摘要
合成孔径图像分类,在军事和民事上具有重要意义。提出一种基于纹理特征提取与降维的合成孔径细微图像的分类算法。特征集包括图像的纹理特征。采用二维离散小波变换对其进行分析处理。最后,选择人工神经网络作为默认分类器。实验部分选择WML算法与HA算法作为对比算法,对不同的实测数据进行分类,本文提出的算法为94.94%精确度,高于WML的91.32%与HA的88.43%精确度。因此,本文算法优于WML与HA算法。
Is proposed based on the texture feature extraction and dimension reduction of synthetic aperture subtle image classification algorithm. Feature set including image texture feature. The two-dimensional discrete wavelet transform to carry on the analysis. Finally, the choice of artificial neural network classifier as the default. The part of the selection WML algorithm and HA algorithm for comparison algorithm for different measured data classification, this paper proposes the algorithm is 94.94% accuracy, higher than 91.32% of the WML with 88.43% accuracy HA. Therefore, this article algorithm is better than the WML and HA algorithm.
出处
《科技通报》
北大核心
2013年第12期100-102,共3页
Bulletin of Science and Technology
关键词
合成孔径雷达
纹理分析
图像分类
synthetic aperture radar
texture analysis
image classification
