一個模型是由數個特徵所構成，其中刻字特徵屬於較細微的特徵，作為商標、文字註釋與型號的功用廣泛，各種產品皆有可能存在刻字特徵，並以群體的方式出現，而扁短的刻字特徵在成長網格上容易產生品質較差(狹長、不平滑)的網格，故本研究以刻字特徵為主要研究對象，期望以自動化來取代傳統人工方式辨識刻字特徵，並提供使用者移除刻字特徵功能。由於刻字特徵可能處於多重曲面上，在刻字辨識前先加入虛擬迴圈(Virtual loop)辨識技術，先找出所有拉伸物件(Extrusion)的位置，而刻字特徵與非刻字特徵的Extrusion並列而難以區別，因此先將Extrusion群辨識後，藉由相同高度與所在的基準面進行分組，並由本研究開發的多樣性(Diversity)來得到每一組Extrusion的種類數量，以進行刻字判定，最後以B-rep資料結構記錄辨識後的刻字特徵相關資訊，並將這些資訊移除來達到提升前處理效率的目的。本研究測試之36個CAD模型中，藉由自動辨識方法成功辨識的有34個(94%)，2個未完全辨識的特例情形可藉由手動方式選取來達到100%的辨識率。;A group of characters are typically engraved on a computer aided design (CAD) model to represent a trademark, text notes, product model, etc. Such a group of characters are called lettering. Lettering recognition is a technique to recognize all kinds of characters on a CAD model. The objective of this study is to propose an automatic lettering recognition algorithm to recognize all characters from CAD models. The proposed lettering recognition algorithm is as follows. The virtual loops of a boundary representation (B-rep) model are recognized first; with the virtual loops, all extrusion features can be identified. Then, a series of rules are employed to separate “lettering” and “non-lettering” extrusion features. We propose a concept called “diversity” to classify and group extrusion features of similar geometry, and recognize all characters on the model. Once all characters are recognized, a character-removing algorithm can be implemented to remove all characters on the model. The feasibility of the proposed algorithm was tested by using 36 realistic CAD models, in which all characters on 34 models (94%) were successfully recognized, while some of the characters on 2 models were not completely recognized. A manual lettering recognition algorithm was also proposed to remove those characters which cannot be recognized by using the automatic letting recognition algorithm.
