摘要
光纤传感器校准中,小样本数据因样本量少、分布局限且代表性不足,导致信号处理时难以精确确定小波变换的最佳尺度因子,造成去噪后信号残留干扰多、特征提取不完整。同时,小样本易引发采集偏差和信号偏移,最终影响周期原位校准的线性度拟合优度。为此,提出一种光纤传感器周期原位校准技术。利用小波变换对含电磁干扰的序列信号进行去噪,经离散化处理与三层小波分解,结合峰值信噪比确定最佳尺度因子,提取干扰抑制后的信号特征。引入子空间高斯混合模型构建偏移度量模型,采用模板匹配与归一化互相关算法实现圆心精确定位和特征区域提取,并通过直线检测获取边缘斜率特征。基于首帧参考角位移,利用旋转角度拟合公式实现多时刻信号与原位信号的精确对齐,最终实现光纤传感器周期原位校准。实验结果表明,所提方法在25℃~45℃温度范围内基本误差不超过0.1、线性度拟合优度达0.99,处理后图像的峰值信噪比均高于0.95,在极端小样本条件(3帧~5帧)下,除第4帧因复合干扰未收敛外,其余帧均稳定收敛,显著提升了小样本干扰工况下的校准精度和可靠性。
In the calibration of optical fiber sensors,it is difficult to accurately determine the optimal scale factor of wavelet transforma⁃tion during signal processing due to the small sample size,limited distribution and insufficient representativeness of small sample data,resulting in many residual interferences and incomplete feature extraction after denoising.At the same time,small samples can easily cause acquisition bias and signal shift,which ultimately affects the goodness of linearity fitting of periodic in⁃situ calibration.Therefore,a periodic in⁃situ calibration technology for fiber optic sensors is proposed.The wavelet transform is used to denoise the sequence signal containing electromagnetic interference,and the optimal scale factor is determined through discretization and three⁃layer wavelet decom⁃position,combined with the peak signal⁃to⁃noise ratio,and the signal characteristics after interference suppression are extracted.The subspace Gaussian hybrid model is introduced to construct the offset measurement model,and the template matching and normalized cross⁃correlation algorithms are used to achieve accurate positioning of the circle center and the extraction of feature areas,and the edge slope features are obtained by linear detection.Based on the reference angular displacement of the first frame,the rotation angle fitting formula is used to achieve accurate alignment between the multi⁃moment signal and the in⁃situ signal,and finally realize the periodic in⁃situ calibration of the optical fiber sensor.The experimental results show that the basic error of proposed the method is not more than 0.1 and the goodness of linearity fitting is 0.99 in the temperature range of 25℃-45℃,and the peak signal⁃to⁃noise ratio of the processed images is higher than 0.95.
作者
罗波
王献云
LUO Bo;WANG Xianyun(School of Information Engineering,Beijing Polytechnic College,Beijing 100042,China;School of Civil Engineering,North China University of Technology,Beijing 100144,China)
出处
《传感技术学报》
北大核心
2025年第12期2217-2222,共6页
Chinese Journal of Sensors and Actuators
基金
2024年北京工业职业技术学院校立科研项目(BGY2024KY-14Z)。
关键词
小样本条件
干扰工况
光纤传感器
原位校准
condition of small sample
interference condition
optical fiber sensor
in situ calibration
