用統計和物形特徵作商標影像辨認

Abstract

商標審查的目的是要將相似的新申請商標予以駁回。為了簡化人工審查商 標的繁雜處理過程並提高審核的效率，商標審查系統自動化處理的研究是 亟需的。本論文提出一套新的商標影像辨認方法，此方法的基本原理是先 從商標圖形中抽取出特徵，再利用這些特徵值在資料庫中找尋較相像的商 標。首先系統根據商標圖案中的適當形狀特徵，將不同的商標加以大略分 類，再依統計式和形狀式的特徵值，擷取相似的商標。商標圖形千變萬化 ，並且審查過程主觀意識濃厚，使得系統的困難度頗高，故本方法目前以 半自動處理為主。此外，系統提供交談式操作，以方便作前處理和特徵抽 取。實驗結果顯示所提的方法可行，而且實用價值相當高。 The goal of trademark examination is to reject infringement cases when trademarks are registered for patents. In order to simplify and speed up the trademark examination process, the development of a trademark image recognition system is desired. A new approach to trademark image recognition is proposed. The basic idea is to extract features from the input trademark image, and to search similar trademarks in a feature database. The proposed system provides friendly interactive interfacing to perform the preprocessing and the feature extraction works. Appropriate shape features of the trademark images are utilized to perform preclassification. Detailed matching is performed next according to the combination of the statistical and shape features. Trademark graphics are of very great variety. Due to the subjective judgment on trademark examination and the difficulty of this task, only semi-automatic processing is implemented in this study. Several experimental results show the feasibility and practicability of the proposed approach.