Improved time series land cover classification by missing-observation-adaptive nonlinear dimensionality reduction

Abstract

Dimensionality reduction (DR) is a widely used technique to address the curse of dimensionality when highdimensional remotely sensed data, such as multi-temporal or hyperspectral imagery, are analyzed. Nonlinear DR algorithms, also referred to as manifold learning algorithms, have been successfully applied to hyperspectral data and provide improved performance compared with linear DR algorithms. However, DR algorithms cannot handle missing data that are common in multi-temporal imagery. In this paper, the Laplacian Eigenmaps (LE) nonlinear DR algorithm was refined for application to multi-temporal satellite data with large proportions of missing data. Refined LE algorithms were applied to 52-week Landsat time series for three study areas in Texas, Kansas and South Dakota that have different amounts of missing data and land cover complexity. A series of random forest classifications were conducted on the refined LE DR bands using varying proportions of training data provided by the United States Department of Agriculture (USDA) National Agricultural Statistics Service (NASS) Cropland Data Layer (CDL); these classification results were compared with conventional metricsbased random forest classifications. Experimental results show that compared with the metrics approach, higher per-class and overall classification accuracies were obtained using the refined LE DR bands of multispectral reflectance time series, and the number of training samples required to achieve a given degree of classification accuracy was also reduced. The approach of applying the refined LE to multispectral reflectance time series is promising in that it is automated and provides dimensionality-reduced data with desirable classification properties. The implications of this research and possibilities for future algorithm development and application are discussed.

Demonstration of Percent Tree Cover Mapping Using Landsat Analysis Ready Data (ARD) and Sensitivity with Respect to Landsat ARD Processing Level

“…As expected, the accuracies increased as more training data were used due to an increased likelihood of capturing spectral and temporal differences within and among percent tree levels [15][16][17]52]. The accuracy differences were negligible (≤0.07% RMSE) when the most training data (8% per tree) were used.…”

“…Similarly, the western sides of the spurs of the Rocky Mountains in the east of the study area have a high percent tree cover. 15.41% (ARD SRF), 15.42% (NBAR SRF ARD), and 15.48% (NBAR SRF WEEKLY) with the smallest (0.08%) amount of training per tree to 13.86%, 13.79%, 13.82%, and 13.81%, respectively, when the greatest amount (8%) of training per tree was used. The resulting maps based on the TOA ARD were more accurate than the ones based on the other processing levels when sparse training data (≤0.16% per tree) were used and were the least accurate when the most training data were used.…”

An Automated Approach for Sub-Pixel Registration of Landsat-8 Operational Land Imager (OLI) and Sentinel-2 Multi Spectral Instrument (MSI) Imagery

“…This was found to be helpful because of the spectral and temporal differences between the Sentinel and Landsat images. The SAM is insensitive to exogenous reflectance brightness variations as demonstrated using hyperspectral data [48][49][50] and multi-spectral multi-temporal data [51]. The SAM is conventionally derived between the spectral reflectance values of two pixels by calculating the angle subtended between their points in spectral feature space and the feature space origin (i.e., zero reflectance).…”

Rapid Land Cover Map Updates Using Change Detection and Robust Random Forest Classifiers

“…The current state of the practice for large area multi-temporal land cover Landsat classification is to derive metrics from the satellite time series and then classify the metrics bands with a supervised non-parametric classifier [60]. The monthly WELD composites were processed to rank-based metrics which have been shown to provide a generalized feature space that has advantages over time-sequential composite imagery in mapping large area [28].…”