Outlier Detection of Leaf Images Based on One-Class Support of Vector Machine

Abstract

In this paper, a novel approach was proposed to use the One Class Support Vector Machine (OCSVM). The growth status of fruit tree's leaves reflects the health level of the plant. Using computer system to detect the abnormal condition of plant leaves will help producers take measures to maintain the health of plants. We focus on the outlier leaf images which are obviously different from with the health leaf. A modified One Class Support Vector Machine is presented as a very efficient method to detect the outlier data from massive data of the leaves image dataset. The results show it worked with the test dataset and with very good performance.