Boosting minimalist classifiers for blemish detection in potatoes

This paper introduces novel methods for detecting blemishes in potatoes using machine vision. After segmentation of the potato from the background, a pixel-wise classifier is trained to detect blemishes using features extracted from the image. A very large set of candidate features, based on statistical information relating to the colour and texture of the region surrounding a given pixel, is first extracted. Then an adaptive boosting algorithm (AdaBoost) is used to automatically select the best features for discriminating between blemishes and non-blemishes. With this approach, different features can be selected for different potato varieties, while also handling the natural variation in fresh produce due to different seasons, lighting conditions, etc. The results show that the method is able to build “minimalist” classifiers that optimise detection performance at low computational cost. In experiments, minimalist blemish detectors were trained for both white and red potato varieties, achieving 89.6% and 89.5% accuracy respectively.