ADAPTIVE SORTING BY PROTOTYPE POPULATIONS

An algorithm for unsupervised adaptive sorting is presented, based on a finite number of 'prototype populations', with distinctly different feature distributions, each representing a typically different source population of the inspected products. Updated feature distributions, o f samples collected from the currently sorted products, are compared t o the distributions of the stored prototype populations, and according ly the system switches to the most appropriate classifier. Although th e goal is similar to the objectives of previously proposed 'Decision D irected' adaptive classification algorithms, the present algorithm is particularly suitable for automatic inspection and classification on a production line, when the inspected items may come from different sou rces. The practical feasibility of the approach is demonstrated by two synthetic examples, using Bayes classifiers. This is followed by an a pplied example, wherein two prototype populations of apples are sorted by size, derived by machine vision. It is shown that misclassificatio n by adaptive classification is reduced, in comparison to non-adaptive classification.