Image Analysis For Leaves Recognition And Retrieval System

• Main Authors: Huang,Jiayang, 黃嘉洋
• Other Authors: Lawrence Y. Deng
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/37751069051751021080

碩士 === 聖約翰科技大學 === 資訊工程系碩士班 === 100 === This research aims to create a leaf recognition system, which helps recognize unknown plants by their leaves. User captured images of the leaves and uploaded to our system. The system will return the name and characteristics of this unknown plant to the user immediately. The most challenged issue is how to recognize leaves within the complex background. A normalization method for sampling was created based on the experiment and discussions. Firstly, using an A4 sized paper as the background when capturing pictures. So, the leaf could be easily extracted from the simple background. Then, adjusting the distance between the camera and the sample till the entire A4 background paper was covered. The exact size of the leaf could be calculated based on the A4 size. However, it was found that the illumination could affect the accuracy of the recognition. Therefore, a series of experiments and analyses were conducted and the best method would be applied to the normalization (process). Every island has specific plants. Users can upload the samples to the database, following the normalization process. A critical characteristic of this research is that users are involved in inserting samples in the database. It is expected that the normalization method and the control of sampling circumstance can increase the convenience and the scalability of the database by experiments, analyses, and discussions.