危险化学品立体仓库火灾预警系统采用光纤传感技术测量仓库内各位置温度值, 在处理光纤测温数据时, 传统小波去噪算法易产生人为振荡。针对此问题提出了基于改进阈值函数的分段小波去噪算法。将数据中易引发伪吉布斯现象的部分划分为准高温段采用平移不变法处理, 其他部分采用传统小波去噪算法。采用阈值法处理小波系数时, 设计了改进阈值函数以减小数据失真和数据波动。实验结果表明, 与传统小波去噪相比, 算法在不同信噪比情况下都能有效消除数据震荡, 具有良好的去噪效果。

The fire warning system of hazardous chemical warehouse uses optical fiber sensing technology to measure the temperature in different positions in the warehouse. When dealing with optical fiber temperature data, the traditional wavelet de-noising algorithm is prone to artificial oscillation. The segmented wavelet de-noising algorithm based on improved threshold function is proposed. The part of the data that can easily lead to the pseudo-Gibbs phenomenon is divided into high temperature section and use translation invariant algorithm to de-noise. The other part uses the traditional wavelet de-noising algorithm. An improved threshold function is designed to reduce data distortion and data fluctuation. Experimental results show that compared with the traditional wavelet de-noising algorithm, the algorithm can effectively eliminate data concussion in different SNR situations, and has good de-noising effect.