基于遥感影像时空谱特征联合的城市更新知识图谱构建研究

新型城镇化背景下，城市更新是面向存量优化的重要内涵发展方式。与遥感影像解译、野外实地调查等传统技术手段相比，高分场景提供了功能结构更新变化的重要观测。针对场景高频变化与分类多任务特性，需要高时间和空间分辨率影像的协同完成。现有研究对场景检测在城市更新领域的认知还不够深入，缺乏场景变化本质的理解以及时空协同的表达机制和模型支撑。针对此，本研究从遥感信息图谱视角出发，探究基于傅立叶谐波变换的时-空-谱特征表达，构建场景时空谱特征相关的深度学习变化分析模型，发展街景品质评分学习的微更新变化感知方法，探明集成场景街景品质变化的时-空-谱-属协同表达机制，有助于实现更新知识图谱的多层次表达与构建，揭示场景时空谱与更新图谱间的相互映射，提供图谱视角下面向更新检测的关键理论、技术方法和模型支持，为长时序大范围的城市更新快速检测提供理论参考，为拓展遥感信息在城市相关领域的应用提供崭新的思路和方法。

In the development of new urbanization, urban renewal is an important way for urban land stock optimization and high-quality development. Compared with the traditional technical ways such as remote sensing image interpretation and field investigation, the high resolution scene provides important observations on the changes of the urban functional structure. However, for the high-frequency changes of scenes and the multi-tasking characteristics of classification, high spatio-temporal resolution images are required to be coordinated. Previous research lacks an accurate understanding and representation of the nature of scene changes and the coordination of spatiotemporal feature. In response to this, inspired by the geography information spectrum theory, this study intends to focus on "feature extraction-feature spectrum representation-spectrum mapping", aiming to the improvement of the theoretical recognition that high- resolution scene detection applied to the construction of urban renewal spectrum. An representation analysis model combining spatio-temporal spectral features, street vision and quality features to solve the inconsistencies problem of data and logic in time and space. By the collaboration of multi-dimensional scene, the urban renewal map representation, mapping method and scientific basis are explored to provide the theories, technical methods and model construction for change detection. Moreover, it will be help to reveal the interaction between remote sensing spatio-temporal information maps and the urban renewal spectrum, and provide new ideas and methods for the rapid detection of long-term and large-scale urban renewal. At last, it can provide theoretical reference to using remote sensing spectrum in urban-related fields application.