序列蒙特卡洛算法的数字地图线状特征跟踪模型  

A sequential Monte Carlo based linear feature tracking model for digital maps

作　　者：丛超 米曽真 CONG Chao;MI Zengzhen(School of Electrical and Electronic Engineering, Chongqing Univer sity of Technology, Chongqing 400054, China;College of Mechanical Engineering, Chongqing Uni versity of Technology, Chongqing 400054, China)

机构地区：[1]重庆理工大学电气与电子工程学院,重庆400054 [2]重庆理工大学机械工程学院,重庆400054

出　　处：《测绘科学》2018年第5期5-12,17,共9页Science of Surveying and Mapping

基　　金：重庆市科委自然科学基金一般基金项目(cstc2016jcyjA0436);重庆理工大学项目(2015XH13)

摘　　要：针对数字地图中的线状特征,如道路、河流的提取问题,该文提出一种基于序列蒙特卡洛的特征提取算法。算法利用粒子滤波方法与贝叶斯模型的建模思路,描述数字地图道路的状态及其观测估计方法;在此基础上,讨论跟踪过程中的模拟退火处理以及粒子聚类分析;最后,将跟踪模型应用在数字地图的特征提取中。实验表明,与传统的数字栅格地图道路提取算法以及基于其他粒子滤波的算法相比较,基于跟踪模型的算法具有较强的鲁棒性以及精度,同时能够很好地处理扫描栅格地图中缺失的道路信息提取。To extract road from scanned raster maps,a sequential Monte Carlo based road tracking algorithm is proposed in this paper.The algorithm describes geometrical growing feature,likelihood feature and discrete topological feature as different aspects of the state of road features.Based on which,simulated annealing and particle clustering analysis are discussed in particle tracking process.Finally,the road tracking algorithm is applied to scanned raster maps.Experiment results show that the proposed algorithm performs well compared with the traditional road extraction method and particle filter algorithm,and can handle the problem of missing information in raster maps.

关 键 词：模式识别 道路提取 序列蒙特卡洛算法 栅格地图 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP753[自动化与计算机技术—计算机科学与技术]