Flores, Huber, juhendajaRao, KarinaTartu Ülikool. Loodus- ja täppisteaduste valdkondTartu Ülikool. Arvutiteaduse instituut2023-10-302023-10-302023https://hdl.handle.net/10062/93851While produce quality estimation across various stages in the value chain is essential to tackle food loss and waste, determining pesticide residue in fresh produce can alleviate the threat to human health and the environment. Light sensing offers a non-invasive and cost-effective method to establish unique fingerprints for fresh produce. During a 12-day produce decomposition period, it was established that light reflectivity is effective for the quality estimation of vegetables. The AdaBoost classification model with blue light reflectivity value, vegetable items and luminosity as input features achieved a performance accuracy of 92.4%. While measuring reflectivity intensity, it is important to account for varying lighting conditions (luminosity). Notwithstanding the success of predicting the quality of fresh produce, light sensing failed in pesticide residue estimation.engopenAccessAttribution-NonCommercial-NoDerivatives 4.0 InternationalLight sensingproduce qualitypesticide residuemachine learningmagistritöödinformaatikainfotehnoloogiainformaticsinfotechnologyThesis