A rotating hydroponics for lettuce cultivation with fuzzy-based adaptive speed control using computer vision-based spectral phenotypes

• Main Author: Aquino, Heinrick L.
• Format: text
• Language: English
• Published: Animo Repository 2022
• Subjects: Hydroponics; Lettuce—Health; Lettuce—Growing media; Computer vision; Electrical and Computer Engineering
• Online Access: https://animorepository.dlsu.edu.ph/etdm_ece/22; https://animorepository.dlsu.edu.ph/context/etdm_ece/article/1022/viewcontent/A_Rotating_Hydroponics4_for_Lettuce_Cultivation_With_Fuzzy_Based_A__1__Redacted.pdf
• Institution: De La Salle University

In the coming years, the world population is expected to go as high as 9.8 billion in 2050, according to the United Nations organization. This phenomenon leads to the increasing requirement for food supply and space. To address these issues, experts see indoor urban hydroponic farming as a solution to meet the demand. Various studies about the applications and yield of different hydroponic configurations, such as the Nutrient Film Technique (NFT), are available; however, experiments about the potential of rotating hydroponics are still minimal. The health of leafy vegetables, such as Lactuca sativa, is commonly determined through inspection of their appearance. Hence, early detection of any manifestation of diseases on lettuce leaves could prevent the further destruction of the yield. In this research study, computational intelligence models were utilized in the development of lettuce health classification through extracted spectral features of lettuce. This computer vision-based and non-invasive method of assessing healthy and chlorotic leaves, or the yellowing and white discoloration on leaves caused by drought or lack of light, eradicates the manual, labor-intensive, and subjective assessment of lettuce health. The classification result was then used in adjusting the adaptive rotation of the growery through fuzzy logic. This is responsible for the fertigation or the absorption of the water-nutrient mixture appropriate for cultivar's needs; hence experiments to know lettuce growth under different rotation speeds were conducted. The system is operated through sensors in control for data acquisition, microcontroller, and actuators. Data logging was done wirelessly and can be monitored via a cloud-based website. The computational models were trained using the 533 cultivar images collected in the initial cultivation and evaluated for their accuracy. System performance was evaluated by its average fresh weight yield with adaptive rotation. The results of this work showed that L*,a*,b* spectral features and SVM model is most suitable in this application of lettuce health classification, with SVM having 100% accuracy and the fastest machine learning model with 36.66 seconds inference time. The 0.75 rpm to 2 rpm growery speed provided good lettuce growth, and it was observed that greater rpm means greater lettuce growth performance due to more nutrient absorption opportunity. However, as speed increases, the percentage increase in growth decreases. Hence, energy-wise, faster speeds than 2 rpm will not be practical anymore. Compared to rotating hydroponics without rotation, the experiment with adaptive rotation recorded an increased 39.61% of average fresh weight yield. In addition, the system fresh weight yield is 12.92% and 44.12% higher than the recorded yield in other NFT and rotating hydroponics studies. Index Terms—rotating hydroponics, lettuce, computer vision, machine learning