機車車牌辨識與移動安全之研究

本論文為了改善現有機車族群停車場進場方式而提出，利用車牌辨識系統取得機車車牌號碼，並利用YoloV3的高FPS優點實現快速定位車牌位置。伺服器會將接收到的資料進行分析，並將結果結合停車場閘門控制車輛進出，並依據不同車型種類指示車主停車方向，停車場管理員可以查看目前停車場各車種車輛停放情況。在機車鑰匙方面，使用手機透過藍芽與車輛進行配對，與目前市售車之i-key或APP有所不同，不須拿出手機打開APP，只要攜帶手機在身上靠近車輛後車輛會自動啟動，車主離開後車輛將自動熄火，適合經常上下車之送貨員使用。本論文的目的是希望能減少民眾機車進入停車場所花費的時間、減少上班上課尖峰時段高車流量造成交通堵塞問題、解決票卡遺失問題、未申請車證無證停車問題及提升機車安全便利性。

關鍵字

車牌辨識； YoloV3 ； Tensorflow ； Keras

並列摘要

Motorcycle parking identification and its security are always problems in the research field. This paper proposes a license plate recognition system to obtain the license plate number of a motorcycle, and uses high FPS YoloV3 to quickly locate the license plate. The server of the proposed system will analyze the received data, and combine the results with the parking gate to control the vehicle in and out. The proposed system is able to indicate the parking direction of the vehicle according to different types of vehicles. The parking lot administrator also can check the parking conditions of the vehicles in the current parking lot. In terms of the motorcycle security issue in the study, the mobile phone is paired with the vehicle through the Bluetooth. It differs from the i-key or some APPs currently available in the market. The proposed system does not need to take out the mobile phone to open the app, as long as the mobile phone is close to the vehicle, the vehicle engine will start automatically. By contrast, it will automatically shut off if the vehicle is far from the phone. This feature is especially suitable for the delivery men. The purpose of this thesis is to reduce the time spent by motorcycles entering the parking lot, reduce the traffic jam caused by the high traffic volume during the rush hours, solve the problem of ticket card loss, the problem of unlicensed parking without applying for a license, and improve the safety and convenience of motorcycles.