基于张量低秩分解和稀疏表示的红外微小气体泄漏检测

Micro gas leakage detection based on tensor low rank decomposition and sparse representation from infrared images

为了检测石化工业生产过程中微小气体的泄漏,提出了一种应用红外成像技术的单帧红外小目标检测方法。研究了低秩稀疏分解理论和稀疏表示理论,并提出了一种新的基于张量低秩分解和稀疏表示的小目标检测方法。该方法基于张量分解的形式充分发掘背景矩阵所包含的信息;利用先验知识构造微小气体泄漏的目标字典;同时利用背景的低秩约束和小目标的稀疏表示约束分解出微小气体的泄漏目标。最后基于非精确增广拉格朗日乘子法(IALM),对本文算法进行最优化求解,并通过实验分析比较了本文方法和已有方法的优缺点。结果表明:本文方法的检测效果优于其他已有方法,并且具有较好的ROC(受试者工作特征)曲线,可以满足工业生产中对微小气体泄漏检测的要求。

To detect the micro gas leakage in petrochemical production,a single-frame small target detection method was proposed by using infrared images.The low-rank sparse decomposition theory and sparse representation theory were researched and an innovative method to detect a micro-target was proposed based on tensor low-rank decomposition and sparse representation. The tensor decomposition form was employed in exploiting the information contained in background matrices,The priori knowledge was used to construct a micro gas leakage target dictionary,meanwhile,the micro-gas leakage targets were decomposed by low-rank constraint in the background and sparse representation in the micro-target.Finally,the algorithm was solved optimally by using Inexact Augmented Lagrange Multiplier（IALM）method and its merits were compared with that of commonmethods.The results indicate that the proposed algorithm has better detection efficiency than that of common methods and it shows better ROC（Receiver Operating Characteristics）curves.It concludes that these results meet the requirements of micro gas leakage detection during industrial productions.