MONITORING CROP HEALTH IN MULTI-CROPPING SYSTEMS USING UAV-BASED LEAF SPECTRAL INDEX ANALYSIS AND GIS INTEGRATION

Abstract

The growing complexity of multi-cropping systems demands innovative monitoring solutions to ensure sustainable and efficient crop management. This study presents an integrated approach combining Unmanned Aerial Vehicle (UAV)-based spectral imaging, Geographic Information Systems (GIS), and machine learning techniques to monitor crop health across diverse cropping zones. UAVs equipped with multispectral sensors were deployed over intercropped fields, generating high-resolution imagery from which key vegetation indices—NDVI, SAVI, and GNDVI—were derived. These indices enabled the detection of plant stress, nutrient deficiencies, and disease symptoms with high spatial precision. Unlike ground truth, spectral measurements were proved to closely link to the state of healthy crops. Machine learning was used on spectral data to teach SVM, RF and CNN methods to identify crop health without being assisted by hand. CNNs often got it right, spotting damaged and stressed areas on 91 percent of the images and remembering them on 88 percent. Using GIS maps and tools, I carried out studies that map out ground areas and then produce buffer maps showing conservation areas. As a result, it became easy to notice variations in the field and take the right actions right there on the site. Stability of the system was illustrated through precision-recall curves, heatmaps and confusion matrices for diseases. According to data, using UAVs and AI in farming protects the environment, improves its accuracy and can scale output for large farms.