Workflow Optimization for Photogrammetric Survey on Arctic Regions through Auto-GCP

The Arctic region's rapidly changing environment, driven by climate shifts and technological advancements, has heightened its strategic importance in commercial and military sectors, including navigation mapping and Intelligence Surveillance and Reconnaissance (ISR) activities. The Arctic's extreme conditions, characterized by illumination variations, sparse distinguishable objects, and monotonous surface features, challenge traditional GPS navigation systems, necessitating alternative technological solutions. Unmanned Aerial Systems (UAS), equipped with advanced optical sensors, offer a promising solution, providing high-resolution imagery while mitigating risks associated with crewed aircraft operations.One primary challenge in Arctic navigation and mapping is the correspondence problem—identifying matching features in overlapping images to accurately estimate the sensor's geometry and platform's position. Current technologies relying on feature detectors and descriptors often prove inadequate under Arctic conditions. This research aims to enhance the photogrammetric workflow, particularly the matching process, to improve the accuracy and reliability of UAS-based ISR operations in Arctic environments.The objective is to develop a comprehensive photogrammetric workflow for UAS-based ISR in Arctic regions. This involves evaluating commercial software performance under Arctic conditions, comparing feature extraction and matching techniques to support automated Ground Control Point (GCP) generation, and assessing the feasibility of deep learning (DL) based tie point extraction and matching. Additionally, the research aims to develop a combined hand-crafted and DL-based feature matching process optimized for Arctic imagery and evaluate the accuracy of aerial triangulation (AT)-adjusted Auto-GCPs, considering their dependence on environmental conditions. The study also seeks to analyze navigation performance in GPS-denied environments using Visual Terrain Referencing.The research will integrate traditional and DL-based feature extraction methods to enhance matching robustness. Initial matches will be identified using DL techniques in challenging areas, followed by refinement using traditional methods. Performance will be evaluated through experiments using Department of Defense (DoD)-provided Arctic imagery, focusing on geometric accuracy in relative orientation and digital surface model (DSM) generation. The proposed workflow aims to interface with major Geographic Information System (GIS) software, supporting both onboard and real-time operations.The anticipated outcomes of Phase I include the development of a set of algorithms and software tools that will require further refinement and development in Phase II to transition towards commercialization. The goal is to create a software product that can be seamlessly integrated into existing DoD workflows and commercial markets, particularly in austere environments like the Arctic.