Herbs identification trough leaves recognition / Hayatun Syamila Sholeh

The leaf contains important information for plant identification including herb plants. This project presents a prototype designed to recognize herbs plant using leaf images. The prototype focuses on six types of herbs plants available in Malaysia, that are Cabang Tiga, Bebuas, Kaduk, Ginseng Jawa, Sireh and Pegaga. The prototype uses Zernike Moment Invariant features and similarity match Template Matching Classification to classify the leaf. Twenty four images have been collected. Twelve from top and twelve from bottom view. All the images captured with white background. Firstly, the image will be resized, converted into grayscale and enhanced to get local range of the image. Then, the value of Zernike moment invariant is computed for each leaf image. The plant identification is performed using template matching where the rules for each leaf is constructed based on the collection of values of the Zernike moment of invariant for eight leaves from each herb type. The output of this prototype is the name of herbs species that have been tested with test image. Accuracy of recognition calculated by count how many true species name displayed using Precision and recall rate. It is found that the Zernike moment invariant feature enables the template matching classifier to identify the herb better based on the bottom view of the leaves image. Thus, the finding from this prototype can provide useful information for developing an automated plant identification tool.