焊接诊断中瞬态光谱单次编码可靠降噪机理研究

High signal-to-noise ratio (SNR) and fast measurement speed have always been the goal of spectral measurement technology, and it is the inevitable requests of spectral measurement in welding diagnosis. Neither computational spectral measurement methods based on compressed sensing, deconvolution, computational slit nor traditional multiplexing methods based on Hadamard transform, interference modulating Fourier transform can balance transient measurement and stable denoising. Thus these measurement methods are not suitable for welding diagnosis. This project plan to explore a new single shot code aperture technology innovatively for stable and reliable denoising measurement based on the achievements of coded measurement. The essential idea of the project is to realize snapshot, stable and reliable measurement through the robustness of Hadamard encoding and positive definiteness of encoding matrix. The main research contents are included as below: (1) Research encoding measurement model based on the physical process of wide-slit deconvolution and Hadamard encoding measurement. (2) Research the equivalence of metastable encoding error and measurement noise. (3) Research the stableness of denoising capacity and the upper and lower bounds of denoising capacity, and verify the results through experiment. The implementation and achievement of this project will rich encoding measurement denoising theory, and provide a new theoretical and technical support for high precision snapshot spectral measurement in complex industrial scenarios such as welding spectral diagnosis.

高精度与快速度一直是光谱测量技术追求的目标也是焊接诊断对光谱数据的必然要求。现有以压缩感知、逆卷积、计算狭缝为代表的计算型测量方法和以哈达玛变换、干涉调制傅里叶变换为代表的多路复用型测量方法均难以兼顾瞬态测量和稳定降噪，均不适用于实时焊接诊断。针对这一缺陷，本项目拟基于编码测量研究成果，创新地探索一种基于单次编码的稳定、可靠降噪技术，其核心思想是结合统一的编码测量数学模型与编码误差、测量噪声之间的等效关系，利用哈达玛编码降噪鲁棒性与编码矩阵的正定性，实现瞬态、稳定、可靠的测量。本项目主要研究内容包括：（1）宽狭缝逆卷积与哈达玛编码物理过程与数学建模研究；（2）亚稳定编码误差与测量噪声的等效性研究；（3）瞬态编码降噪稳定性、性能上下界以及实验研究。本项目研究及其成果可丰富编码测量降噪理论的内涵，为基于光谱的焊接诊断等复杂工业场景下的高精度瞬态光谱测量提供一种全新的方法和技术支撑。

光谱作为物质的指纹信息，广泛应用于工业监测、物质鉴别、科学研究等领域。为了满足高精度、高灵敏、高速度的光谱测量需求，本项目研制了一种适用于复杂场景下的高精度瞬态光谱测方法。在深入分析现有狭缝光谱测量、编码光谱测量优缺点与测量物理过程的基础上，结合光栅色散的线性特性，建立了统一的编码光谱测量数学模型和基于编码矩阵特征值分布的编码光谱测量信噪比提升分析方法。以此为基础，研制了一种基于单次编码的稳定、可靠降噪技术，通过在传统哈达玛变换光谱仪中专门设计用于测量场景光强分布的光路，同时测量场景光强分布与色散叠加后的光谱，然后通过可逆重建得到待测光谱，从而实现瞬态光谱的测量。仿真与实验结果均表明该瞬态哈达玛变换光谱仪拥有与传统哈达玛变换光谱仪类似的性能，相比于传统光谱测量方法，信噪比提升10dB以上，且只需要单次成像即可得到光谱分辨率优于1nm的电弧光谱数据。本项目研究及其成果可丰富编码测量降噪理论的内涵，为基于高灵敏、高精度、高速度光谱的工业监测、科学研究等，提供了一种全新的光谱测量方法与技术支撑。