利用視覺技術辨識導盲磚之研究

Abstract

對盲人來說，導盲磚是一個重要的導引設備。盲人卻只能藉著碰觸這些含有許多資訊的導盲磚，來達到備導引的的目的。在這篇論文中所提出的視覺系統可以自動辨識導盲磚及抽取有用的資訊。此系統是針對複雜的環境，比如人行道，或者一些不同的外在因素，像是天氣，光線或遮蔽等影響的情況下所設計。這個方法包含了三個主要的步驟:前處理，辨識和資訊的抽取。在前處理中，整張影像備分割成有意義的部份。接著屬於導盲磚的部份可以經由所提出的方法辨識出來。最後，利用這些辨識出的導盲磚，一些有用的資訊可以被估測出來。實驗結果顯示，此系統有不錯的辨識及估測資訊的效果。

The guide brick is an important guidance for the blind people. The blind people are guided only by touching these guide bricks which consist of much information. A vision-based system proposed in this thesis can support autonomous guide brick recognition and information extraction. The system is developed for complex scenes, such as sidewalks, and for various outside conditions, such as weather, lighting and occlusion. There are three major stages for this system: image preprocessing, recognition, and information extraction. In the preprocessing stage, the images are divided up into meaningful regions or not. Next, the recognition stage is to extract the parts of the guide bricks by using a classifier to classify the regions. Finally, some useful information can be estimated from these parts of the guide bricks. Experimental results are provided to demonstrate the proposed scheme with good performance.