**4. Proposed Recognition Method**

Image recognition has a fixed framework. In general, for plant recognition, object images acquired by special devices (e.g., camera or scanner) are used. In this paper, we select the leaf datasets with clean background for identification. Some key features which can identify the object are extracted from the images through various algorithms. The classifier is employed for classification after feature extraction. Most classifiers need to be trained by samples before classification. Finally, the result is obtained. The proposed method of leaf image recognition also adopts the above framework. The detailed scheme of the proposed method is shown in Figure 6. It can be divided into three steps: leaf image preprocessing, leaf feature extraction, and recognition. These steps are explained in the following.

**Figure 6.** Scheme of the proposed method.
