A Feature-Based Generic Vision Dimension Measurement System Design

[[abstract]]本文旨在改善傳統視覺量測系統都採個別專案開發方式的缺點，專案開發方式既費時且後續維護壓力大，另雖目前亦有嵌入式視覺架構以及套裝軟體於市場上使用，但在使用者操作上都有一定的困難度，其視覺量測系統可涵蓋檢測之產品類別亦相當有限，而在工具使用上，嵌入式視覺系統難以新增或修改功能，套裝式視覺系統則有許多功能屬於冗餘，有鑑於此。本文結合影像處理演算法及量產尺寸量測的高重複性而設計提出一通式化視覺尺寸量測系統，此系統軟體操作複雜度低，且僅藉簡易之教導即可讓系統涵蓋廣大產品範圍之多樣細微尺寸量測，本文所設計提出之視覺量測系統係以待測物之二明顯特徵為定位基準，依此可量測該待測物之總位置與方位角，亦可量測待測物中各幾何特徵的尺寸及相對位置，系統的運作是在大量重複量測之前先進行量測教導，以定義產品之基準特徵及建立產品中欲量測之幾何特徵及規範。[[abstract]]This research is aiming to overcome the disadvantages of the conventional vision measurement systems which are usually developed case by case due to the variety of application requirements System development through case by case is not only time-consuming but quite difficult to maintain Although there are some embedded vision architecture and software packages in the market they are pretty limited to cover wide domain of products and hard to add or modify features that are required In view of this this research proposed a generic vision measurement system that can be applied to wide domain of products directly with just simple training The basic ideas behind for a generic vision system to be feasible are that the geometry feature need to be measured are limited and can be well defind and the vision systems are mainly applied for mass production The proposed generic vision measurement system assumed that there are at least two significant features on the product to serve as the locating basis Once the locating features have been measured the position and orientation of the whole product has been determined It’s then also possible to measure the dimensions of the various geometry features on the product and the relative position among the