智慧型車輛周圍環境偵測型與安全駕駛研究

Abstract

本計畫之目標為智慧型車輛周圍環境偵測與駕駛行為安全之研究，在演算法核心的 部分，本計畫將基於前期計畫的執行結果，開發前方以及盲點的行人偵測，讓智慧車在 往前行駛以及過彎時，能徹底避免不慎與行人碰撞之事故。除此之外，本計畫亦會開發 出一套停車輔助演算法，除了讓駕駛能輕鬆簡單的倒車入庫之外，也會將這套演算法移 植至總計畫的輪型運輸機器人上。 本計畫與總計畫、子計畫四、五、七均有緊密的連結，透過 3.5G/3G 無線車用嵌入 式平台的建立。當車輛在行駛中，可觀看到其他路口的即時影像，進而判斷行車途徑。 若車輛不幸遭竊，除了竊賊的面容可即時看到，搭配GPS 資料，可鎖定車輛位置，快速 找回失車！ 在內容方面，本計畫主要分為三個主題：(1) 建立 3.5G/3G 車載平台、開發倒車輔 助演算法，(2) 開發前方行人偵測技術、並結合子計畫七之高效率視訊演算法，將路口影 像即時傳送至車載平台以及行動通訊裝置、將倒車輔助演算法移植至輪型運輸機器人， (3) 在車載平台上開發盲點行人偵測技術以及先進車輛防盜系統，並將駕駛者資訊透過無 線傳送給分析系統。本計畫之宗旨為完成一具有國際水準之全方位智慧車，期盼提升台 灣之車用電子於全球市場佔有率並藉由科技的力量提高台灣的國際形象及競爭力！。

This subproject is a three-year project, which focuses on the development of the automatic vision-based detection system for the environment surrounding a vehicle and the research of the vehicle driver’s behaviors. The content of this project is separated into three parts: (1) Establish the 3.5G/3G wireless embedded platform; develop the parking assistance algorithm; and draw the corresponding map using the GPS data. (2) Derive the front view pedestrian detection technique; transfer the videos of the intelligent intersection systems, built in the main project, to the embedded platform and the mobile devices through the efficiency video processing algorithms, provided by the subproject seven; and port the parking assistance algorithm to the transportation robots which are implemented in the main project. (3) Design the blind spot pedestrian detection algorithm and the advanced burglarproof system. The final objective of this project is to accomplish an intelligent vehicle which can automatically detect the lanes, the obstacles, and the pedestrians around it. We hope to improve the technologies in the autotronic fields, and better the international images of Taiwan.