基于纹理与特征选择的前视红外目标识别

Forward-looking infrared target recognition based on texture and feature selection

针对前视红外(FLIR)目标自动目标识别(ATR)问题,提出了一种基于纹理特征的ATR方法。不同于传统基于学习、基于模板以及基于稀疏表示的方法,从图像灰度入手,提出采用局部三值模式(LTP)描述图像纹理特征,同时结合FLIR图像的特点,对LTP进行了增强处理;然后针对特征的高维问题,采用特征选择方法进行降维处理;最后采用降维后的特征实现ATR。实验结果表明,本文方法取得了比传统方法更好的效果;同时也证明,仅从纹理分析入手,也能较好地实现前视红外目标的ATR。

Unlike the traditional learning-based, template-based and sparse-based approaches, this paper presents a novel feature extraction algorithm based on texture and feature selection for automatic target recognition （ATR） in infrared imagery. Firstly, combining with the characteristics of the forward-loo- king-infrared （FLIR） images, we introduce a new concave-convex local ternary pattern （CCLTP） opera- tor by incorporating global intensity information, which divides the local features LTP into two distinct groups,namely,convex LTP and concave LTP. After that, different feature selection methods are dis- cussed and tested to reduce the dimensionality of the features. Finally, the reduced feature is used for for- ward-.looking infrared target recognition. Experimental results demonstrate that the proposed method can achieve competitive results （at lower computational complexity） compared with the state-of-the-art methods.