基于粒子群优化的SAR图像舰船目标特征选择

SAR图像特征提取和分类器设计是进行目标识别的关键,通常情况下分类器性能与特征数量之间并不存在线性关系,相反过度冗余的特征甚至会导致分类器性能严重下降,因此特征选择成为必要。提出一种粒子群优化与Wrapper策略相结合的特征选择方法,针对包含待识别舰船目标的SAR图像,提取其3类共16个典型特征,利用本文所提算法筛选出最佳的特征组合。实验结果表明,将本文所提取的特征组合用于目标识别,分类精度提高了22%,分类时间缩短了2.16 s。

The feature selection of SAR image ship target based on PSO

SAR image feature extraction and classifier design are the key to target recognition.Generally,there is no linear relationship between classifier performance and feature quantity.On the contrary,over-redundant features may even lead to serious degradation of classifier performance.Feature choice becomes necessary.A feature selection method combining particle swarm optimization and Wrapper strategy is proposed.For the SAR image containing the ship target to be identified,16 typical features of 3 categories are extracted,and the best combination of features is chosen out using the proposed algorithm.The experimental results show that the feature combination extracted in this paper is used for target recognition,the classification accuracy is improved by 22%,and the classification time reduced by 2.16 seconds.