论文部分内容阅读
合成孔径雷达(SAR)图像自动目标识别是图像识别领域的一个重要方向。受视觉细胞感受野模型的启发,该文提出了一种从图像局部点出发,对图像进行多分辨分解的图像处理方法。采用一组简单的八邻域正交基对图像进行多级滤波采样处理,得到原图像的多级类Gauss差分图像尺度空间,并将其应用到MSTAR数据集中的SAR图像目标的特征提取;同时,基于多级特征的整合思想,运用基于多尺度核方法的SVM模型,对不同级别图像特征采用不同尺度的核函数分别映射,然后进行合成,实现多类目标的分类。对MSTAR数据集的实验结果表明,该方法具有很高的正确率,并且实现简单快速。此外,该方法还可方便地应用于SAR图像场景中多类、多个目标的分割与自动目标识别,并且对相干斑噪声具有较强的鲁棒性。
Synthetic aperture radar (SAR) image automatic target recognition is an important direction in the field of image recognition. Inspired by the visual cell receptive field model, this paper presents an image processing method for multiresolution decomposition of an image starting from the local point of the image. A set of simple orthonormal basis of eight neighborhoods is used to perform multistage filtering and sampling processing to obtain the multiscale Gaussian difference image scale space of the original image and to apply it to the feature extraction of SAR image targets in the MSTAR dataset. Based on the idea of multi-level integration, the SVM model based on multi-scale kernel method is used to map different levels of image features using different scales of kernel functions, respectively, and then synthesized to achieve the classification of multiple targets. Experimental results on MSTAR datasets show that this method has high accuracy and is simple and fast. In addition, this method can also be conveniently applied to the segmentation and automatic target recognition of multiple classes and multiple targets in SAR image scene, and has robustness to speckle noises.