12 Bands vs. NDVI: Sentinel-2 Crop Identification
12 Bands vs. NDVI for Crop Identification
Published in: Sensors (Volume 23)
Publisher: MDPI
DOI: 10.3390/s23167132
Abstract: Today, machine learning applied to remote sensing data is used for crop detection. This makes it possible to not only monitor crops but also to detect pests, a lack of irrigation, or other problems. For systems that require high accuracy in crop identification, a large amount of data is required to generate reliable models. The more plots of and data on crop evolution used over time, the more reliable the models. Here, a study has been carried out to analyse neural network models trained with the Sentinel satellite’s 12 bands, compared to models that only use the NDVI, in order to choose the most suitable model in terms of the amount of storage, calculation time, accuracy, and precision. This study achieved a training time gain of 59.35% for NDVI models compared with 12 band models; however, models based on 12 band values are 1.96% more accurate than those trained with the NDVI alone when it comes to making predictions. The findings of this study could be of great interest to administrations, businesses, land managers, and researchers who use satellite image data mining techniques and wish to design an efficient system, particularly one with limited storage capacity and response times.
This peer reviewed research paper investigates the trade off between using all 12 spectral bands of Sentinel 2 satellite imagery versus the widely used Normalized Difference Vegetation Index (NDVI) for neural network based crop classification.
The Research Question
Remote sensing for agriculture traditionally relies on vegetation indices like NDVI, which compress multispectral information into a single value. But Sentinel-2 provides 12 distinct spectral bands: from visible light to short-wave infrared. Does using all that raw spectral data actually improve crop identification accuracy?
Methodology
We trained Convolutional Neural Networks (CNNs) to classify ten different crop types across the Extremadura region of Spain. Two approaches were compared:
- 12-Band Input: Feeding the full multispectral stack directly into the network
- NDVI Input: Using only the derived vegetation index as input
Both models were trained on the same geographic regions and temporal windows using Sentinel-2 imagery from the EU's Copernicus programme.
Key Findings
The 12-band approach achieved marginally higher classification accuracy (averaging 0.9334 compared to 0.9141 for NDVI), but at a significantly higher computational cost. The NDVI-based approach proved remarkably effective for its simplicity, achieving a 59.35% training time gain over the 12-band models. This suggests that for many practical agricultural monitoring applications, especially those with limited storage or requiring rapid response times, the simpler approach may be the more efficient choice.
The study contributes to the ongoing discussion about the balance between precision, efficiency, and computational resources in remote sensing workflows.
Citation
Published in MDPI Sensors, August 2023.