People Counter with Object Detection using Mobilenet-SSD Method
Eni Dwi Wardihani, Wimba Zainrona, Sarono Widodo, Amin Suharjono
Telecommunication Engineering, Semarang State Polytechnic, Jl. Prof. Soedarto, Tembalang, Kec. Tembalang, Kota Semarang, Jawa Tengah Indonesia 50275
*edwardihani[at]polines.ac.id
Abstract
Monitoring of a room is required to observe attendance of the visitor as well as to know the visitor-s behaviour. A surveillance camera is one of the solutions for the development of computer vision technology and IoT. This study aims to create people counter by implementing object detection using MobileNet-SSD method in which a webcam camera is connected to a laptop. The data will be sent to IoT platform Thingspeak. The system obtain outputs of object id, percentage of confidence, virtual boundary line and tuples information. The virtual line is a boundary for knowing which objects will enter or exit the room. The analysis was carried out to determine the factors that influence the people counter using statistical test of the binary logistic regression. The results show the most optimal performance are at a distance of 300 cm, an angle of 160 degre, light intensity of 300 lux, an accuracy of 100 %, sensitivity of 90 % and precision of 100 % in normal walking characteristics.
Keywords: People Counter, Computer Vision, MobileNet-SSD, Internet of Things, Thingspeak
Topic: Electrical Engineering and Computer Science