Google Earth Engine-based Machine Learning Tool for Fast Crop Classification

The study suggests a cloud-integrated crop classification system based on the Google Earth Engine (GEE) platform for real-time monitoring, analysis, and visualization of agricultural crops. The users have the option of defining their region of interest by drawing a boundary, selecting specific date ranges, and applying cloud cover thresholds to clip the satellite images available. The software allows for visualization of NDVI temporal curves by crop type, allowing the user to choose and remove specific cloudy or noisy images from the temporal data set to improve accuracy. The application hosts three machine learning algorithms, Support Vector Machine (SVM), Classification and Regression Tree (CART), and Random Forest (RF) to generate the crop classification maps in near real-time. The tool was validated using data collected from the Central State Seed Farm in Hisar district, Haryana, India to classify major Rabi season crops (Wheat, Mustard, Barley, Oat, and Gram). During the iterative training and testing phase within the GEE platform, the machine learning models achieved a training accuracy of up to 96%. The system automatically identifies the most accurate trained model with which to create the final classified map. The resulting map is displayed together with the original true-color composite for quick reference and comparison. The tool is able to distinguish between major crop types and detect mixed cropping regions based on discriminative NDVI temporal patterns. By means of automation, visualization, and real-time processing, the tool provides a practical and scalable solution for operational crop mapping in regions where in situ training data are available. It is a vital decision-making tool for policymakers and farmers to offer evidence-based agricultural planning and management.