增强现实系统中电磁跟踪器注册精度校正研究

RESEARCH ON CORRECTION OF REGESTRATION ACCURACY FOR ELECTROMAGNETIC TRACKER IN AUGMENTED REALITY SYSTEM

针对电磁跟踪器磁场空间数据之间的复杂的模糊关系,提出基于T-S模糊系统的BP神经网络和最小二乘支持向量机相融合的方法对电磁跟踪器的注册精度进行校正。方法首先采用K-means对空间数据进行聚类分析,随后在局部上采用T-S模糊系统进行预处理,再从全局上利用BP神经网络进行训练,根据最终校正精度动态调整BP神经网络的训练目标,初步校正后再采用最小二乘支持向量机进行求解。实验结果表明,该方法适用于非线性空间数据校正,能有效提高电磁跟踪器的注册精度,有助于提高增强现实系统的交互精度。

Aiming at the relationship among the spatial data of the electromagnetic tracker is complex and fuzzy in magnetic field,the fusion method which combined of BP neural network and least squares support vector machine based on T-S fuzzy system is proposed. This method corrects the registration accuracy of the electromagnetic trackers. First,kmeans algorithm was employed in the cluster analysis,and then T-S fuzzy system was used to pre-process the spatial data in local. Second,training was used by BP neural network in global and the training target of BP neural network was dynamically adjusted by the final correction accuracy. At last,least squares support vector machine was used for solving after preliminary correction. Our results indicate that this method is applicable to nonlinear spatial data correction,which can effectively improve the registration accuracy of electromagnetic tracker and to help improve the accuracy of an interactive system with augmented reality.