Plant encyclopedia with image matching on a web-based application: Evaluation and implementation of SURF algorithm

Abstract

The Plantcyclopedia application is developed to provide plant information. The difference between this research application and others is the search feature based on an image. The search feature is designed to match the shape and orientation. This application uses a Speeded Up Robust Features (SURF) algorithm to conduct the matching feature process. The SURF algorithm is divided into three main steps. The first step is to retrieve the image information and turn it into a matrix, using a Hessian - Laplace Technique to define the feature point following the integral to reduce the computational time. It also uses a Difference of Gaussian in order to handle the object with an invariant scale. This process is called Fast Hessian. The second step is to define the orientation of the image using a Haar Wavelet Responses Technique. The third step is to match the feature points from both images using a four dimensional vector and K-Means as a classification technique. The result of this research is an application which is able to collect plant information using image matching.