| Title |
The Smart Door System Using Pedestrian Trajectory Prediction Based on Machine Learning |
| Authors |
최재훈(Jae-Hoon Choi) ; 김동진(Dong-Jin Kim) ; 김정주(Jung-Ju Kim) |
| DOI |
https://doi.org/10.5370/KIEE.2024.73.3.618 |
| Keywords |
Vision Sensor; Machine Learning; Polynominal Regression; Trajectory Prediction; Computer Vision; Automatic Doors |
| Abstract |
This paper proposes a system that utilizes vision sensor and machine learning to predict pedestrian trajectories, aiming to reduce the high malfunction rate and safety incidents associated with existing infrared sensor-based automatic doors. The system employs vision sensor and machine learning to gather pedestrian location information, which is then input into a polynominal regression-based pedestrian trajectory prediction algorithm. The predicted pedestrian trajectory data is applied to the automatic door opening or closing determination algorithm and the and the automatic door opening or closing is determined by the pedestrian trajectory pattern.
To compare the accuracy of the automatic door system and the existing automatic door system proposed in this paper, we conducted an experiment to find out accuracy by classifying three walking trajectory patterns based on malfunctions occurring in infrared sensor-based automatic doors. The experimental results show that the automatic door system proposed in this paper accurately predicts the trajectory of pedestrians and effectively controls the automatic door opening or closing. This paper explores potential applications in smart city development, energy conservation, and enhancing pedestrian convenience. Furthermore, the system combining vision sensor, machine learning, and polynominal regression for trajectory prediction and door operation decision-making can significantly contribute to the advancement of other automation technologies. |