Study on Seamless Target Modeling System for Three Dimensional Vision System

点击翻译按钮获取中文摘要

In this study, we developed some techniques for building a seamless object modeling system using three-dimensional vision that does not require an off-line operation for modeling and can construct a model for the operation target of the robot and a part of the robot itself during on-line operation.Firstly, as a basic technique for finding the same data in the multiple three-dimensional data, we developed a method to find the same data in multiple two-dimensional data. Secondly, we developed a new feature descriptor for three-dimensional object recognition system and constructed an object recognition system based on the feature descriptors. Finally, we developed a prototype of life support robot system equipped with a three-dimensional vision system that can accomplish a pick-and-place task in terms of everyday items.