互動式的花朵辨識系統

Abstract

本篇論文中我們提出一個互動式花朵辨識系統，針對由數位相機 直接從真實景色中拍攝的花朵影像進行辨識。花朵的形狀以及顏色是 辨識上相當重要的特徵，在本篇論文中，總共使用了十四種特徵，針 對花朵外型有三種，另外十一種為顏色特徵。我們提出一個與使用者 互動的方法，加以限制花朵擷取的範圍，解決在花朵叢生所造成的重 疊影像中，單一花朵輪廓擷取的困難。接著利用平均成本方法(NC method)進行邊界擷取。系統針對348 張影像(共24 種)做實驗，實驗 結果顯示可以有效的進行輪廓擷取並且辨識花朵的種類。

In this thesis, we present an interactive method for recognizing flower on a flower image taken by a digital camera in the real scene. Shapes and colors of flower are very important features for recognizing, so fourteen features are used in our recognition system: three for shapes and eleven for colors. We provide an interactive method to limit the extraction area to solve the problem at extracting flower boundary in the overlapped flower image. Then we implement a normalized cost (NC) method for boundary extraction. Experiments are conducted on the database of 348 flower images to demonstrate the performance of the method, results show the effectiveness for recognition.