Acquisition of labeled data for supervised Hyperspectral Image (HSI) classification is expensive in terms of both time and costs. Moreover, manual selection and labeling are often subjective and tend to induce redundancy into the classifier. Active learning (AL) can be a suitable approach for HSI classification as it integrates data acquisition to the classifier design by ranking the unlabeled data to provide advice for the next query that has the highest training utility. However, multiclass AL techniques tend to include redundant samples into the classifier to some extent. This paper addresses such a problem by introducing an AL pipeline which preserves the most representative and spatially heterogeneous samples. The adopted strategy for sample selection utilizes fuzziness to assess the mapping between actual output and the approximated a-posteriori probabilities, computed by a marginal probability distribution based on discriminative random fields. The samples selected in each iteration are then provided to the spectral angle mapper-based objective function to reduce the inter-class redundancy. Experiments on five HSI benchmark datasets confirmed that the proposed Fuzziness and Spectral Angle Mapper (FSAM)-AL pipeline presents competitive results compared to the state-of-the-art sample selection techniques, leading to lower computational requirements.

Food fraud, the sale of goods that have in some way been mislabelled or tampered with, is an increasing concern, with a number of high profile documented incidents in recent years. These recent incidents and their scope show that there are gaps in the food chain where food authentication methods are not applied or otherwise not sufficient and more accessible detection methods would be beneficial. This paper investigates the utility of affordable and portable visible range spectroscopy hardware with partial least squares discriminant analysis (PLS-DA) when applied to the differentiation of apple types and organic status. This method has the advantage that it is accessible throughout the supply chain, including at the consumer level. Scans were acquired of 132 apples of three types, half of which are organic and the remaining non-organic. The scans were preprocessed with zero correction, normalisation and smoothing. Two tests were used to determine accuracy, the first using 10-fold cross-validation and the second using a test set collected in different ambient conditions. Overall, the system achieved an accuracy of 94% when predicting the type of apple and 66% when predicting the organic status. Additionally, the resulting models were analysed to find the regions of the spectrum that had the most significance. Then, the accuracy when using three-channel information (RGB) is presented and shows the improvement provided by spectroscopic data.