传统村落空间特征快速量化与比较研究

The three levels of spatial feature indexes statistics -- folk dwellings, street and alley together with terrain, are the common groundwork in various researches of traditional village. However, distinguished from official architecture, in traditional villages, extensive folk buildings, winding streets with twists and turns as well as rugged terrain cause great difficulties and inefficiency to spatial index statistics. At present, as there are already about 6000 national level traditional villages in China, the current working pattern---Field measurement-->Manual tracing-->One-by-one statistics-->Multi-village comparison---which occupies excessive precious resources has become the bottleneck of researches on traditional villages. .This project is about building a relatively complete three-dimensional computing work chain for spatial features, based on a novel 3D point cloud direct calculation technique and supplemented with other known computing means, such as GIS, spatial syntax, etc., aiming to obtain abundant approximate calculation results of village spatial indexes in a semi-automatic and efficient way without tracing. Thereby, not only meet needs in the progress of village field surveys but deeply and widely satisfy the requirements of comparative studies among multiple villages across regions..In order to verify the accuracy and reliability of the method, those approximate calculation results mentioned above will be compared with both artificial actuarial values and subjective feeling in the field. In addition, an exemplary study will be done from three aspects of traditional village which are folk dwelling’s styles, street and alley identification, location environment in order to test the practical effect of obtained spatial estimate indexes, and meanwhile try to quantify and visualize inherent wisdom of space contained in Chinese traditional villages.

民居、街巷、地形三个层级的空间特征指标统计是传统村落各类研究共同的基础性工作，但与官式建筑相比传统村落中民居建筑数量多、街巷迂回曲折、地形起伏多变，空间指标统计难度大、效率低；且全国范围内仅国家级传统村落数量已约六千个，“外业测量->人工描图->逐一统计->多村落比较”的工作模式占用过多宝贵资源，成为传统村落研究的瓶颈。本课题基于新颖的三维点云直接计算技术，辅以GIS、空间句法等其他已知计算手段，构建相对完整的空间特征三维化计算工作链，以免描图、半自动、高效率的方式获得丰富的村落空间指标概算结果，满足村落现场调查的进度要求和跨地域多村落比较研究的深度、广度要求。上述概算结果亦与人工精算值和现场主观感受比对，以验证方法的准确度、可靠性。在前述基础性研究之上，本课题还从民居风貌、街巷辨识、选址环境三个层面以示例性比较研究测试概算指标的实用效果，同时尝试量化和可视化我国传统村落蕴含的空间智慧。

在乡村建设、西部发展等的社会需求背景下，本课题主要利用低空信息采集技术、基于三维点云的地物分类技术、图像及三维目标的人工智能识别提取技术，针对传统聚落的各级多类空间特征实施快速采集、跨地域统计、量化的模拟分析等工作。在这一过程中，针对传统聚落所在的复杂自然环境总结优化的数据采集作业方法，针对聚落中的民居建筑、常见设施构件等目标，利用机器学习技术获得比较成熟的识别模型、总结人工智能新技术应用的经验，并凝练传统聚落在选址、规划、设计方面蕴含的智慧。本课题已采集17个省市超过132个传统村落的三维数据，自主编程构建了大型图像与三维数据库，支撑了“中国传统村落数字博物馆”这样的重点工程实施，尤其是对如何深挖数字博物馆中的数据潜力，对本领域提到了一定的启发作用。对于国内外各类线性遗产、线路遗产、群体遗产，本课题的研究都具有一定的参考价值。