Optimized Layout Of Spatial Objects Based On Genetic Algorithms And Scene-Building

In the"text-to-scene"research, which is an area in combination of natural language understanding and virtual scenes generating, the key is to understand the spatial concepts. Nature Science Foundation project"the 3D visualization of spatial relationships in text based on ontology"is just focus on the extraction of the spatial relationships and their visualization in articles. Our research work is an important part of the project, focusing on the research and implementation of optimized layout of spatial objects and automatically generating of 3D scenes.To layout objects in a finite three dimensional space efficiently and automatically is an important issue in many application fields. Especially in the field of Text-to-Scene and Virtual Reality, automatic and optimized layout of objects is the prime problem in creating 3D scenes.Currently, there are mainly two kinds of research methods on this problem: meta-heuristics approach and randomized algorithms. Because our research is about optimized layout of entities whose quantity is arbitrary, it's difficult to find the applicable meta-heuristics rules. And in random algorithms, simulated anneal and simple genetic algorithm are not only inefficient, but also can not guarantee the completeness of the results. In this paper, we propose the idea of solving optimized layout of spatial objects based on the algorithm of OMSGA, which remedies the deficiencies of the methods above.Specifically, my work includes three parts, the following details:1. The TTS entity-model database is built based on HowNet. Firstly, the standard of each VRML entity-model and the methods to standardizing the model are proposed. Secondly, the entity models are classified based on the hyponymy relationship of sememes in HowNet. Last, the TTS entity-model database is implemented.2. To automatically generate 3D scenes from spatial relationships by text input is studied. Based on the TTS entity-model database, realize the automatically generating of 3D scenes using VRML Automation techniques provided by Cortona SDK.3.Based on the research above, the optimized object-layout system is...