深度学习与大模型技术在北京中轴线实景重建中的应用
The application of deep learning and large model technology in the real reconstruction of Beijing Central Axis
王天舟 1张华 1刘奇申 1黄超1
作者信息
- 1. 腾讯科技(深圳)有限公司,深圳 518057
- 折叠
摘要
参考卫星地景数据以及北京市政公开的城市规划数据,并用大模型技术对该数据进行样本扩充,然后对这些样本数据进行多模型联合学习训练.通过模型训练对北京市区的高清卫星底图以及多光谱数据进行高精度的测量还原(包括城市建筑物底座、植被和道路),并运用大语言模型技术进行程序化三维建模,最后在虚幻引擎中自动生成.该流程对北京中心1 000 km2范围内进行了地景资产的复刻重现.本方法不仅提高了城市建模的效率和准确性,而且为城市规划、历史保护等领域提供了新的研究工具和视角.
Abstract
This paper refers to satellite landscape data and urban planning data published by the government,and uses large model technology to expand the sample size.Then,the sample data are used for multi-model joint learning training.Through model training,the high-definition satellite base map and multi-spectral data of Beijing urban area are measured and restored with high precision,including the urban building base,vegetation,and road.The large language model technology is used to program three-dimensional modeling,and finally automatically generates in the Unreal Engine.The process reproduces landscape assets within a 1 000 square kilometer area of central Beijing.This method not only improves the efficiency and accuracy of urban modeling,but also provides a new research tool and perspective for urban planning,historical preservation and other fields.
关键词
深度学习/大模型/重建Key words
deep learning/large model/reconstruction引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据