摘要
提出了利用BP神经网络对跟踪器进行校正。针对神经网络训练速度慢、容易陷入局部极值的情况,首先利用具有良好全局搜索能力的遗传算法来优化BP神经网络的各层初始权值和阈值,为后续神经网络的搜索定位出一个优化的搜索空间。实验结果表明,利用该遗传神经网络方法进行跟踪器校正,能够显著提高增强现实系统的精度,有助于提高增强现实系统的真实感。
An approach using BP neural network is proposed to rectify the error. Considering that the neural network is prone to get local extremum and its convergence speed is slow, it utilizes the excellent global searching ability of genetic algorithm to optimize the initial weights and threshold of neural network. The results of experiment show that this method can not only improve the convergence precision of weights but also insure the neural network to get global convergence fleetly. After the tracker is rectified with the above method, the precision of AR system is improved prominently, consequently to enhance the third dimension of AR system.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2006年第10期28-30,共3页
Computer Engineering
基金
上海市自然科学基金资助项目(025115008)
关键词
BP神经网络
遗传算法
增强现史
磁力跟踪器
校正
BP neural network
Genetic algorithm
Augmented reality
Magnetic force tracker
Rectification