Novel Dictionary Pair Learning Through Locality-Constrained Profiles for the Identification of Chili Pepper Varieties via E-Noses

In modern society, to make it easier for customers to assess product quality and regularize distribution and supply, almost all crops are sorted and labeled. This sorting and labeling process also facilitates packaging and transportation of products and benefits farmers. As an emerging technology, electronic noses (E-noses) can be applied to food quality control. In this study, we proposed a novel dictionary pair learning (DPL) model named DPL based on locality constraint of profiles (DPL-LCPs), which achieved better performance in the identification of chili pepper varieties using E-noses. This model incorporates profile vectors to define a profile locality constraint term and integrates it with a dictionary reconstruction error term into a unified dictionary learning (DL) model to fully learn the characteristics of samples from various aspects. This approach not only reduces the similarity between dictionary atoms but also avoids the problem of weakening the connections between characteristics of different classes of samples caused by using the block diagonal constraint. In addition, the model utilizes the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${l} _{{2},{1}}$ </tex-math></inline-formula> norm to regularize the analysis dictionary matrix, enhancing the sparsity of features and thereby enhancing the flexibility and robustness of the model. The experimental results show that the proposed model can achieve better classification results and higher stability than other advanced DL models can. With this method, it is possible to reliably identify 13 varieties of chili peppers based on flavor assessment.