Neural Network Model for Land Cover Classification from Satellite Images

Abstract

A B S T R A C TLand cover data represent environmental information for a variety of scientific and policy applications, so its classification from satellite images is important. Since neural networks (NN) do not require a hypothesis about data distribution, they are valuable tools to classify satellite images. The objectives of this work were to develop NN models to classify land cover data from information from satellite images and to evaluate them when different input variables are used. MODIS-MYD13Q1 satellite images and data of 85 plots in Córdoba, Argentina, were used. Five NN models of multi-layer feed-forward perceptron were designed. Four of these received NDVI (Normalized Difference Vegetation Index), EVI (Enhanced Vegetation Index), red (RED) and near infrared (NIR) reflectance values as input patterns, respectively. The fifth NN had RED and NIR reflectances as input values. By comparing the information taken in the field and the classification made during the validation phase, it can be concluded that all models presented good performance in the classification. The model that shows better performance is the one that jointly considers RED and NIR reflectance as input; this model shows an overall classification accuracy of 93% and an excellent Kappa statistic. The networks constructed with NDVI and EVI values have a similar perfomance (86 and 83% accuracy, respectively). The Kappa statistics correspond to the categories of very good and good, respectively. The networks including only RED or NIR reflectance values get the lowest accuracy results (76 and 81%, respectively) and Kappa values within fair and good ranks, respectively.Keywords: modeling, back-propagation neural networks, remote sensing, crops-bare soil. R E S U M E NLos datos de cobertura de suelo representan información ambiental clave para aplicaciones científicas y políticas, por esto su clasificación a partir de imáge-nes satelitales es importante. Las redes neuronales (NN) constituyen una herramienta valiosa para clasificar imágenes satelitales pues no requieren hipótesis sobre la distribución de los datos. Los objetivos de este trabajo fueron desarrollar modelos de NN para clasificar datos de cobertura de suelo a partir de información proveniente de imágenes satelitales y evaluarlos cuando se utilizan diferentes variables de entrada. Se utilizaron imágenes satelitales MODIS-MYD13Q1 y datos de 85 parcelas en Córdoba Argentina. Se diseñaron cinco NN del tipo perceptrón multicapa feed-forward. Cuatro de ellas recibieron como patrones de entrada valores de NDVI (Normalized Difference Vegetation Index), EVI (Enhanced Vegetation Index), de reflectancias en la bandas roja (RED) y en infrarroja cercana (NIR), respectivamente. La quinta NN tuvo como valores de entrada las reflectancias RED y NIR. La validación, permitió concluir que todos los modelos presentan un buen desempeño. El modelo que muestra mejor comportamiento es aquel que considera conjuntamente valores de reflectancias RED y NIR, cuya precisión en la clasificación es del 93% con un ...

Spectral vegetation indices performance evaluated for Cholistan Desert

“…In 'increase' category, the MSAVI2 showed minimum increase, the EVI showed increase and the result of the SAVI, the NDVI and TNDVI was same. The modeling process is effective to estimate land cover from satellite images, even using a limited number of data (Bocco et al, 2007). The EVI is the best to optimize the vegetation signal with improved sensitivity in high biomass regions by incorporating both background adjustment and atmospheric resistance concepts.…”

Clasificación de uso del suelo y vegetación con redes neuronales convolucionales

“…Durante la fase de aprendizaje, se aplicó una función de transferencia a través de una serie de iteraciones para comparar los valores predichos con los valores observados (Bocco et al, 2007). El conjunto de pruebas no es visto por el modelo en el entrenamiento y se utiliza después, tras el ajuste de los hiperparámetros para proporcionar una evaluación imparcial del modelo final.…”

Estimating soybean ground cover from satellite images using neural-networks models

“…NN architecture, in general, consists of three layers: the input, the hidden and the output; input and output layers contain neurons that correspond to input and output variables, respectively (Bocco et al 2007). NNs learn from the existing information through a training process by which their parameters (weights) are adjusted so as to provide an approximate output close to the desired one.…”

“…NNs appear to be quite powerful and easy to operationally implement because they have the ability to compute, process, predict and classify data (Bocco et al 2007, Verger et al 2008; they also allow adjustment of a response even with a complex and non-linear problem, input-output mapping, adaptivity, generalization and fault tolerance (Panda et al 2010).…”