Classification of Similar 3D Objects with Different Types of Features from Multi-view Images: An Approach to Classify 100 Apples

摘要

This paper proposes a method for classifying 3D objects with similar appearances using different types of features from multi-view images. We can find this type of task in various practical applications, such as flaw inspection of industrial components, quality checking, ans screening of fruits and vegetables. In this paper, as an example such a concrete task, we will deal with the problem of classifying apples, a task that is difficult even for human vision. To tackle this task, we will introduce the mutual subspace method (MSM)-based methods as a weak classifiers in an ensemble learning framework. In addition, we will consider three types of features: shape, texture and color in the terms of invariants of position and scale, as input vectors of each MSM-based classifier. The effectiveness of the proposed method will be demonstrated through the results of evaluation experiments using 100 apples.