Search

One of the main applications of machine learning (ML) in remote sensing (RS) is the pixel-level classification of satellite images into land cover types. Although classes with different spectral signatures can be easily separated, e.g. aquatic and terrestrial land cover types, others have similar spectral signatures and are hard to separate using only the information within a single pixel. This work focused on the separation of two cover types with similar spectral signatures, cocoa agroforest and forest, over an area in Pará, Brazil. For this, we study the training and application of several ML algorithms on datasets obtained from a single composite image, a time-series (TS) composite obtained from the same location and by preprocessing the TS composite using simple TS preprocessing techniques. As expected, when ML algorithms are applied to a dataset obtained from a composite image, the median producer's accuracy (PA) and user's accuracy (UA) in those two classes are significantly lower than the median overall accuracy (OA) for all classes. The second dataset allows the ML models to learn the evolution of the spectral signatures over 5 months. Compared to the first dataset, the results indicate that ML models generalize better using TS data, even if the series are short and without any preprocessing. This generalization is further improved in the last dataset. The ML models are subsequently applied to an area with different geographical bounds. These last results indicate that, out of seven classifiers, the popular random forest (RF) algorithm ranked fourth, while XGBoost (×GB) obtained the best results. The best OA, as well as the best PA/UA balance, were obtained by performing feature construction using the M3GP algorithm and then applying XGB to the new extended dataset. [ABSTRACT FROM AUTHOR]

Data gathered in the real world normally contains noise, either stemming from inaccurate experimental measurements or introduced by human errors. Our work deals with classification data where the attribute values were accurately measured, but the categories may have been mislabeled by the human in several sample points, resulting in unreliable training data. Genetic Programming (GP) compares favorably with the Classification and Regression Trees (CART) method, but it is still highly affected by these errors. Despite consistently achieving high accuracy in both training and test sets, many classification errors are found in a later validation phase, revealing a previously hidden overfitting to the erroneous data. Furthermore, the evolved models frequently output raw values that are far from the expected range. To improve the behavior of the evolved models, we extend the original training set with additional sample points where the class label is unknown, and devise a simple way for GP to use this additional information and learn in a semi-supervised manner. The results are surprisingly good. In the presence of the exact same mislabeling errors, the additional unlabeled data allowed GP to evolve models that achieved high accuracy also in the validation phase. This is a brand new approach to semi-supervised learning that opens an array of possibilities for making the most of the abundance of unlabeled data available today, in a simple and inexpensive way. [ABSTRACT FROM AUTHOR]

The objective of this work is to develop a dynamic geographical information system (DGIS) applicable for the simulation of spatio-temporal phenomena. DGIS is designed and implemented as a dynamic raster GIS in a fully integrated systems strategy using theory of modelling and simulation, discrete event systems specifications, and knowledge-based simulation methodologies. The approach accommodates the representations and operators of current GIS in a broader systems representation scheme that includes the constructs for handling continuous time, model base, multiple resolution, hierarchy, taxonomy, variable structure, and intelligent agents. DGIS is defined as an endomorphic system that can be transformed by the user into different, application-specific, spatial dynamic simulation environments depending on the models selected to fit the various aspects of its structure. In this paper we describe the theory, methods, and technologies upon which the system is designed and implemented and discuss how it can support complex spatial simulations. To probe operational applicability, an instance of DGIS is generated: a fire spread simulation environment, which is tested in the simulation of 15 real fire events. [ABSTRACT FROM AUTHOR]

Genetic programming (GP) is a powerful machine learning (ML) algorithm that can produce readable white-box models. Although successfully used for solving an array of problems in different scientific areas, GP is still not well known in the field of remote sensing. The M3GP algorithm, a variant of the standard GP algorithm, performs feature construction by evolving hyperfeatures from the original ones. In this work, we use the M3GP algorithm on several sets of satellite images over different countries to create hyperfeatures from satellite bands to improve the classification of land cover types. We add the evolved hyperfeatures to the reference datasets and observe a significant improvement of the performance of three state-of-the-art ML algorithms (decision trees, random forests, and XGBoost) on multiclass classifications and no significant effect on the binary classifications. We show that adding the M3GP hyperfeatures to the reference datasets brings better results than adding the well-known spectral indices NDVI, NDWI, and NBR. We also compare the performance of the M3GP hyperfeatures in the binary classification problems with those created by other feature construction methods such as FFX and EFS. [ABSTRACT FROM AUTHOR]