摘要
方位角估计是进行目标识别的重要步骤.本文首先介绍了一种基于CFAR的目标图像分割算法.并讨论了目标区域的抽取问题,然后提出了一种快速有效的方位角估计算法.该算法首先线性拟合目标的距离向主导边界,若直线倾角落在近水平的某个范围内,则拟合方位向主导边界.从拟合直线的倾角可以得到目标方位角.该算法能够检测并去掉主导边界上的毛刺.对分割误差具有一定的鲁棒性.用MSTAR公开数本库中的10类目标进行实验,结果表明该算法具有较高的估计精度.
Azimuth estimation is very important for target recognition. Frist, the paper presents a segmentation algorithm based on CFAR technique, and the target region extraction problem is discussed. Then we propose an efficient azimuth estimation algorithm. A linear fit is first performed on the major edge of the range direction. If the obliquity of the fitted line is near horizontal, the linear fit performs on the major edge of the azimuth direction. The target azimuth angle can be obtained from the obliquity of the fitted line. The algorithm can detect and get rid of burrs on the major edge, and is robust to some segmentation errors. By using the MSTAR public data set experimental result shows the efficiency of this algorithm.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2004年第4期462-466,共5页
Pattern Recognition and Artificial Intelligence