딥러닝과 깊이카메라를 활용한 안정성이 강화된 전동 킥보드

초록

This paper proposes an enhanced safety electric scooter(S-scooter) integrating a deep learning-based forward obstacle detection module and an incident notification module. The forward obstacle detection module uses a deep learning model to detect obstacles such as cars, motorcycles, pedestrians, bicycles, and scooters, as well as defective road surfaces (potholes and speed bumps). When a detected object is identified as an obstacle, the system measures the distance using a depth camera. The incident notification module outputs a warning sound on the measured distance and sends incident information to a guardian in case of an accident. Performance evaluation shows that the deep learning model achieved a high mAP of 0.8118, and the obstacle distance measurement experiment demonstrated an average error rate of less than 10%. The proposed system is expected to reduce accidents caused by inattentive driving and inexperienced operation while significantly enhancing user safety.

키워드