Passenger’s Behavior Recognition System Using Computer Vision

Abstract

One of the problems that occurs inside public transport is that passengers often overlook or ignore public transport’s rules such as in regards to eat and drink. Eating foods and drinks are not even allowed in public transport in avoiding excessive littering and drink spills which could potentially cause unwanted accidents. This system aims to recognize two actions such as eating and drinking through image processing in real-time environment. It also aims to classify and label the behavior of the passengers. This research is on a passengers behavior recognition system using computer vision (PBRSUCV) as a prototyping model. The method consists of image processing with a faster region-based convolutional neural networks (faster R-CNN) classification project that can be implemented in public transport to solve the stated problem. The system consists of a camera and laptop. The camera is used as a sensor to detect the targeted behaviors while the laptop will be the main device where every image processing takes place. The system was tested in real-time and is able to detect and label the eating and drinking behavior correctly with 99% accuracy on a single and two people in the image frame. It is certain that this system is capable to accurately recognize and classify the targeted behaviors in the public transport without any problem with the help of the faster R-CNN deep neural network.