摘要
在基于实测图像模板的SAR目标分类方法中,环境因素、成像参数等差异性造成了模板库存储量过大与完备性需求的矛盾。针对此问题提出一种基于仿真图像模板的SAR目标分类方法,通过减少模板库的存储量来降低实测数据的采集成本。该方法首先采用正则化方法对图像进行预处理来减少噪声,在此基础上提取目标峰值特征,然后利用几何哈希匹配算法实现目标的分类。实验中的仿真图像模板库采用RadBase软件生成,结果表明新方法对MSTAR实测数据进行目标分类的精确度较高,适应性较好。
The conventional SAR target classification methods are usually based on real image templates.However,the excessive storage is in contradiction with the completeness demand in the library of real image templates due to environmental factors,imaging parameters,etc.Therefore,a novel method based on simulated image templates was proposed to reduce the template library storage and economize the cost of acquiring real experimental data.At first,the images were preprocessed by regularization approach in order to suppress speckles.Then,the peak feature of the target was extracted.Finally,target classification was carried out using the geometric Hash match algorithm.Combining MSTAR image data with the simulation image templates which are generated by Radbase software,the experiment results show that the new approach is accurate and adaptable.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2013年第6期1359-1363,1375,共6页
Journal of System Simulation
基金
国家自然科学基金(60772045)
关键词
SAR图像
正则化处理
峰值提取
几何哈希
目标分类
仿真图像
SAR image
regularization
peak extraction
geometric hashing
target classification
simulated images