摘要
利用最大后验概率-马尔可夫随机场(maximum posteriori-Markov random field,MAP-MRF)模型,从合成孔径雷达图像中提取道路是一种全局最优方法。该方法克服了SAR图像固有的相干斑干扰,结合道路的空间分布特征,构建了马尔可夫随机场关于道路分布的先验概率模型,并利用遗传算法获得后验概率最大意义下的最优结果。通过对真实SAR图像的试验表明,在主干道路被提取的同时,一些人眼难于辨识的低对比度细小的道路也能被辨识出来,从而说明了该方法对于SAR图像道路提取的有效性。
A road extraction method in EAR image based on MAP-MRF （maximum posteriori-Markov random field） frame is proposed, which performs good result in spite of speckle. The method models the prior probability distribution in combination with space distribution of roads. A genetic algorithm is applied to generate optimal results in terms of maximal posterior probability. The proposed method is testified on ERS-1 SAR image. Result shows that not only main roads＇are extracted, roads almost submerged by speckle are also detected, which indicates the method has a low minimum detectable contrast.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2008年第10期2028-2030,共3页
Systems Engineering and Electronics
