摘要
根據臺灣衛生福利部統計指出，2018年臺灣約有3千人口死於車禍事故。而世界衛生組織（WHO）指出先進國家頻繁發生車禍事故將會造成3~5%GDP的損失。因此發展智慧駕駛輔助系統或智慧交通技術是全球先進國家一致的目標，故近年來車聯網正在快速地發展。有鑑於此，本團隊開發設計並建置危險駕駛分析平台，進行駕駛者行駛於道路上的潛在危險與自身危險駕駛行為分析。此平台設計車輛感測資料收集系統與雲端儲存伺服器收集車載資料，針對危險駕駛行為進行分析。本研究運用深度學習智慧辨識技術，結合OBD-II即時偵測的引擎狀態資料，針對一連串連續性的行車影像進行我方與他方車道位置檢測、車距測量以及車速評估等分析。除此之外，本團隊建置危險駕駛分析網站，提供駕駛者行駛軌跡狀態以及統計資訊，協助駕駛者或監控單位快速檢視該駕駛者是否時常處於危險駕駛當中。
關鍵詞：先進駕駛輔助系統、巨量資料分析、雲端計算、深度學習、機器學習Abstract
According to Taiwan's Ministry of Health and Welfare 2018 report, around 3000 people died in car accidents in Taiwan. Furthermore, The World Health Organization (WHO) indicated that frequent car accidents in advanced countries resulted in a loss of 3~5% of GDP. Therefore, the development of smart driver assistance systems or smart transportation technologies is the common goal of the world's advanced countries which leads the rapidly developing of the Internet of vehicles in recent years. The purpose of this study is to develop and build a hazardous driving analysis platform to analyze the potential hazards and their own dangerous driving behavior on the road. This platform designs a vehicle sensing data collection system and a cloud storage server to collect in-vehicle data for analysis of dangerous driving behavior. Deep learning intelligent recognition technology was applied and was also integrated with real-time OBD-II detection engine status data and a series of continuous driving images, which could perform lane position detection, distance measurement, and the speed assessment. In addition, the study built a hazardous driving analysis website that provides driver trajectory status and statistical information to help drivers or monitoring units quickly detect if the driver is constantly driving dangerously.
Keywords: Advanced Driver Assistance Systems (ADAS), Big Data Analytics, Cloud Computing, Deep Learning, Machine Learning