Filter‐based feature ranking technique for target recognition by radar infrared combined sensors

Abstract

Feature selection plays a key role in target recognition especially for multiple combined sensors’ applications. This study presents an improved relief feature selection (Relief-F) model to rank features for multi-target recognition. Besides, the evaluation of the effectiveness of the method is conducted based on the following three measuring indices: feature redundancy, class separability and overall accuracy. The simulated data acquired by radar infrared combined sensors are used to verify the metrics of the proposed method. The experimental results demonstrate that the Relief-F method reveals the best results compared to other feature selection methods, including principal component analysis and Fisher linear discriminant analysis. It indicates that Relief-F can be an effective option for feature selection with high-class separability and a low feature redundancy rate. Furthermore, high overall accuracy can be obtained using a relatively small amount of features selected by the Relief-F method.