以影像處理估測交通流量及其應用於獨立路口的模糊適應性交通號誌管理之研究

Abstract

交通資訊偵測以及適應性交通號誌管理，是現今智慧型交通系統(intelligent transportation system, ITS)的發展重心。將偵測到的交通資訊應用在獨立路口的適應性交通號誌管理上，可以動態地依照當時的交通狀況調整時相長度，減少該路口擁塞的情況發生。目前使用影像處理技術來偵測交通流量的研究相當受歡迎，但它們對交通資訊的偵測大都使用微觀尺度的量測，這對交通號誌管理的應用其實是不必要的；另外在適應性交通號誌管理的方面，由於具有動態變動的週期長度，所以在設計時常常不會考慮到最佳週期而使得週期過長。有鑑於此，本論文的主要目的是發展一套有效率的交通流量估測系統與具有適當週期長度的適應性號誌系統。 本論文的主要貢獻有三，第一，利用透視轉換(perspective transformation)得到前景像素總數與前景物件數目之間的關係，搭配預先建立好的橢圓行人樣版，可以大約地估測出行人穿越道上的人數。第二，利用前述的方法並配合影像紋理分析(texture analysis)，可以將道路或是特定車道上的車流程度以[0, 1]之間的數表示。第三，提出一個以模糊推論系統(fuzzy inference system, FIS)為核心的模糊適應性交通號誌管理系統，並將最佳週期以及相對路口飽和度考慮於系統的設計當中。實驗顯示本論文所提出的影像處理演算法及模糊適應性號誌管理系統在不同的實驗場景之下皆有不錯的效果。

Traffic information detection and adaptive traffic signal control are vital to the development of intelligent transportation system. The cycle length of a traffic signal controller can be adjusted dynamically by applying gathered traffic information to adaptive traffic signal control at the isolated intersection, thus the traffic jams can be reduced. Detecting traffic information by use of image processing has become a trend, however, most of the previous research use microscopic measurement, which is unnecessary in application to adaptive traffic signal control. For adaptive traffic signal control system, the topic of optimal cycle is rarely considered since it has various lengths in each cycle; therefore, a very phase could have an overlong phase time. In view of this, an efficient image processing algorithm is proposed to estimate traffic flow in this thesis. In addition, an adaptive traffic signal control system that takes optimal cycle into account is presented. The contributions of this thesis may be summarized as follows. First, the relation of the number of foreground pixels and the number of foreground objects can be obtained by using perspective transformation. With the pre-constructed ellipse human template, the number of pedestrian on a crosswalk can be estimated approximately. Second, use the method mention above together with texture analysis of an image, the traffic flow can be normalized and represented by a number between 0 and 1. Third, a fuzzy adaptive traffic controller based on fuzzy inference system is proposed. The design of the system also takes optimal cycle and relative saturation degree of different roads into consideration. The image processing algorithm and fuzzy traffic signal controller have been tested in various situations; the system shows promise and the experiment results are satisfactory.