摘要
论文主要探讨了如何构建科学、合理的SAR图像适配性指标体系的过程。在适用于SAR图像的图像特征参数指标体系构建基础之上,首先对各个图像特征参数进行单次独立实验,验证其正确性以及对不同类型图像的区分性;再对各个参数在大量训练集的基础上进行统计实验,分析其对正负样本的区分性。提出了两种图像预处理的方法,极大地消除了训练集图像之间、训练集图像与使用集图像之间,因为成像环境、气候、飞行器飞行模式、SAR极化方式等原因造成的图像质量差异,与此同时,还使得部分特征参数获得了对图像更好的描述能力。
This paper mainly discusses how to construct a scientific and reasonable SAR image adaptation index system. First-ly,based on the construction of the index system of image characteristic parameters for SAR images,single independent experi-ments are carried out for each image feature parameter. These experiments verify the correctness and differentiation of different typesof images. Then,the statistical experiments are carried out on the basis of a large number of training sets. These experiments analyzethe distinction between positive and negative samples. Two methods of image preprocessing are proposed. The image quality differ-ences between the training set images,the training set images and the set images are greatly eliminated,such as imaging environ-ment,climate,flight modes,and SAR polarization modes. At the same time,some feature parameters are also given better imagedescription ability.
出处
《计算机与数字工程》
2017年第12期2415-2421,2464,共8页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号:61273241)资助