Resumen de Discriminación de trigo y cebada empleando imágenes satelitales ópticas y radar. Estudio de caso: partido de Coronel Rosales (Argentina)
- español
El partido de Coronel Rosales (Buenos Aires, Argentina) se halla localizado dentro de la región pampeana austral, una de las de mayor relevancia agro productiva del país. En este contexto, el conocimiento de la superficie cultivada adquiere significativa importancia para la posterior planificación agrícola y económica. En tal sentido, la discriminación de cultivos mediante teledetección se dificulta cuando se trata de los de ciclo fenológico muy similar, como el trigo y la cebada. En este estudio se realizó una discriminación de dichos cultivos empleando imágenes de Radar de Apertura Sintética (SAR) Sentinel-1A SLC, imágenes ópticas Sentinel-2 y una combinación de ambos tipos de datos. Se incorporaron medidas de coherencia, textura e intensidad de retrodispersión extraídas de los datos SAR durante el ciclo fenológico completo. Sobre cada escena Sentinel-2 se obtuvo el Índice de Diferencia Normalizada de Vegetación (Normalized Difference Vegetation Index - NDVI). Se emplearon tres algoritmos de clasificación: Máxima Verosimilitud (Maximum Likelihood - MLC), Máquinas de Soporte Vectorial (Support Vector Machines - SVM) y Random Forest (RF). Los mejores resultados se obtuvieron al combinar imágenes ópticas y SAR empleando el clasificador RF. La combinación de las retrodispersiones VV y VH junto a la coherencia y la textura de las imágenes SAR, sumada al apilado de NDVI de imágenes ópticas, arrojó los máximos valores de precisión de la clasificación. El valor de F1 fue de 87.27% para el trigo y de 89.20% para la cebada.
- English
TIn Argentina, the farming industry is considered one of the main economic resources in terms of income and domestic market supply. Thus, the study, inventory, and knowledge of the cultivated surface area are key cornerstones for agricultural and economic planning. Agriculture focuses mainly on cereals such as wheat, barley, maize, oat, and sorghum, as well as on oilseeds such as soybeans, sunflower, and peanuts. The most important productive areas of Argentina include the Pampean region, where the Coronel Rosales Department is located (Buenos Aires, Argentina). In this context, the knowledge of the cultivated surface area is particularly important to support agricultural and economic planning. In this regard, crop discrimination based on remote sensing is difficult for crops with highly similar phenological cycles, as is the case of wheat and barley. To address this issue, the standard satellite image classification methods have been based on the spectral response of each individual pixel using optical images. Crops are also monitored using Synthetic Aperture Radar (SAR) images; these have several advantages over optical imagery because radio waves are unaffected by the presence of clouds. This provides the benefit of recording satellite data throughout the whole phenological cycle. The aim of this work is to discriminate wheat from barley crops grown in the Coronel Rosales Department by using SAR data, optical images, and the combination of both approaches. To this end, we used coherence (C), texture (T), and backscatter intensity images extracted from SAR data derived from the Sentinel-1A satellite over the whole phenological cycle of both winter crops. The optical images used were Sentinel-2, with six representative dates selected from the different phenological states of each culture. To have comparable parameters over time, we obtained the reflectance value for each band used, and the Normalized Difference Vegetation Index (NDVI) for each scene. SAR images from the Sentinel-1 satellite were also used, including seven dual polarization images (VV and VH), and the Single Look Complex (SLC) product. Three classification algorithms were used: Maximum Likelihood (MLC), Support Vector Machines (SVM), and Random Forest (RF). To evaluate each classification, a confounding matrix was generated, from which two accuracy measurements are derived: Global Accuracy (P), i.e., the ratio between the total number of pixels correctly classified and the total number of pixels corresponding to field-based (true) classifications, and the F1 coefficient. The latter uses the accuracy percentages corresponding to the producer and the user, derived from the confounding matrix The study showed that the use of SAR data allowed an optimal discrimination of both winter crops. The combination VV + VH + C+ T yielded the best results for wheat (F1: 79.74%), while the combination C + T was the best for discriminating barley (F1: 84.94%). From the above, we can conclude that combining consistency, texture, and backscatter images improves classification accuracy. Besides, our results corroborate that optical data can be replaced by Sentinel-1A images when the scenes from the former are limited by the presence of clouds. Classification using optical images (NDVI image stacking) yielded optimal results using the MLC algorithm (P: 80.62%). In contrast, for winter crop discrimination, the highest F1 values were obtained using the RF method (82.95% for wheat and 84.16% for barley). The best results of the overall classifications performed were obtained by combining optical images and SAR. The combination of all SAR images (VV + VH + C+ T) with NDVI stacking yielded the highest Overall Accuracy value (89.37%). For winter crop discrimination, the combination of Sentinel-1A data with optical data was also most accurate. The combination VV + VH + C+T+ NDVI yielded F1 values of 87.27% for wheat and 89.20% for barley using the RF algorithm.
