深度學習之預防車門碰撞控制系統

Deep Learning System for Preventing Collisions with Car Doors

隨著現代生活步調的加快，人們在日常活動中的節奏也日益緊湊，導致道路交通事故頻繁，根據警政署近十年統計資料顯示，由於駕駛人或乘客在開啟車門時未注意後方車流所引發的事故逐年攀升，且其造成的傷亡情形往往相當嚴重。為降低此類事故的發生，本研究以深度學習之YOLOv8（You Only Look Once version 8）物件偵測技術為核心，建置一套可辨識車輛與行人之智慧防碰撞系統，並結合門鎖控制模組實現主動安全防護機制。
研究過程中，蒐集多樣化之道路與車流影像資料，經由資料標註與模型訓練後，系統可即時判斷車輛周遭動態情形。整體系統以Raspberry Pi 4為主要運算平台，搭配鏡頭執行影像擷取與分析，於偵測到後方有汽、機車或行人靠近時，能自動鎖止車門並進行警示，以確保乘客開門時的安全。
實驗結果顯示，基於YOLOv8的辨識模型具備高準確率與良好反應速度，能有效降低開門碰撞風險，提升行車與用路安全，所以本研究成果不僅證實深度學習在車輛主動安全領域的應用潛力，也提供未來車用智慧防護系統發展之參考方向。

Opening car doors without checking rear traffic can lead to collisions that result in serious injuries or fatalities, which have increased in frequency. The present study developed an intelligent system designed to prevent such collisions. The system employs the You Only Look Once version 8（YOLOv8）deep-learning-based object detection algorithm and operates on a Raspberry Pi 4 platform. The system performs real-time image acquisition and analysis and includes a door-lock control module. When vehicles, motorcycles, or pedestrians are detected as approaching from behind, the system automatically locks the car door and triggers a visual warning to ensure the safety of passengers when exiting the vehicle. The proposed system was tested on diverse traffic and road images, which were annotated for model training. The experimental results indicated that the YOLOv8-based detection model achieved high accuracy and rapid response in detecting potentially dangerous scenarios, thus effectively reducing the risk of door-opening collisions and enhancing overall driving and pedestrian safety. The findings of this study verify the applicability of deep learning in vehicle safety systems and serve as a valuable reference for the future development of intelligent automotive protection technologies.