基于加权RPCA的单帧红外图像小目标检测

Single-Frame Infrared Image Small Target Detection Based on Weighted RPCA

红外图像目标检测作为红外探测系统的关键技术。本文提出一种改进的基于加权鲁棒性主成分分析的单帧红外小目标图像检测算法。该方法首先建立红外图像的小目标模型,其中包含目标、背景和噪声及杂波;其次将红外图像构造成块图像,转换为RPCA问题;采用加权鲁棒性主成分分析算法进行求解,将目标矩阵和背景矩阵进行分离,并通过图像重构得到目标图像和背景图像。使用包含天空、海洋和沙漠不同背景下的单帧红外小目标数据集验证了本文提出算法的有效性。

Infrared image target detection is the key technology of infrared detection system. This paper proposes an improved single-frame infrared small target image detection algorithm based on weighted robust principal component analysis. Firstly, this method establishes a small target model of the infrared image, which contains the target, background, noise and clutter;secondly, the infrared image is constructed into a block image and converted to the RPCA problem;finally, the weighted robust principal component analysis algorithm is used to solve the problem, and the target matrix and the background matrix are separated, and the target image and the background image are obtained through image reconstruction. Using single-frame infrared small target datasets under different backgrounds of sky, ocean and desert verifies the effectiveness of the proposed algorithm.