VISION BASED FRUIT INSPECTION USING INDEPENDENT COMPONENT ANALYSIS

we will propose a vision-based fruit inspection method that can robustly discriminate fruit defects. This is achieved by automatic feature extraction using ICA. Comparing with PCA and HSI decompositions, we will show that ICA results in better defect discrimination. Although no order is defined for ICA components, we observed that the defect unmixing vector of each fruit is clustered in a compact region of RGB space and a simple classifier can pick up the correct unmixing vector.

کلیدواژه ها:

Fruit Inspection ، Computer Vision ، Independent Component Analysis