局部对比度先验下基于低秩模型的红外小目标检测方法  

Infrared Small Target Detection Method Based on Low Rank Model with Local Contrast Prior

作　　者：何巍 安博文[1] 潘胜达[1] HE Wei;AN Bowen;PAN Shengda(College of Information Engineering,Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海海事大学信息工程学院,上海201306

出　　处：《光子学报》2021年第11期342-358,共17页Acta Photonica Sinica

基　　金：国家自然科学基金(Nos.61302132,61504078,41701523);国家重点研发计划项目(No.2017YFC1405402)。

摘　　要：为了解决红外小目标检测算法容易在复杂背景边缘和拐点处误检的问题,本文提出了一种局部对比度与非局部低秩张量模型相融合的红外小目标检测算法。首先采用双窗口结构的局部对比度算法提取目标和背景的局部先验信息。然后在所获取的局部先验信息约束下,对标准的红外块张量模型进行重新构建,并通过引入加权张量核范数最小化来进一步抑制背景和提高迭代效率。最后,将目标和背景的分离问题,转化成了一个张量鲁棒性主成分分析问题,并用交替方向乘子法实现该问题的求解。实验表明,在不同的复杂背景下,本文方法的性能均优于现有的典型红外小目标检测方法。In order to solve the problem that infrared small target detection algorithm is easy to detect falsely at the edge and inflection point of complex background,an infrared small target detection algorithm based on the fusion of local contrast and non-local low-rank tensor model is proposed in this paper.First,Double window local contrast measure algorithm is used to extract the local prior information of target and background.Then,under the constraints of local prior information obtained,the standard IPT model was reconstructed,and weighted tensor nuclear norm minimization was introduced to suppress the background and improve the iteration efficiency.Finally,the separation problem of target and background is transformed into a tensor robust principle component analysis problem,and alternating direction method of multipliers is used to solve this problem.Experimental results show that the performance of the proposed method is better than the existing typical infrared small target detection methods under different complex backgrounds.

关 键 词：红外小目标检测 加权张量核范数最小化 双窗口局部对比度算法 张量鲁棒性主成分分析 交替方向乘子法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]