Sign in Subscribe
Machine learning

Optimised DNN-Based Agricultural Land Mapping Using Sentinel-2 and Landsat-8 with Google Earth Engine

Strategic Insights

— 2 min read

Punjab, India's agricultural sector is a crucial aspect of the country's economy, with a significant portion of the population dependent on it. To ensure sustainable development and effective management of agricultural lands, researchers have turned to remote sensing and machine learning techniques to classify and monitor land use. A recent study from Chandigarh University has made significant strides in this area by developing an optimized Hyper-tuned Deep Neural Network (Hy-DNN) model for land use and land cover (LULC) classification.

Key Takeaways:

  • The study used a combination of remote sensing data from Sentinel-2 and Landsat-8, processed on the Google Earth Engine (GEE) platform, to classify agricultural lands into four classes: agricultural land, vegetation, water bodies, and built-up areas.
  • The optimized Hy-DNN model achieved an overall accuracy of 97.60% using Sentinel-2 and 91.10% using Landsat-8, outperforming traditional classifiers such as Convolutional Neural Network (CNN), Random Forest (RF), Classification and Regression Tree (CART), Minimum Distance Classifier (MDC), and Naive Bayes (NB) by 3.19% and 7.59% respectively.
  • The study highlighted the superiority of the optimized Hy-DNN in agricultural land mapping and its potential use in crop health monitoring, disease diagnosis, and strategic agricultural planning.
  • The researchers used performance metrics such as producer's and consumer's accuracy, Kappa coefficient, and overall accuracy to measure the classification performance of the Hy-DNN model.
  • The Hy-DNN model was able to classify agricultural lands more accurately than traditional classifiers, with an overall accuracy of 97.60% and 91.10% for Sentinel-2 and Landsat-8 respectively.

Statistics:

  • Sender-2 multispectral data was used to achieve an overall accuracy of 97.60% with the Hy-DNN model.
  • Landsat-8 multispectral data was used to achieve an overall accuracy of 91.10% with the Hy-DNN model.
  • The optimized Hy-DNN model outperformed all traditional classifiers by 3.19% and 7.59% respectively using Sentinel-2 and Landsat-8 data.
  • The study was conducted by researchers from Chandigarh University's University Institute of Computing, led by Dr. Nisha Sharma.

Sources:

  • Optimised DNN-Based Agricultural Land Mapping Using Sentinel-2 and Landsat-8 with Google Earth Engine. Land, 2025,14(8):1578. (Land - http://www.mdpi.com/journal/land).
  • NewsRx. Chandigarh University Researchers Publish New Studies and Findings in the Area of Landscape Ecology (Optimised DNN-Based Agricultural Land Mapping Using Sentinel-2 and Landsat-8 with Google Earth Engine). Ecology, Environment & Conservation. September 12, 2025; p 52.

Read more