齿轮测量的粒子滤波方法

Measurement of Gear Based on Particle Filter

提出一种基于粒子滤波的齿轮检测新方法。提出了核概率区域抽样方法,避免了抽样不全面和样本特征不全面的问题;通过对每个样本(粒子)赋予相应的权值来表示样本准确度以表征所测参数的分布,进而由分布求得具体测量值。该方法不受测量数据实际分布的限制,适用于非线性、非平稳、非高斯分布情形的测量。实验结果表明,所提方法有效而稳健,可应用于不同形状、不同大小的齿轮的测量。

This paper presented a particle filtering algorithm for measurement of gears. First, a sampling method based on nucleus probability function was proposed to avoid sampling or characteristics of samples deficientiy. For scaling nicety of samples each sample was weighted, so as to take distribution of measured parameters, which can be obtained from the distribution. The proposed method is not restricted to any contribution. It can effectively deal with non--linear, non--placidity and non--Gaussian problem. Experimental results show that the proposed algorithm is effective and robust for measurement of gears, and can be applied to various gears with different sizes and shapes.