面向地理位置感知的移动服务技术方法研究

发布时间：2018-03-25 10:41

本文选题：位置感知服务　切入点：地理信息检索　出处：《浙江大学》2017年硕士论文

【摘要】：近年来,随着移动互联网技术和定位技术的快速发展,百度地图导航、大众点评等基于地理位置感知服务的应用为人们的生活带来了极大的便利。目前地理位置与用户需求之间存在着"知识鸿沟"的问题,缺乏对于空间位置关系和语义关系的查询支持,阻碍了基于地理位置感知服务的进一步发展。本文针对面向地理位置感知的移动服务技术进行了研究,将语义理解、图像识别与具体应用服务相结合,力求解决以下两个方面的技术问题:(1)常规的地理信息查询服务不能满足特定情境下用户的需求,如何在该情境下为用户提供符合语义的地理空间位置查询服务成为亟待解决的问题。本体自身具有良好概念层次结构的特点,能对逻辑推理提供较好的支持。本文提出了一个基于本体的地址推理和POI(Point of Interest,兴趣点)查询模型。该模型通过特定情境下本体的构建以及神经网络语言模型的辅助,为用户对其所处环境资源的空间关系查询和语义关系查询提供支持。(2)针对图像识别中空间金字塔匹配模型缺乏对视觉物体旋转、平移和缩放的考虑的问题,顾及移动端设备的性能瓶颈,本文提出了一种基于特征点群相似度计算模型的图像表示方法。通过引入特征点群相似度计算模型对金字塔匹配模型进行优化,得到图像抗旋转、抗平移和抗缩放的表示,可以有效提高移动客户端标志物图像识别的准确率。最后,本文设计了相应的实验验证上述两个关键技术问题的有效程度。并在此基础上,在Android平台设计并实现了相应的移动应用示例。实验结果及应用示例表明,本文所设计的方案符合预期,具有一定的可行性。
[Abstract]:In recent years, with the rapid development of mobile Internet technology and positioning technology, Baidu map navigation, The application of geographical location awareness services such as Dianping has brought great convenience to people's lives. At present, there is a "knowledge gap" between geographical location and user needs. The lack of query support for spatial location relation and semantic relationship hinders the further development of geographical location awareness service. Combining image recognition with specific application services, we try to solve the following two technical problems: 1) conventional geographic information query services can not meet the needs of users in a specific situation. How to provide semantic geo-spatial location query service for users in this context has become an urgent problem to be solved. Ontology itself has the characteristics of good conceptual hierarchy. This paper presents an ontology-based address reasoning and POI(Point of interest) query model, which is based on ontology and assists with the language model of neural network through the construction of ontology in a specific situation. To provide support for spatial relation query and semantic relational query for users' environmental resources.) in view of the lack of consideration of rotation, translation and scaling of visual objects in spatial pyramid matching model in image recognition, Considering the performance bottleneck of mobile devices, this paper presents an image representation method based on feature point group similarity calculation model. By introducing feature point group similarity calculation model, the pyramid matching model is optimized, and the image anti-rotation is obtained. The anti-translation and anti-scaling representation can effectively improve the accuracy of mobile client marker image recognition. Finally, this paper designs corresponding experiments to verify the effectiveness of the above two key technical problems. An example of mobile application is designed and implemented on Android platform. The experimental results and application examples show that the scheme designed in this paper meets the expectations and has certain feasibility.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P209;P208

【相似文献】