行人偵測與行人模擬模式之研究

Study on Pedestrian Detection and Simu1ation Model

為改善長久以來交通系統中行人現況資訊不易獲得之弊，並提升行人相關設施配置事前事後績效評估的效率，本文透過日新月異的電腦科技，藉由影像處理及微觀系統模擬，建立自動化行人偵測及行人模擬評估系統。研究方法係以行人為研究對象，以在自然光或人造光源下攝影所得行人影像，經由影像擷取、影像區分、影像校正等電腦處理步驟，獲得各種不同偵測環境下的行人流量、密度、速度等相關交通參數，再經由行人二維移動行為的觀察分析，以向量分解組合的概念，發展出適用於描述行人運動特性的二維模擬模式及電腦動畫模擬輸出系統，並透過具體實例說明行人偵測與行人模擬的應用潛力與未來發展方向，最後則提出相關的結論與建議。

This paper investigates the feasibility of using computer image processing and simulation techniques to estimate pedestrian traffic parameters automatically and to evaluate pedestrian facility performance effectively. The pedestrian traffic parameters such as the flow rate, density and speed are obtained. A two-dimensional microscopic pedestrian behavior model has been developed and va!idated. The application potential and further research directions of pedestrian detection and simulation are discussed by real-world cases.