| Summary: | 碩士 === 國立交通大學 === 多媒體工程研究所 === 97 === In this thesis, we propose a plant recognition system based on leaf and flower images. The images are taken in the Taichung Metropolitan Park by the camera mobile phone. There are three parts in the recognition system: flower, leaf, and combination of flower and leaf. In the flower recognition part, 14 features of flowers, including 3 shape features and 11 color features, are used. We propose a fast and automatic object segmentation method and combine user’s interaction to extract the flower region. In the leaf recognition part, 5 shape features are extracted. A two-stage approach is provided for automatic leaf recognition. Firstly, some impossible species are pruned according to the first three features. Next, the remaining species are tested to do recognition based on all five features. In the combining recognition part, a pair of leaf and flower images are recognized respectively. Then, an effective method is presented to do recognition by combining the recognition results of leaf and flower. According to experimental results, the combining recognition part can improve the recognition rate effectively.
|
|---|
