IT And Machine Vision For Closed Plant Production Systems

The closed system is characterized by a warehouse-like structure covered with opaque thermal insulators which ventilation is kept at a minimum, and artificial light is used as the sole light source for plant growth. Advantages of the closed system for producing high quality plantlets using photoautotrophic micropropagation were demonstrated in a 20 m² practical closed system. Contamination in the closed system was controlled with cleanness of ISO7-grade. The physical environment in the closed system was maintained in high relative humidity within gradient of 10% and in a uniform distribution in temperature, CO₂ concentration, and photosynthetic photon flux. It can be concluded that quality and productivity of transplants are definitely higher and the growth period of transplants can be shortened by approximately 30% when produced in the closed system using lamps than in the greenhouse using sunlight. The closed system is energy and material efficient especially with respect to the amounts of water required for irrigation and energy required for heating in winter.An environmental measurement system using wireless networks and web-server-embedded technology is developed to measure various data and image information as a digital infrastructure for contributing the digitalization and remote sensing. The developed system equipping various sensors such as temperature, relative humidity, photosynthetic photon flux, CO₂ concentration, UV etc. sensors and web-cameras is achieved to remote sensing in real-time and low-cost. The actual running experiments were examined in Japan, China, United State, Thailand, and Demark to test the remote sensing performance. Comparisons between the developed systems to the commercial systems for environmental measurement were conducted in a field and a tissue-culture room. The compared results could indicate that the developed system is one of the most promising tools for environmental measurement and will be very powerful for digitalization and remote sensing in precision agriculture.In order to achieve optimization of transplant production using IT and artificial intelligence, a time-delay artificial neural network (ANN) model combined with a binocular stereovision system was developed for predicting growth variables of the transplant population. The ANN model consisting of 9, 8 and 5 processing nodes in the input, hidden and output layers, respectively, was trained based on backpropagation algorithm. Four growth variables of average height, leaf area, projected leaf area, and mass volume of the transplant population obtained by the machine vision system, four environmental factors of daily average temperature, photosynthetic photon flux, CO₂ concentration and relative humidity for the following day, and days after planting were used as the ANN model inputs. The ANN model outputs were five growth variables of average height, number of unfolded leaves, leaf area, and fresh and dry masses of the transplant population for the following day. The predicted errors of the growth variables using a sweet potato (Ipomoea batatas (L.) Lam.) transplant population obtained from the ANN model in testing were less than those obtained from a traditional regression model. The growth prediction of the transplant population could be realized successfully using the ANN model combined with the image analysis system.