Assisted approaches for forest and non-forest ground truth annotation in satellite remote sensing images / Imran Md Jelas, Mohd Asyraf Zulkifley and Mardina Abdullah

This paper presents a methodology for enhancing ground truth annotations in satellite remote sensing images using Python and OpenCV's SimpleBlobDetector. Focusing on a dataset from Chini Lake, Malaysia, the study adjusts threshold values and employs additional algorithms to reduce noise. The findings demonstrate significant accuracy improvements, with an average of 97.60% using the DeepLabV3+ algorithm. This paper highlights the importance of robust ground truth annotations for ecological monitoring and suggests future research directions.