Impact of Environmental Factors on Water Body Segmentation: A Study of Deep Learning Models

Water body segmentation is a critical task in remote sensing and environmental monitoring, with applicationsranging from flood management to natural resource assessment. This task is complex due to many factors such aswater color, vegetation, and land type variations. One major and often neglected challenge is the impact of nightlighting conditions. In this paper, we present a novel dataset specifically designed to enhance performance underdiverse conditions especially with nighttime scenarios. Utilizing nighttime imagery offers advantages such asreduced solar reflection, minimized glare, and clearer detection of water boundaries, which can be particularlyuseful during periods of haze, cloud cover, or when daytime observations are limited. Our methodology includes thecollection, annotation, augmentation, and performing selective slicing on this dataset, followed by the training and evaluation of advancedsegmentation models. The results demonstrate significant improvements in model performance, including accuracy,IoU, and precision, across various scenarios, particularly in handling previously challenging conditions. This workhelps advance the state of water body segmentation and provides a valuable resource for future research inenvironmental monitoring and remote sensing.