Enhancing Aerial Image Georeferencing with Innovations in Pixel-Level Semantic Segmentation for Improved Precision in Mapping

In the realm of aerial image georeferencing and mapping, the primary challenge is to enhance georeferencing accuracy and segmentation efficiency for precise remote sensing applications. The core obj ective is to refine georeferencing precision, ensuring heightened mapping reliability, while concurrently developing efficient segmentation methods for valuable geospatial data extraction. The incorporation of the Coco dataset validates and enhances the effectiveness of the segmentation technique devised in this paper. This paper introduces an innovative approach centered on refining pixel-level semantic segmentation for both aerial images and georeferencing processes. This proposed method marks substantial advancements in accurately categorizing and delineating objects within aerial imagery, contributing to an elevated precision in geo-referencing. The emphasis on improved pixel-level semantic segmentation underscores the commitment to enhancing the efficacy of georeferencing in the context of aerial images. The proposed approach demonstrates remarkable performance metrics, including an accuracy of 96.71 %, precision scaling to 98.75%, and a commendable recall of 90.62 %. Through comprehensive comparative analysis with established models, such as semantic segmentation, panoptic segmentation, and 3D semantic segmentation, this method emerges as a leader in the field.