Improving Land Cover Classifications with Multiangular Data: MISR Data in Mainland Spain

Abstract

In this study, we deal with the application of multiangular data from the Multiangle Imaging Spectroradiometer (MISR) sensor for studying the effect of surface anisotropy and directional information on the classification accuracy for different land covers with different rate of disaggregation classes (from four to 35 different classes) from a Mediterranean bioregion in Iberian, Spain. We used various MISR band groups from nadir to blue, green, red, and NIR channels at nadir and off-nadir. The MISR data utilized here were provided by the L1B2T product (275 m spatial resolution) and belonged to two different orbits. We performed 23 classifications with the k-means algorithm to test multiangular data, number of clusters, and iteration effects. Our findings confirmed that the multiangular information, in addition to the multispectral information used as the input of the k-means algorithm, improves the land cover classification accuracy, and this improvement increased with the level of disaggregation. A very large number of clusters produced even better improvements than multiangular data.

Sand Dunes Spectral Index Determination Using Machine Learning Model: Case study of Baiji Sand Dunes Field Northern Iraq

“…Several machine learning (ML) models have been developed for modeling heavy metals over the past two decades, with outstanding progress (Yaseen, 2021). During the past two decades, the (SVM) and (RF) classifiers have brought image classification to the forefront of remote sensing applications (Novillo et al, 2018). The relatively high classification accuracy of Support Vector Machine SVM and Random Forest RF places them among the most popular machine learning classifiers in the remote sensing domain.…”

A review of global land cover maps in terms of their potential use for habitat suitability modelling

“…Land cover maps can be created from satellite images in free open-source programs (Manton et al 2005;Ndegwa Mundia and Murayama 2009;Barik et al 2021). Precision of such maps depends not only on the quality of satellite imagery (Manton et al 2005), but also on the classification approach (Li et al 2017) and on type of data used (Novillo et al 2018). However, for a global scale study, creating land cover maps from individual satellite images is extremely time consuming and this process will often exceed storage memory and processing capacities of a personal computer.…”