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

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。%This paper uses the Shenzhen Futian district as a case study, to mine the massive amounts of taxi GPS data through time and space, for testing the correlation between the activities of residents, on and off taxi, and the urban land use pattern. It classiifes the taxi GPS data and veriifes its accuracy using the SVM classiifcation algorithm, while the optimal parameters are determined using the particle swarm optimization method. The results show that, the matching accuracy of the classiifcation of the taxi GPS data and the actual land use pattern of Futian district can be as high as 97.16%, which proves that the taxi GPS data can be used for real-time, accurate, dynamic awareness on urban land use pattern.