大数据条件下城市用地类型辨识研究——基于出租车GPS数据的动态感知

发布时间：2018-03-05 16:21

本文选题：城市用地　切入点：类型辨识　出处：《上海国土资源》2016年01期 　论文类型：期刊论文

【摘要】：随着大数据的广泛应用以及城市规划管理的智慧化、精细化转变,土地管理需要丰富城市探查视角,开拓用地分类方式。传统遥感影像分类具有时效性低、不能反映真实用地功能的弊端,而基于价值密度低、时效性高的GPS数据进行土地分类可以有效克服这一不足。本文以深圳市福田区为案例,通过海量出租车GPS数据的时空挖掘,在验证居民活动、出租车上下车活动与城市用地类型之间相关关系的基础上,利用SVM监督分类算法和粒子群优化方法得出的最优参数对出租车GPS数据进行分类和准确率验证。结果显示:出租车GPS数据的分类结果与福田实际的土地利用类别匹配精确度高达97.16%,可以证明出租车GPS数据能够实现对城市土地利用类型实时、精确的动态感知。
[Abstract]:With the extensive application of big data and the intelligent transformation of urban planning management, land management needs to enrich the urban exploration perspective and develop land classification methods. Land classification based on GPS data with low value density and high timeliness can effectively overcome this shortcoming. This paper takes Futian District of Shenzhen City as a case study, through spatiotemporal mining of massive taxi GPS data. On the basis of verifying the correlation between residents' activities, taxi boarding and alighting activities and urban land types, The optimal parameters of SVM supervised classification algorithm and particle swarm optimization are used to classify taxi GPS data and their accuracy is verified. The results show that the classification results of taxi GPS data and the actual land use categories of Fukuda are obtained. The accuracy of the match is as high as 97.16, which can prove that taxi GPS data can achieve real-time urban land use types. Accurate dynamic perception.
【作者单位】：北京大学城市与环境学院;北京大学深圳研究生院;
【基金】：国家自然科学基金项目(41371534)
【分类号】：P228.4;F299.27

【相似文献】