摘要
遥感图像在森林管理中有重要的作用,随着数据量的增加和分辨率的提高,从图像中提取树冠参数成为需要和可能。该文根据树冠的特征,使用标记点过程对树冠建模,采用可逆马尔科夫链蒙特卡罗算法(MCMC)配合模拟退火算法提取树冠的参数,提出新数据项以使其更好适应图像;提出数据驱动的生灭核,同时提出使用随机扩散方法代替非跳转转移核加快算法收敛速度并简化了该方法的实现。最后通过对遥感图像的实验验证了该方法的有效性。
Remote sensing imagery plays a key role in forestry management. The increasing availability of data and their high spatial resolution make the tree crown extraction necessary and possible. Marked point processes are employed to model the tree crowns, based on the characteristics of them. The parameters of the model are optimized by RJMCMC （Reversible Jump Markov Chain Monte Carlo sampler） and Simulate annealing. New data term is proposed to give better description of the local pattern of the tree crown; Data-driven Birth-and-Death and Stochastic Diffusions are introduced to reduce the complexity of RJMCMC kernel and accelerate convergence speed. The method is verified on remote sensing image.
出处
《电子与信息学报》
EI
CSCD
北大核心
2008年第9期2103-2106,共4页
Journal of Electronics & Information Technology
关键词
图像处理
树冠提取
数据驱动的生灭
随机扩散
Image processing
Tree crown extraction
Data-driven birth and death
Stochastic diffusions
