Advancing Forest Segmentation: Prioritizing lightweight DeepLabV3+ for Precision Environmental Monitoring

Accurate forest segmentation from satellite imagery is crucial for environmental monitoring, conservation, and natural resource management. Forests, as significant carbon sinks, play a vital role in absorbing atmospheric carbon dioxide and mitigating climate change. Traditional methods like UNet and SEGNet often struggle with limitations in precision and adapt-ability to diverse forest conditions, impeding efforts to monitor and manage these vital ecosystems sustainably. This study aims to elevate forest segmentation performance by prioritizing the DeepLabV3+ model with lightweight EfficientNet backbones over UNet and SEGNet. These models are evaluated using forest satellite imagery, incorporating advanced hyperparameters optimization techniques and two-stage learning strategies. Our research demonstrates that DeepLabV3+, with its superior architecture and enhanced decoder capabilities, significantly outperforms UNet and SEGNet in key performance metrics, including Mean Intersection Over Union, Mean Pixel Accuracy, and Dice Coefficient. The results underscore the efficacy of DeepLabV3+ for precision environmental monitoring and highlight its potential to improve carbon sequestration insights, thereby supporting effective climate change mitigation strategies and contributing to the achievement of Sustainable Development Goals.