| Title |
Crop yield estimation using multispectral satellite imagery and machine learning |
| Translation of Title |
Derliaus prognozavimas naudojant multispektrinius palydovinius vaizdus ir mašininį mokymąsi. |
| Authors |
Kaedbey, Jad |
| Full Text |
|
| Pages |
58 |
| Keywords [eng] |
crop yield estimation ; multispectral satellite imagery ; machine learning ; Sentinel-2 ; vegetation indices |
| Abstract [eng] |
This Master's thesis outlines the design and implementation of an automated system for estimating crop yields on a large scale using Sentinel-2 satellite imagery and machine learning. The research focuses on integrating time-series multispectral data with ground-truth statistics from the US Department of Agriculture's National Agricultural Statistics Service (NASS). A diverse suite of vegetation indices (including NDVI, EVI and NDRE) was calculated and combined with historical meteorological data to develop a comprehensive feature set that captures the complex dynamics of crop growth. The study compares the performance of tree-based ensemble models (specifically XGBoost and Random Forest) with that of experimental deep learning architectures, such as Long Short-Term Memory (LSTM) networks. The results show that, when enhanced with engineered weather features, tree-based models provide robust and scalable yield predictions at county level, offering a viable alternative to traditional reporting methods in precision agriculture. |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
English |
| Publication date |
2026 |
