually, so it requires much effort. This study proposes a prototype design of the system which can automatically calculate visitors by utilizing Single Shot Detector (SSD) method. This method is performed by operating OpenCV library. It is used to detect a human object marked as in-store visitor or people passing. The embedded computer is conducted to process images captured by Pi Camera. Based on the study, the result accuracy is 65.08% for the system counts in-store visitors, and 66.12% for the system marks objects as people pass around in front of the store. Although the accuracy values obtained is not high, but all patterns show that the highest average values of in-store visitors and people passing occur on the days nearing weekend and also on the weekend, such as Friday, Saturday and Sunday. The peak time of in-store visitors (e.g. power hour) on Friday is between 12 PM and 1 PM. The peak time of in-store visitors on Saturday is between 3 PM and 4 PM, and on Monday, it is between 4 PM and 5 PM.