长波红外高光谱成像气体鉴别方法研究

Research on Gas Identification Method of Long Wave Infrared Imaging

红外高光谱成像技术能准确地估计和显示目标气体在空间中的分布情况和扩散态势,传统的方法难以直接扣除背景辐射和干扰物信号,制约了目标信号的特征提取和识别分类。针对红外高光谱技术探测气体的问题,基于红外光谱探测技术的分析,重点选择了时空混合调制高光谱的技术途径。研究了红外光谱成像气体鉴别算法并行处理算法。基于GPU平台,设计了并行处理算法,解决了准实时的海量数据处理难题,通过实物样机进行室内外验证试验,实现了外场准实时气体鉴别。整个并行算法架构不仅可适用于时空混合调制型红外成像光谱系统,也可适用于时间调制型红外成像光谱系统、空间调制型红外成像光谱系统。

Infrared hyperspectral imaging technology can accurately estimate and display the distribution and diffusion situation of target gas in space.The traditional methods are difficult to directly deduct the background radiation and interference signals,which restricts the feature extraction,recognition and classification of target signals.Aiming at the problem of detecting gas with infrared hyperspectral technology,based on the analysis of infrared hyperspectral detection technology,the technical way of spatiotemporal mixed modulation hyperspectral is selected.This paper studies the parallel processing algorithm of gas identification algorithm in infrared spectral imaging.Based on the GPU platform,aparallel processing algorithm is designed to solve the problem of quasi real-time massive data processing.Through the indoor and outdoor verification test of the physical prototype,the quasi real-time gas identification in the field is realized.The whole parallel algorithm architecture can be applied not only to spatiotemporal hybrid modulation infrared imaging spectrum system,but also to time modulation infrared imaging spectrum system and space modulation infrared imaging spectrum system.