01558285

Component

In the era of big data, harnessing remote sensing data for transportation decision making has become an achievable task. This paper focuses on the land use classification on the finest parcel scale by using the remote sensing data as the input. Different from other relevant research, the authors utilized the multinomial logitistic regression, or called multinomial logit (MNL) models, whose great potentials have been overlooked for remote sensing based land use classification. In addition, the authors also suggest using transportation related attributes, such as the distances from a parcel of land to the nearest road or intersection, as the ancillary attributes to improve classification performance, in addition to spectral features collected by remote sensing. The MNL models were tested on the land use data collected in the City of Buffalo, New York. The best model achieves an average prediction accuracy of 83.7%. For the residential and commercial parcels, the prediction accuracy reaches up to 94.5%. In addition, the suggested transportation attributes were also found significant in discriminating land use classes. Two main conclusions were raised from the research, including remote sensing as a reliable data source for timely updating land use and land cover, and the applicability of the MNL models for land use classification with remote sensing.

This paper was sponsored by TRB committee ADD30 Transportation and Land Development.

01550057

15-5128

English

16p

2015

Transportation Research Board 94th Annual Meeting

Digital/other

Figures; Maps; References; Tables

Commercial strips; Land use; Multinomial logits; Remote sensing; Residential areas; Transportation planning

Buffalo (New York)

Highways; Planning and Forecasting; Public Transportation; I72: Traffic and Transport Planning

Transportation Research Board Annual Meeting 2015 Paper #15-5128

TRIS, TRB, ATRI

Dec 30 2014 1:43PM