A Hybrid Pipeline for Date Palm Tree Detection in High-Resolution Satellite Imagery

A novel hybrid method combining image processing with deep learning significantly enhances the detection of date palm trees in high-resolution satellite imagery, outperforming Mask R-CNN by 7.8% in complex, dense farms typical of Al-Ahsa Oasis. This approach merges spectral band analysis and peak detection to generate initial detections, which a convolutional neural network then refines using manually labeled data. Applied to WorldView-3 imagery, the method demonstrates its potential for large-scale palm tree estimation in this UNESCO-listed cultural landscape. The findings highlight the effectiveness of integrating traditional image processing with machine learning for precise agricultural monitoring in challenging environments.

Aljishi, F., Alharbi, A., De Weck, O., & Habib, A. (2023). A Hybrid Pipeline for Date Palm Tree Detection in High-Resolution Satellite Imagery. IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium, 6572–6575. https://doi.org/10.1109/igarss52108.2023.10282206