紀要論文

Study on Algorithms for Part Recognition Problem in the Smart Surface

eng

18810667

https://doi.org/10.15002/00009542

法政大学大学院情報科学研究科

8

43

48

2013-03

Version of Record

This study looks into the part recognition problem in the smart surface and proposes a modified Modular Principle Component Analysis (MPCA) algorithm for it. The smart surface aims at designing a microrobotics system for conveying, sorting and positioning. The part recognition is the first important step. In this research, the part recognition block is analyzed and simulated. The previous approach and algorithm by Boutoustous, et. al. PCA and MPCA are implemented for the same two sets of experiments. Aiming at a better performance in terms of differentiation rate, computing time, memory cost, a modified modified MPCA algorithm is proposed. It extracts local and global information of the parts in order to reduce the effects caused by local array distribution and global array profile, and chooses a suitable eigenspace in order to reduce memory cost and improves accuracy. Two sets of experiments, i.e. three simple models and three complicated models, are conducted. The experimental results are analyzed and comparisons are made, which indicates that the performance of the proposed approach, the modified MPCA is more effective and robust than others.

Article

AA12222297