植物與蝴蝶自然影像之自動化辨識系統

Abstract

二十世紀以來，隨著自然保育觀念的推廣，人們逐漸意識到生態教 育的重要性，例如今年於台北市舉辦的「2010 臺北國際花卉博覽會」， 從而發展相關的應用工具來輔助教學，例如：圖鑑、海報與專業的生物 及生態查詢網站。然而對一般民眾而言，這些工具除了有「攜帶不便」 的問題外，在檢索資料時，常須輸入關鍵字，然而，民眾本身多不具相 關專業背景，「關鍵字選擇」成為大多數人的一大難題。 為了解決以上問題，自然影像自動化辨識的概念遂被提出。使用者 可藉由簡單的操作，在影像中選出想要辨識的物種，再經由自動化影像 識別的技術，回饋使用者所要查詢的物種名稱及相關資訊；如此將可大 幅降低使用者的負擔與檢索資料的門檻，從而帶動生態教育的推廣與落 實。 由於其便利性及實用性，自動化辨識已成為未來的趨勢，惟考量植 物與昆蟲影像之不同特性，因此本計畫擬分兩年開發植物與蝴蝶的自動 化辨識應用系統。第一年擬設計一個植物影像自動化辨識系統，利用從 真實自然環境中拍攝的花朵以及葉片影像進行辨識。第二年擬設計一個 全新的蝴蝶影像自動化辨識系統，針對從真實自然環境中拍攝的全方位 蝴蝶影像進行辨識。藉由自動化辨識系統，提供使用者所要查詢的資 訊。民眾透過此系統可以簡單且快速的學習植物、蝴蝶方面的知識。

Since the twentieth century, with the promotion of nature conservation, people gradually realize the importance of ecological education, for example, the Taipei International Flora Exposition which is held by Taipei City Government, and start to develop the related application tools to assist it, such as field guide, poster and professional biological/ecological website. Nevertheless, for ordinary people, these tools are not only unportable but also have a serious problem. Since ordinary people do not have enough professional background knowledge, it is hard for them to choose the suitable keywords while searching. To overcome these problems, the technology of automatic recognition on natural image has been proposed. Through simple user interaction, the position of interested creature in the image can be located. Then, by the process of automatic image recognition, user can easily obtain the corresponding scientific name and related information of the queried creature. By automatic recognition, the burden and the searching qualification of user can be substantially reduced, and the nature conservation can hence be popularized. Due to the convenience and practicality, automatic recognition has become the future trend. Therefore, our goal is to build two automatic recognition systems for plant and butterfly. In the first year, we will focus on designing an automatic recognition system for plant on natural image. This system can recognize flower and leaf images taken in natural scene, and provide the corresponding information to users. People can learn the plant's information simply and effectively by the system. In the second year, we will focus on designing a novel automatic system for butterfly recognition on natural image. This system can recognize butterfly from a natural image which can be taken from various shooting directions in the real scene. With this system, user can effectively acquire the information of desired butterfly just by simple operations.