Addressing mixed pixel challenges in crop mapping: An integrated remote sensing and machine learning framework

Effective classification of crop types in mosaic farmland remains challenging due to spectrally mixed pixels and irregular field shapes. This study presents an integrated remote sensing framework combining texture analysis, subpixel classification, Ehlers fusion, and Object-Based Image Analysis (OBIA) to address these issues in Bhopal, Madhya Pradesh, India. The evaluation employed LISS-IV imagery (5.8 m) and extensive ground truth data collected during the crop season. Integrating spectral, spatial, and textural features significantly improved class separability and field boundary delineation. Random Trees achieved the highest performance with an overall accuracy of 98% and a Kappa coefficient of 0.971, while OBIA further strengthened mixed pixel handling using shape and geometrical features. Rather than proposing new algorithms, this work provides a reproducible, adaptable pipeline suitable for heterogeneous agro-ecosystems. The approach supports precision agriculture, food security planning, and crop monitoring, offering a scalable solution for crop inventory and resource management. Future work will focus on integrating deep learning and multi-temporal analysis to enhance operational performance.