適應天候亮度之車輛防偏即時影像系

Abstract

本論文以影像為基礎，藉由一些簡單的影像處理方法與判斷法則，來實現一套能夠適應天候亮度變化的車輛防偏即時警報系統。 由於在以影像為基礎的系統，最常碰到的問題就是天候因素的影響，因其通常會而使得系統效能大大地降低，而其中又以天候亮度改變的影響最為重要，故若要使我們的防偏系統能夠更實用，首先必須先克服天候亮度的因素的影響，所以我們提出了以光敏電阻作為感測元件，隨著天候亮度的變化，來適當地調整二值化臨界值，以使系統能在任何亮度下皆能適用。 為了達到即時處理的目的，在作處理前先在影像上取兩個小視窗，僅對視窗內的影像作處理，不僅能減少資料量且能消除不必要的影像雜訊。為了得到清楚的車道線以利於作判斷，我們使用了影像處理中的梯度法邊緣強化的方法來強化車道線影像，並使用均化濾波的技巧，有效地減少了天候較暗時及陰影下會有較多雜訊的現象，最後，提出一套簡單的判斷法則，以偵測車輛是否有偏離車道的狀況，並將其實現在即時的防偏系統上。

In this thesis, a vision-based lane departure warning system can adapt the variation of the brightness is accomplished. Certain image processing methods and decision rules are developed to achieve the goal. The most important factor of the weather is the variation of brightness. A method, which uses the photo-resistor as the sensor to measure the brightness, is proposed in this thesis. According to the variation of the brightness, the adaptive threshold of dichotomy is adjusted automatically. For the purpose of real-time processing, data in two windows on the image are sampled. By applying the proposed image processing techniques: (a) lane strengthen (b) dichotomy (c) smoothing filter, suitable image of lines on the road is obtained. Furthermore, smoothing reduces noise under darkness and shadow. Finally, a simple decision rule used to realize the lane departure system is developed.