【摘要】 提出一种光强-波长模型和径向基函数网络（radial basis function network,RBFN）相融合的光谱共焦信号峰值提取算法，简称RBFN-I-λ。首先通过高斯拟合法拟合离散光谱响应信号的差分信号粗略得到初始峰值波长，然后基于泰勒近似法得到理想峰值波长并计算初始峰值波长和理想峰值波长之间的波长差，最后利用RBFN-I-λ建立光谱共焦响应信号与波长描述误差之间的映射关系。实验结果表明，RBFN-I-λ算法的精度与传统抛物线法、质心法和高斯拟合法等方法相比，至少提升30%。更多还原

【Abstract】 A peak extraction algorithm for spectral confocal signals with the integration of light intensitywavelength model and radial basis function network(radial basis function network, RBFN) is proposed,referred to as RBFN-I-λ. Firstly, the initial peak wavelength is roughly obtained by fitting the difference signals of the discrete spectral response signals through the Gaussian fitting method, then the ideal peak wavelength is obtained based on the Taylor approximation method and the wavelength difference between the initial peak wavelength and the ideal peak wavelength is calculated, and finally the mapping relationship between the spectral confocal response signals and the wavelength description error is established by using RBFN-I-λ.Finally, the mapping relationship between the spectral confocal response signal and the wavelength description error is established using RBFN-I-λ. The experimental results show that the accuracy of the RBFN-I-λalgorithm is at least 30% higher than that of the traditional parabolic, center-of-mass and Gaussian fitting methods.更多还原