Research On Intelligent Physiological Engine For Virtual Plant And Its Key Technologies

Research on virtual plants is one of the important aspects of digital agriculture and agricultural informationization, and is presenting great theoretical and practical value. However, there are still some problems and difficulties between research and application because of the complexity of virtual plants being as an interdisciplinary work, such as being difficult to extract growth rules, difficult to configure complex parameters, difficult to present complicated functional structure of organs by using the iteration mechanism of the traditional L-system, difficult to accurately model matter transport, difficult to model the intelligent behavior of plant based on the traditional modeling techniques, etc.According to these limitations, this thesis conducted intensive studies on virtual plants modeling and its key technologies. The most important feature of this thesis is that the Intelligent Physiological Engine incorporated artificial life, machine learning, computational intelligence, simulation technologies, plant science, plant physiology and system biology, etc. can automatically extract growth rules of plant without help of specialist. Moreover, the complex intelligent behaviors of plant can emerge from simple interactions of intelligent organs using our methods. Main contributions are as follows:First, multiple images, contour recognition and 3D skeleton recovery have been used to automatically measure the shape, topology and geometry of metamers and growth units. The hidden Markov tree model has been used to analysis aforementioned information according to the time-sequence of plant growth and development. Hidden states and their transitions of hidden Markov tree are mapped to the differential states of meristems and finally are transformed into growth rules by means of the construction of the Two Dimensional Hierarchical Automata (2DHA).Second, the structural-functional model of virtual metamer has been established using intelligent agent approaches. The virtual metamer with inbuilt physiological status and procedures, capability of dynamic deformation and reasoning can autonomously respond to environmental heterogeneity. It also has the key features of keeping dynamic balance between carbon production, reserve, consumption and transport, and finding the best position for child metamer to maximizing light interception.Third, the discrete pressure-flow model using the pattern of graph automaton has been established based on the discrete model of plant vascular system. The water, nitrogen and carbon are simultaneously transported and the lateral osmosis between xylem and phloem is considered as well. Simulation results show that our model can conform to the theory of plant physiology.Fourth, the concept of complex adaptive system has been used to model the features of plant: modularized structure, intelligent behaviors of phenotype plasticity emerged from relatively simple physiological procedures and interactions. The structure of plant has been decomposed into individual virtual metamer constructed using intelligent agent. Simulation results show that some emergent behaviors have been revealed from different level of the virtual plant.Finally, the Intelligent Physiological Engine (IPE) for virtual plants has been implemented using modeling approach and technologies aforementioned. It is a simulator with loosely coupled components, such as user interface, virtual environment, dynamic branching network, matter transport engine, branching controller and virtual organ library. The IPE takes physiological status and environmental data as input, processes plant structure variation resulted from internal physiological procedures, and outputs plant status and structure for 3D rendering. Several simulations of eggplants'growth under normal, lower temperature, light and water and nitrogen deficit environment have been conducted using IPE. The comparison between simulated and real data shows that the IPE can effectively and vividly model the growth and development of virtual plant and its intelligent behaviors.