基于国产GF-1遥感影像的面向对象桥梁提取方法研究  

Object-Oriented Extracting Bridges Information Based on China-Made GF-1

作　　者：袁泽[1,2] 丁建丽[1,2] 王娇[1,2] 陈文倩[1,2] 李相[1,2] 黄帅[1,2] 

机构地区：[1]新疆大学资源与环境科学学院,新疆乌鲁木齐830046 [2]绿洲生态教育部重点实验室,新疆乌鲁木齐830046

出　　处：《传感技术学报》2015年第5期690-696,共7页Chinese Journal of Sensors and Actuators

基　　金：高分辨率对地观测重大专项项目(民用部分)(95-Y40B02-9001-13/15-03-01);国家自然科学基金项目(U1303381;41261090;41130531;41161063);教育部新世纪优秀人才支持计划项目(NCET-12-1075);教育部长江学者计划创新团队计划项目(IRT1180);新疆维吾尔自治区青年科技创新人才培养工程项目(2013711014);霍英东青年教师基金项目(121018)

摘　　要：以国产"高分一号"2m/8m高空间分辨率遥感图像为数据源,使用基于规则的面向对象的方法实现了对高分辨率影像中桥梁目标的精确提取。首先,经过多尺度分割实验并结合下垫面特征选择最优分割尺度;其次,利用水体指数、阈值函数等方法建立规则集,逐步获取水体和桥梁潜在区的矢量文件;最后,通过二值化、数学形态学处理、叠加分析等方法成功提取桥梁目标。实验结果表明,该方法可以准确、高效地提取出桥梁信息,总体精度在90%以上,Kappa系数在85%以上,在高分辨率遥感影像高精度提取桥梁的实践中具有广泛的适用性,该研究成果或对国产高分影像处理系统的研究与应用提供了一定的科学参考。Based on China-made GF-1 ,high spatial resolution remote sensing image as data sources ,using object-o- riented method based on rules to achieve the goal of high spatial resolution image in the bridge of the extraction. First, through a large number of multi-scale segmentation experiments, the optimal segmentation scale according to the underlying surface feature selection;Secondly, useing of water body index, the threshold function method to establish rules, such as gradually access to water and the potential bridges vector files;Then, by using binarization processing, mathematical morphology algorithm, overlay analysis and so on operation, The ultimate success of extraction bridge targets. Results that the method can accurately and efficiently extract the information of bridges, the overall accuracy of over 90% ,the Kappa coefficient of more than 85%, the results of the study or to the research and appli- cation of domestic high image processing system to provide some scientific reference.

关 键 词：高分一号 桥梁 最优分割尺度 面向对象 数学形态学 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]