摘要
本文提出了本地图像块模型,该模型使用概率隐含语义分析pLSA(probabilistic Latent Semantic Analysis)作为数据项的点过程对人造目标进行建模,用可逆马尔科夫链蒙特卡洛算法(RJMCMC)配合模拟退火算法提取模型的参数。在提取模型参数的过程中使用了数据驱动生灭和统计扩散方法加快算法收敛速度。最后通过对实际遥感图像的实验验证了该方法的有效性。
This paper proposes local patches model, and the probabilistic Latent Semantic Analysis is employed to model the man made object as data term of marked point process in this model, the parameters of the model are optimized by Reversible Jump Markov Chain Monte Carlo （RJMCMC） and Simulate annealing. Data-driven Birth-and- Death and Stochastic Diffusions are used to reduce the complexity of RJMCMC kernel and accelerate the speed of convergence. The method is verified on remote sensing image.
出处
《电子测量技术》
2008年第9期117-120,共4页
Electronic Measurement Technology
关键词
高分辨率遥感图像
人造目标提取
概率隐含语义分析
可逆马尔科夫链蒙特卡洛
数据驱动的生灭
统计扩散
high resolution remote sensing image
extracting man made objects
probabilistic Latent SemanticAnalysis
Reversible Jump Markov Chain Monte Carlo sampler （RJMCMC）
data-driven birth and death
stochastic diffusions
