摘要
针对道路提取方法存在的漏提、误提等问题,提出一种基于局部结构约束的线段标识点过程提取算法。将道路网视作道路段集合,以随机点过程定义位置,以线段为标识定义几何结构,根据道路局部特征构建自适应的约束模型,结合道路的光谱同质性和异质性建立光谱测度模型,在贝叶斯理论构架下,综合上述模型和参数分布建立后验概率模型,设计可逆跳变马尔可夫链蒙特卡洛(RJMCMC)算法模拟采样,并为加快收敛速率设计高效的RJMCMC转移核,以最大化后验概率为准则,获取最优道路网。采用全色遥感影像进行实验、缓冲区评价方法进行定性和定量分析,计算得出道路提取的精准率和完整率分别能达到52%和98%以上,验证所提算法能准确有效地提取出道路网络。
The existing extraction methods of road extraction have problems of mission, errors, etc. To address these issues, this paper presents a new algorithm based on the line marked point process with local structure constraints. The road network is regarded as a collection of road segments. The random point process defines their locations and the line marks associated to the points indicate their geometric structures. Accordingly, an adaptive structure constrained model is established with respect to the local features of road network. The spectral measurement model is formulated by combining the spectral homogeneity and heterogeneity of road segments. By using the Bayesian theory, the posterior probability of the road extraction model is achieved by integrating the aforementioned models and correlative parameters distributions. The reversible jump Markov chain Monte Carlo(RJMCMC) algorithm is designed to simulate the posterior probability, and efficient transfer kernels are designed to accelerate the convergence rate. Then, the optimal road network is realized under the maximum a posterior(MAP) criterion. Panchromatic remote sensing images are utilized to carry out experiments and results are qualitatively and quantitatively analyzed by the buffering assessing method. The precision and completeness of road extraction are both larger than 52% and 98%, respectively. It demonstrates that the proposed algorithm can realize the road extraction accurately and efficiently.
作者
赵泉华
吴优
张洪云
李玉
Zhao Quanhua;Wu You;Zhang Hongyun;Li Yu(School of Geomatics,Liaoning Technical University,Fuxin 123000,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2020年第7期185-195,共11页
Chinese Journal of Scientific Instrument
基金
辽宁省高等学校创新人才支持计划项目(LR2016061)资助
关键词
遥感影像
道路提取
线段标识点过程
可逆跳变马尔可夫链蒙特卡洛
转移核
remote sensing images
road extraction
line marked point process
reversible jump Markov chain Monte Carlo
transfer kernels
