基于实景显著性的室内外行人导航模型研究

The spatial saliency feature of real-scene environment is crucial to the reduction of pedestrian navigation cognition load, the improving of pedestrian navigation efficiency, and the avoiding of getting lost.However,current GIS navigation model cannot model saliency feature for guiding pedestrians with less cognition burden. Different from these traditional navigation approaches, this project makes full use of saliency feature in real-scene environment to address at the problem. Two key research questions, like the spatiotemporal modeling of real-scene saliency in pedestrian navigation environments and its derived pedestrian navigation data model, the multiobjective optimization of navigation path calculation and augmentation by using saliency feature to reduce recognition load, will be studied. This project will define and attract the real-scene saliency feature in pedestrian navigation environment, and then propose a spatiotemporal data model for pedestrian navigation which is based on the fusion of real-scene image and pedestrian walking network.After that,this project will propose the multiobjective optimization approaches of calcuating pedestrian navigation path and augmenting guidance information on real-scene image according to the proposed data model. Finally, this project will implement the proposed data model and their algorithms to carry out an prototype of pedestrian navigation application for both outdoor and indoor environments, and will evaluate the efficiency and adaption of pedestrian navigation prototype in different typical environments. As result, this project will fill in the blanks of GIS capabilities in modeling, representing,analyzing and making full use of the saliency cognition knowledges for navigation theory and location-based services. It will create an innovation of pedestrian navigation data model which integrates real-scene saliency feature, which could improve the adaption abilities of pedestrian navigation applications for different complex pedestrian navigation environments,and provide some solidate foundations of theory and method for real-time navigation servcies. This study can facilitate navigation application of public life popularity, improve the identity of discipline of geography for more people.

实景环境的显著性空间特征,对降低行人导航认知负担、提高导航效率、避免迷路等起关键作用。然而目前GIS导航研究难以支撑实景空间显著性等认知规律的科学建模与分析利用。本项目研究新思路是以显著性认知规律支撑行人导航理论,针对行人导航中实景显著性认知规律的时空建模与融合、顾及实景显著性的行人导航路径计算与增强表达多目标快速优化等关键科学问题,具体研究:实景要素的视觉显著性特征描述与提取,实景显著性特征时空融合的室内外一体行人导航数据模型,实景导航路径计算与增强表达的多目标快速优化,行人导航实验及三种环境的适应性分析。预期突破目前GIS导航对实景空间显著性认知规律科学建模与分析利用的瓶颈科学难题,实现基于实景显著性的室内外行人导航模型理论与方法原始创新。为提高行人导航的复杂环境适应能力、构建GIS与认知学科交叉的低认知负担行人导航理论等做科学贡献，促进行人导航技术变革与普及，提高地理学科大众认同感。

本课题围绕行人导航中实景显著性认知规律的时空建模与融合、顾及实景显著性的导航路径计算与增强表达多目标快速优化两个关键科学问题展开了研究，四年来的主要研究内容归纳为如下四个方面：（1）采集实景影像、路网等数据，对行人导航场景中的实景要素的视觉显著性特征描述与提取；（2）建立实景显著性特征时空融合的室内外一体化行人导航数据模型，构建基于实景显著性的行人导航理论；（3）研究实景导航路径计算与增强表达的多目标快速优化；（4）在理论研究基础上，构建实验原型系统开展行人导航实验分析与评估。.课题基本按项目预定的计划进行，达到了项目预期的目标，完成了项目预期的研究成果。基于以上研究内容，形成了理论结合实践的比较系统的研究成果。（1）理论成果：相关研究成果整理论文发表35篇发表在国内外期刊和著作上。其中SCI索引论文16篇，EI索引论文7篇。（2）实验原型系统及专利：基于实景显著性特征时空融合的行人导航数据模型，在GIS空间数据库中进行实现，构建了基于iOS平台的行人导航应用，申请获得一项软件著作权；对利用离散全景图构建连续场景的图像投影方法研究，申请了一项国家发明专利。（3）学术交流：与美国田纳西大学、深圳大学、南京师范大学、河南理工大学、武汉市交通发展战略研究院、四维图新等进行学术交流与技术合作等，多次主持并参与地理信息科学理论与方法年会导航与位置服务专场。（4）人才培养：指导了2名博士生、7名硕士生、2名本科生顺利毕业。