臺灣博碩士論文加值系統

彩色影像處理技術的研究，是一項新的發展趨勢，尤其在國內交通領域上
，更是全新的嘗試。研究中發現，中值化處理可使原始相片之灰度值變異
降低、色彩穩定，同時由實驗觀察證明本研究建立出之色彩析出準則可代
替繁複的色彩光線計算方式。因此本研究試圖結合傳統影像處理技術與自
行發展之色彩析出方法，建立一套道路標誌自動擷取、分類系統。　　本
研究利用中值濾波法、色彩析出、收縮膨脹所建立之擷取、分類系統，經
實證研究後，標誌正確達到擷取、分類功能者有 93.3﹪，失敗之因素包
括雨天影像模糊、強烈陽光之背光效果、前車車體與標誌顏色相近等造成
誤判，但基本上色彩析出效果十分良好，證明本研究之可行性。　　本系
統所建立之道路標誌擷取、分類功能，屬於電腦視覺輔助車輛駕駛之範疇
，短期內可引為自動標誌辨識的基礎，長期則可輔助自動駕駛目標的完成
。

The purpose of this article is to build an automatic extraction
and classification system for road signs (i.e. indication sign,
service and rest station signs, and charge station sign) with
color features in the image. Using median filtering decreases
variance and increases the stability of color characteristics.
According to the color features of road signs, we establish
four heuristic decision rules: the range of hue, the fixed
order of the intensity of RGB, the ratio of the intensity of
RGB,and the range of the intensity of RGB.Experimental results
have the 93.3% correct rate, and the reasons of error including
: 1) fuzzy inage due to raining in therain, 2) front vehicular
color to be similar to some sign'' color, and 3) sign in
opposite side of strong light resources. Finally, we obtained
three major contributions in this paper: 1). It is successful
of making prior Knowledge in the establishment of color
segmentation rules. 2). The experimental results show 93.3%
correct rate. 3). A prototype of automatic road signs finder
system have been built.