摘要
引入二进小波变换理论研究动态变形信号分析模型,提出动态变形信号二进小波提取与粗差识别修复的技术路线.实例选择B3-spline小波函数,采用àtrous快速二进小波变换算法的软阈值降噪方法进行动态变形信号提取,三倍中误差作为各细节尺度的降噪阈值,并在分解的细节尺度上进行粗差识别,在此基础上进行粗差修复,获取真实的变形信号.结果表明:用此模型能有效提取动态变形信号,变形信号提取的效果优于中值滤波的方法;二进小波第二分解尺度细节部分能同时定位孤立态、离散态与区域态粗差的位置,并且精度远优于Mallat算法,在区域态粗差边界的定位中表现尤为显著,这是由二进小波的平移不变性决定的.
The dyadic wavelet transform theory was introduced to study dynamic deformation analysis model in this paper. The technique process for signal extraction and gross error recognition and recovery of dynamic deformation was put forward, B3-spline wavelet function was chosed for experimental analysis to denoise the dynamic deformation signal using fast dyadic wavelet decomposition based on d trous algorithm. Soft-threshold noise reduction algorithm was applied to separate deformation trend taking triple-mean square error as the threshold of detail signals and gross error was recognized at detail scales, then the actural deformation signal is obtained after gross error recovery. The results show that the proposed model can efficiently extract deformation signal which is better than that using median filter. The isolated, dispersed and regional gross errors are discerned at the second detail scale of dyadic wavelet decomposition at one time and the positioning precision is better than that using Mallat algorithm, especially for the boundary of regional gross error, which can be explained by the shift-invariant of dyadic wavelet transform.
出处
《中国矿业大学学报》
EI
CAS
CSCD
北大核心
2007年第1期116-120,共5页
Journal of China University of Mining & Technology
基金
高等学校博士学科点专项科研基金项目(20040290503)
中国矿业大学科技基金项目(2005B020)
中国矿业大学引进人才基金项目
关键词
二进小波
à
trous算法
动态变形
粗差识别
dyadic wavelet transform
à trous algorithm
dynamic deformation
gross error detection
