Spectral Detection and Neural Network Discrimination of Rhizopus Stolonifer Spores on Red Tomatoes

Abstract

Rhizopus stolonifer causes significant postharvest losses and about 80% of the total loss in pre-packaged and loose tomato fruits were due Alternaria rot and Rhizopus rot. The feasibility of using near infrared spectroscopy (NIR) for Rhizopus stolonifer conidia detection was studied. Visible and near infrared spectra were acquired before and after inoculating 200 tomatoes in the laboratory. The spectral data were studied using discriminant analysis, and Rhizopus stolonifer conidia were detected with an accuracy of 78%. A test set of 200 tomatoes was used for testing the algorithm, measuring the fruits only once. Spore-free and infected tomatoes were classified with an accuracy of 81 and 75%, respectively, and 96% of the infected tomatoes were properly detected by a neural network method.