Detection of irrigation malfunctions based on thermal imaging

This work presents an algorithm for monitoring and mapping irrigation system malfunctions based on airborne thermal imaging data. Data from 100 ha of olive groves were collected in 2012 using an airborne thermal camera. Ground truth was determined manually by scouting. Image segmentation was performed by merging Continuous Max-Flow-Min-Cut with the Otsu method. This was followed by a Subpixel Edge Detection method to avoid mixed pixels. Irrigation of trees classification was performed by Bagging with Random Forest algorithms using features derived from the thermal images. Leaks or clogging irrigation malfunctions were successfully detected with 89.5 and 87.5% success rates, respectively.