摘要
利用Creator建立目标模型,Vega的TMM工具进行纹理材质映射,基于视景仿真技术建立了多波段多极化SAR图像数据库。设计了融合遗传算法和二值粒子群的混合智能优化算法,对SAR图像的波段极化组合方式进行优化;基于未矫正和矫正后的图像分别提取Zernike矩、Gabor小波系数等构成候选特征序列,进行了多波段多极化SAR图像特征选择实验。实验结果表明,采用仿真技术建立SAR图像数据库是进行多波段多极化SAR图像识别的一种有效手段;采用优化后的特征集合能够提高多波段多极化SAR图像的识别率。
The object model was built based on Creator, and object texture-material mapping was performed by Vega TMM tool The multi-band and multi-polarization SAR image database was built by visual simulation technology. A hybrid intelligent optimization algorithm was designed to optimize combination of band and polarization by genetic algorithm and binary particle optimization. Zernike moment features, Gabor wavelet coefficients, etc were extracted from original image and rectified image to make up of feature candidates, and the feature selection experiments were carried out by using multi-band and multi-polarization SAR images. Simulation results demonstrate that, building SAR image database through simulation technology is an effective method to perform research on multi-band and multi-polarization SAR image recognition; target recognition rate can be improved for multi-band and multi-polarization SAR images using the optimized feature set.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2017年第10期2482-2488,共7页
Journal of System Simulation
基金
国家自然科学基金(61174024
61372024)
浙江省自然科学基金(LQ13F050010)
关键词
SAR
多波段多极化
计算机仿真
特征选择
SAR
multi-band and multi-polarization
computer simulation
feature selection
