Learning Fuzzy Rules with Boosting for Visual Object Classification
Rizwan Ayazuddin · Preprints.org · 2025
Image classification remains a fundamental challenge in computer vision, requiring models that can handle high-dimensional, noisy, and uncertain data. It is a key task in computer vision, but the variability and uncertainty of local image features make it difficult to design accurate classifiers. Fuzzy logic provides a natural way to handle imprecision, while boosting improves weak learners into strong ensembles. Traditional classifiers often struggle to balance interpretability and performance when dealing with high-dimensional descriptors such as SIFT. There is a need for a method that can generate effective, yet understandable, classification rules while maintaining robustness to noise and feature variation. In this work, we examine boosting-generated simple fuzzy classifiers for visual object recognition. We describe the process of generating fuzzy rules from local descriptors, show how boosting combines these weak fuzzy rules into a strong classifier, and discuss the strengths and limitations of this approach compared to conventional methods.