小波多尺度分解及其在SAR图象目标检测中的应用

Multiscale Wavelet Decomposition and Application in Object Detection of SAR Images

合成孔径雷达图象的目标检测是最重要的任务之一.目前用于目标检测的现有方法速度较慢、准确率较低、精度较差.作者提出小波分解的方法用于目标检测.同尺度目标检测采用二维离散二进小波分解方法,提高检测速度.不同尺度的目标检测采用多尺度二维离散小波分解方法,提高检测准确率.图象匹配采用灰度归一化的灰度相关匹配方法,提高匹配精度.实验证明,采用该法进行目标检测能获得较好的效果.

Synthetic aperture radar （SAR） imaging is an absolutely necessary means in modern mapping and military spying, Object detection in SAR images is one of the most important tasks. Nowadays, most of the existing methods used for object detections are of slower computation, less correctness, and lower accuracy. A method using wavelet transform is proposed for object detection. For object detection in same scale, two-dimensional discrete bit wavelet decomposition is applied to speed up computation. In different scales, multiscale two-dimensional discrete wavelet transform is utilized to increase correctness. For image matching, intensity generalization and intensity correlation are used to increase accuracy. Experiments have shown that the object detection using the method proposed here can achieve the results as good as expected.