Machine Learning for Automatic Classification of Tomato Ripening Stages Using Electrophysiological Recordings

• DOI: 10.3389/fsufs.2021.696829
• Publication Date: 2021-10-29T07:16:34Z
• Authors: Gabriela Niemeyer...; Thiago Francisco ...; Ádrya Vanessa Lir...; André Geremia Parise; Danillo Roberto P...; Gustavo Maia Souza
• Keywords: 2. Zero hunger; Global and Planetary Change; 0303 health sciences; Ecology; post-harvest technology; fruit electrical signals; Nutrition. Foods and food supply; Solanum lycopersicum var. cerasiforme; Horticulture; Management, Monitoring, Policy and Law; TP368-456; Food processing and manufacture; 03 medical and health sciences; electrome; TX341-641; plant electrophysiology; Agronomy and Crop Science; Food Science

The physiological processes underlying fruit ripening can lead to different electrical signatures at each stage, making it possible classify tomato through the analysis of signals. Here, activity ( Solanum lycopersicum var. cerasiforme) during was investigated as tissue voltage variations, and Machine Learning (ML) techniques were used for classification stages. Tomato harvested mature green stage placed in a Faraday's cage under laboratory-controlled conditions. Two electrodes per inserted 1 cm apart from other. measures carried out continuously until entire fruits reached light red stage. time series analyzed by following techniques: Fast Fourier Transform (FFT), Wavelet Transform, Power Spectral Density (PSD), Approximate Entropy. Descriptive FFT, PSD, PCA (Principal Component Analysis). Finally, ApEn, PCA1, PCA2, PCA3 obtained. These features ML analyses looking classifiable patterns three stages: green, breaker, red. results showed that is stages using fruit's activity. It also observed, precision, sensitivity, F1-score techniques, breaker most among all found more accurate distinction between × than seem be novel tool classifying obtained electrophysiological have potential supervised training, being able help