多时相SAR和GPS轨迹协同下的城市道路提取研究

随着我国城镇化进程的加快，城市道路随之急剧扩张，道路信息及时获取及更新对城市管理和规划具有重要意义。受限于城市场景复杂性，仅依赖遥感影像的道路提取在精度、完整性和现势性遇到了瓶颈。近年来，SAR影像和GPS轨迹的不断获取为城市道路提取提供了一个新的解决思路。因此，本项目开展GPS轨迹与多时相SAR协同下的城市道路提取研究。具体研究内容包括：1）提出集成SAR影像与GPS轨迹的纠偏模型解决轨迹偏移问题；2）提出基于GPS轨迹和多时相SAR影像的多特征集，增大道路与背景的对比度；3）提出一种基于参数不确定性的单类分类器和一种精准道路中心线提取和融合模型，解决分类参数对分类边界的影响以及道路中心线“毛刺”和断裂问题，提高道路中心线的光滑性、完整性和精度。本项目研究成果将推动多时相SAR在城市道路提取中的应用水平，为城市管理和规划部门提供服务。

With the rapid urbanization in China, the urban road network has dramatically expanded accordingly, and thus, it is critical to monitor and update road information timely for urban management and planning. Due to the complex scenario of urban, the accuracy, completeness and verisimilitude of road extraction using only satellite images has met the bottleneck. Recently, the advent development and accumulation of multi-temporal SAR imagery and GPS trajectories provide a new solution for urban road extraction. Based on this consideration, this project will conduct research on the synergy of multi-temporal SAR imagery and GPS trajectories with applications in urban road extraction. The project consists of three main parts:..1).This project will present a model based on the integration of multi-temporal SAR imagery and GPS trajectories to adjust the GPS offset caused by the positioning error...2).This project will design a multiple-feature-set to measure road information to enhance the contrast of road and background. The new descriptor will solve to some extend the limited accuracy of traditional road extraction methods using few road features...3).This project will present a parameter-uncertainty-based one class classifier (OCC) and a new centerline extraction and fusing model to overcome the influence of OCC parameters on the classification boundary, to eliminate centerline spurs and discontinuities. The results of new method will well retain the smoothness and completeness of road centerlines, which in turn improves the road extraction accuracy...The outcome of this project will accelerate the application of multi-temporal SAR in urban road extraction, and eventually provide service for the government in urban management and planning.