论文部分内容阅读
随着布里渊光时域分析(BOTDA)传感技术在许多大型基础工程设施安全监测中的广泛应用,对测量精度和实时性的要求日益提高。采用传统的最小二乘曲线拟合方式对布里渊散射谱进行布里渊频移提取,其测量结果的精度依赖于参数初始值的选取和噪声的影响,并且拟合算法的参数迭代求解过程增加了数据处理的时间,降低了工程实时性。文章综述了多种非线性参数优化估计的曲线拟合算法和基于神经网络的布里渊散射谱特征提取的混合优化算法,介绍了无需经过曲线拟合的互相关法(XCM)、深度学习法(DL)和亚像素级精度的重心提取算法(CDA),这些算法能适应更大的扫频步长,实时性更好。
With the widespread application of Brillouin Optical Time Domain Analysis (BOTDA) sensing technology in the safety monitoring of many large-scale infrastructure projects, the requirements for measurement accuracy and real-time performance are increasing. Brillouin frequency shift extraction is performed on the Brillouin scattering spectrum using the traditional least-squares curve fitting method. The accuracy of the measurement results depends on the selection of the initial value of the parameter and the influence of noise, and the parameter iterative solution process of the fitting algorithm Increased data processing time, reducing the project real-time. In this paper, we present a hybrid optimization algorithm based on curve fitting of many nonlinear parameters and neural network based on Brillouin scattering spectral feature extraction. Cross-correlation (XCM) without curve fitting, depth learning (DL) and sub-pixel-level centroid extraction algorithms (CDA). These algorithms can adapt to larger sweep steps and have better real-time performance.