Land Use Classification from Lidar Data and Ortho‐Images in a Rural Area

Abstract

Obtaining information on the distribution of rural landscape types is an active research topic within Spanish rural studies. This paper presents a new hierarchical object‐based classification method for the automatic detection of various land use classes in a rural area, combining lidar data and aerial images. In view of the upcoming availability of low‐density lidar data (0·5 pulses/m2) for most of the territory of Spain, this paper assesses the feasibility and accuracy of the proposed method for various lidar data densities. Such an assessment was conducted using two approaches: firstly, based on the final classification, which produced an overall accuracy over 96% and a kappa index above 0·95 for the combinations of the aerial image and lidar data‐sets with four different densities; and secondly, based solely on the areas classified as buildings. In the second approach, the accuracy of the classification for building detection at pixel and object level was assessed. The object‐oriented classification of buildings produced an index of correctness of over 99% and an index of completeness of about 95%. The results reveal a high agreement between classification and ground truth data.

Urban land cover classification using airborne LiDAR data: A review

“…On the other hand, the nDSM derived from the airborne LiDAR data was used to distinguish the elevated objects (building) from ground. Similar design of classification trees can be found in Sasaki, Imanishi, Ioki, Morimoto, and Kitada (2012) and Buján et al (2012). Hartfield et al (2011) reported that the classification accuracy of multi-spectral image together with the Normalized Difference Vegetation Index (NDVI) for 8 classes was 84%.…”

“…The empirical approach does not consider the physical properties of the laser backscattered energy. Instead, it introduces statistical methods to minimize the noise in the intensity data, such as median filter (Buján et al, 2012;Song et al, 2002). Fang and Huang (2004) introduced a discrete wavelet transform approach for noise reduction in LiDAR signal; the results demonstrated that the proposed method outperforms the traditional digital filters by improving the signal-to-noise ratio.…”

“…Bartels and Wei (2006) performed similar experiments and claimed that such feature can improve the accuracy; however, no individual quantitative measure of each feature was included in the paper. Brennan and Webster (2006) and Buján et al (2012) adopted the multiple-return data as one of the criteria in the object-oriented decision tree classifier to distinguish permeable object (e.g. tree canopy) from non-permeable object (e.g.…”

Spatial application of Random Forest models for fine-scale coastal vegetation classification using object based analysis of aerial orthophoto and DEM data

“…The combination of aerial orthophoto imagery and object-based analysis has mostly been applied for target specific mappings such as urban and infra-structural cover (Guan et al, 2013;Poznanska et al, 2013) where target elements are relatively easily discernable. With a rural landscape as the study area, an object-based land cover classification, of similar thematic detail to the coarse structure classes presented here, has achieved an overall classification accuracy of 96% in a small scale site (Bujan et al, 2012). The prediction accuracies in that study were however increased by combining aerial imagery with high point density LiDAR to discern classes which were mainly height related.…”

Land Use/Land Cover Monitoring and Geospatial Technologies: An Overview

“…In applications to distinguish features within shadowed areas, researchers have used Lidar data with aerial images to extract land use for rural Spain [57]. These researchers found that the combination of these two types of data allowed extraction of land uses within shadowed areas.…”