A Study on Fuzzy Dimension Measurement with Computer Vision

[[abstract]]本文旨在改善傳統視覺影像量測中幾何形體影像具模糊性時的處理方法，一般視覺影像量測在擷取幾何形體影像再經由一連串的處理程序後，計算獲得的幾何量測資訊都為單一量測值，不含量測誤差資訊，其量測誤範圍就只能以過去量測最大誤差量為其範圍，如此的量測誤差範圍太不精確。本文結合模糊理論定義了影像中的模糊幾何，包括模糊點、模糊距離、模糊角度、模糊線及模糊圓等以精確表達影像量測誤差，為能量測這些模糊幾何資訊，本文設計提出Fuzzy Thresholding，Fuzzy Hough Transform及Fuzzy Line Fitting方法，並進而偵測出影像中的模糊直線從而可得知模糊尺寸。[[abstract]]This paper is to address the issue of vision-based geometry measurement when the image of the geometry is fuzzy in nature Traditional vision measurement report a crispy dimension without on error associated As a result maximum error possible must be assumed For more precise reporting the geometry measurement some fuzzy geometries such as fuzzy point fuzzy line fuzzy distance fuzzy angle and fuzzy circle are defined to cope with the inaccuracy that must exist in any measurement Furthermore fuzzy thresholding fuzzy Hough transform and fuzzy line fitting were proposed to extract the fuzzy lines from a typical imag