基于无人机遥感与AI深度学习的中东铁路城镇风貌特色规划范式研究

Characteristic townscape, as the vision carrier of towns’ quality, is the crucial part in urban and rural planning and design. The traditional planning and design methods of characteristic townscape are subject to the professional background, emotion and limited energy, which greatly limit the development of its planning and design. This research project is a cross-border subject combining with artificial intelligence(AI) disciplines, supported by the scientific techniques of remote sensing from the discipline of electronic and information engineering, and also integrated with the social science methodology of history science; as a result, the purpose of this research task would be achieved in a way of cross-disciplines and subjects fusion. This research takes station-located towns along Chinese Eastern Railway within the area Heilongjiang Province as the main research objects and UAV is equipped with many types of sensors to obtain a considerable number of images and data. And deep learning algorithm is applied to interpret and analyze the acquired images and data. Based on the typology of towns’ space, the characteristic townscape is identified and classified to constitute the evaluation framework and the townscape guiding management methods, which both breaks through the barriers of traditional planning and design and realizing the paradigm innovation of characteristic townscape planning.

城镇风貌特色作为承载城镇品质的重要视觉载体，是城乡规划与设计工作中倍受关注的内容，鉴于传统城镇风貌特色规划与设计方法受限于规划师的专业背景、情感因素以及有限精力等问题，在某种程度上限制了该方法的发展。本项目跨界人工智能学科领域，借助电子与信息工程专业的遥感技术方法，再辅以历史学的社会科学方法论，以多学科交叉、互动的方式，达成研究目的。以黑龙江省境内的中东铁路设站城镇为主要研究对象，通过无人机搭载多种类传感器，多途径获取海量城镇影像数据，并借助深度学习算法对所获图像数据进行解译与判读，在城镇空间类型学研究的基础上，完成风貌特色要素的识别与分类，构建城镇风貌特色评价体系，并生成导控技术，从而突破传统规划设计方法的壁垒，实现风貌特色规划的范式创新。

中东铁路沿线城镇风貌早期受到俄罗斯建筑风格、规划理论的影响，形成了区别于传统中国城镇的特色风貌，时至今日仍留存分布广泛、数量众多、形式多样的建筑群落，具有研究价值。同时中东铁路作为线性文化遗产，其风貌的保护工作应是一个系统工程，区别于以单个城市、城镇或街区为研究对象的风貌研究，本项目以整体性的视角探讨中东铁路沿线的风貌保护范式，防止线性遗产城镇风貌割裂。在整个研究过程中，无人机、深度学习等新兴的科技设备与技术方法协助了本项目快速的获取并客观的识别风貌要素，相较受限于规划师专业背景、情感因素以及有限精力的传统城镇风貌特色规划方法而言，更加高效、客观与科学。.项目遵循“数据采集与风貌调查——城镇类型划分——风貌特色评价——风貌导控策略提出”四个研究阶段。项目选取中东铁路黑龙江段较为典型的35个城镇为研究对象，应用无人机搭载多种类传感器并结合新兴大数据，获取了海量城镇影像数据。在此基础上，结合深度学习、GIS、量化模型等多元技术方法，完成沿线城镇风貌类型的划分。建立中东铁路风貌特色评价体系，通过主观与客观相结合的评价认知中东铁路沿线城镇的共性特征、聚类特征与个性特征。最后，针对各类特征选取典型城镇制定了相应的城镇风貌导控策略，实现了从单一到整体、从主观到科学的线性特色城镇群风貌保护范式的转变。