Detecting Tree Height and Addressing Challenges Along Power Line Corridors Using Optical and Stereo Satellite Imagery

Vegetation encroachment along power transmission lines poses a critical challenge to infrastructure integrity, often leading to power disruptions. Traditional monitoring methods, such as LiDAR and synthetic aperture radar (SAR), are effective but costly and limited in coverage. This study introduces an innovative approach using Landsat satellite imagery for tree growth analysis. Various vegetation indices, including NDVI, EVI, SAVI, and NDWI, etc. are derived from Landsat data to quantify vegetation health. The final dataset, comprising 107,000 tuples, facilitates a comprehensive analysis of tree growth patterns. Concurrently, a Random Forest Regressor (RFR) is employed to create a predictive model, leveraging machine learning algorithms to correlate satellite indices with tree height. Simultaneously, stereo satellite imagery is utilized to generate Digital Terrain Models (DTM) and Digital Surface Models (DSM). The subtraction of DTM from DSM yields a Canopy Height Model (CHM), providing spatial details on tree heights. This CHM serves as a validation tool for the RFR model, ensuring accurate predictions. The integrated methodology combines Landsat's spectral information and stereo satellite's spatial data, offering a robust and cost-effective solution for monitoring and predicting tree growth along power line corridors. The RFR model exhibits an accuracy of 80%, with a Mean Squared Error (MSE) of 0.29 and an R-squared [10] value of 0.31, indicating its reliability in predicting tree height variations.