Olfactory visualization sensor based on densely connected convolutional networks for in-situ fruit ripeness prediction

Abstract

The classification of ripeness levels is one of the most important indicators to assess fruit quality, and a technique to achieve in-situ ripening grading is urgently needed. We constructed an olfactory visualization sensor based on Pd2+-dye/NH2-UiO-66 as the gas-sensing materials and densely connected convolutional networks (DenseNet) as data classification method for in-situ fruit ripeness differentiation by ethylene detection. The designed Pd2+-dye/NH2-UiO-66 composites showed extremely sensitive sensing performance for C2H4 detection due to the high loading capacity for colorants (Pd2+ and dyes) and pre-concentration for C2H4. A series of Pd2+-dye/NH2-UiO-66 composites were synthesized using eight pH dyes with different pH sensitivity ranges, which were used to fabricate the colorimetric sensor arrays for C2H4 specific detection. Taking advantage of the greatly sensitive and specific nature of C2H4 detection, the colorimetric sensor arrays were applied for in-suit fruit ripeness classification by integrating DenseNet, achieving up to 99.91% classification accuracy. The combination of Pd2+-dye/NH2-UiO-66 composites based on highly sensitive and specific colorimetric sensor arrays and the supervised image classification method DenseNet enables high-precision prediction of different fruit ripeness and opens new possibilities in the low-cost and on-site classification of ripeness.