應用時空分析方法建立動態交通事故預警系統之研究

Abstract

交通事故到了2020年其嚴重性將勝於結核病、癌症、呼吸道疾病，其所造成的死亡人數都超越這些疾病，被認為是當今世界最大公害之一。過往對於事故的預防傾向於較被動的預防，宣導、工程面改善、加強法規等方式，而人們今天上路時仍舊無從得知當下自己所在位置哪邊是危險的。本研究透過時空探索分析證實事故無論是在時間與空間上皆呈現非均勻分布，並接著透過時空掃描方法進行事故熱區的搜尋，原因在於熱區非一成不變，時間尺度深深影響著事故的變化，不同的時間會產生不同的熱區，透過時空掃描所得到的結果，除了空間上的熱區範圍，更能知道熱區何時發生、何時結束、持續時間長短、風險值，而經過分析，發覺事故確實會因為不同時間產生出不一樣的熱區結果，事故在一個禮拜中不同天與一天中不同小時皆存在顯著的時間規律性，預警將不再是一成不變，而是動態變化的過程，時間上的動態變化與地圖上的動態變化，這也是本文與過往研究最大不同與創新之處，如此的預警將更具有意義。 將時空掃描結果應用於事故預警系統。主要功能為透過行動裝置地理資訊系統展示地圖，以及動態的早期預警功能。在無線網路、行動裝置的成熟，以行動裝置軟體 (App) 帶動了適地性服務的可能背景，在事故預警App執行後，即時傳遞動態事故熱區資料，並藉由主動的預警方式，以文字、地圖圖像、語音去進行早期預警，提醒使用者避開危險熱區，期望能往更安全的都市道路環境邁進。

The casualty involved in traffic accident will be even worse as a greatest problem than that involved in tuberculosis, cancer, and respiratory diseases in 2020. It is thought of as the biggest public hazard in the world nowadays. In the past, accident prevent method was prone to passive prevention like as propaganda, engineering, strengthen regulations, however people still didn’t understand where was dangerous. This study used spatio-temporal exploratory method to confirm car accidents were not uniform distribution in time and space. We searched car accident hotspots by space-time scan statistics method because it was not always stationary, and the time criterion deep was affecting the accident change deeply, the different time caused the different hotspots. Space-Time Scan can understand hotspots range, starting time, ending time, duration, and risk-value besides research result confirm car accidents appeared at different locations by time flowing. Car accidents were significant timing regularity at days of week and hours of day, so early warning was not same like past it was a dynamic changing results including timing changing and user positions changing. Compare to past, new early warning way was different and creative, and such result was more meaningful. Accident alert system deliver information by internet and provide a new way for active alerting. Main function contain dynamic early warning. Currently wireless internet and mobile devices are ripe, and they drive Location-Based Service (LBS). When accident alert App work, it will deliver dynamic accident hotspots and real-time traffic data immediately actively alerting way by word, map, voice to remind user avoiding from dangerous intersection and sudden road event. Hoping can forge ahead to a safer city road environment.