Location: Water Management ResearchTitle: Scale-aware pomegranate yield prediction using UAV imagery and machine learning
|NIU, HAOYU - University Of California
|EHSANI, REZA - University Of California
|CHEN, YANGQUAN - University Of California
Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/9/2023
Publication Date: 11/1/2023
Citation: Niu, H., Wang, D., Ehsani, R., Chen, Y. 2023. Scale-aware pomegranate yield prediction using UAV imagery and machine learning. Transactions of the ASABE. 66(5):1331-1340. https://doi.org/10.13031/ja.15041.
Interpretive Summary: Fruit trees are subject to high variabilities from tree to tree because of non-uniformities in orchards from natural and anthropogenic factors such as soil properties and irrigation amounts. Historically it has not been possible to document the tree-level variations until recent proliferation of small drones or unmanned aerial vehicles and lightweight optical sensors. Field experiments were carried out to estimate pomegranate fruit yield using drone images and data analyzed with machine learning algorithms. The research found that vegetation indexes derived from drone images could be used as predictors for fruit yield. The findings provide farmers valuable information early in the growing season on potential returns of their crop.
Technical Abstract: Monitoring the development of trees and accurately estimating the yield of trees are important to improve orchard management and production. Growers need to estimate the yield of trees at the early stage to make smart decisions for field management. However, methods to predict the yield at the individual tree level are currently not available because of the complexity of each tree. Therefore, this study aimed to evaluate the performance of an unmanned aerial vehicle (UAV)-based remote sensing system and machine learning (ML) approaches in pomegranate yield estimation. Lightweight sensors, such as multispectral camera, were mounted on the UAV platform to acquire high-resolution images. Eight features were extracted, including normalized difference vegetation index (NDVI), green normalized vegetation index (GNDVI), red-edge normalized difference vegetation index (NDVIre), red-edge triangulated vegetation index (RTVIcore), individual tree canopy size, the modified triangular vegetation index (MTVI2), the chlorophyll index-green (CIg), and the chlorophyll index-rededge (CIre). Research results compared the correlation (R2) of all the features with the yield of trees. Then, machine learning approaches were applied with the extracted features to predict the yield at the individual tree level. Results showed that the decision tree classifier had the best prediction performance, with an accuracy of 85% so far. The study has demonstrated the potential of using UAV-based remote sensing methods for pomegranate yield estimation. Predicting the yield at the individual tree level will enable the stakeholders to manage the orchard at different scales, thus improving the field management efficiency.