GPS-Denied Terrain Relative Navigation (TRN) via Oblique Camera and DSM/DEM Feature Contrast Window Matching for UAVs

Unlike traditional Nadir (straight-down) map-matching algorithms which may require higher flight altitudes to obtain a sufficient Field of View (FOV), this paper discloses a method utilizing an oblique (angled) camera view for absolute position bounding. By tilting the optical sensor at a designated pitch angle relative to the horizon, vertical structures and landscape features can be effectively captured even during low-altitude flights. The algorithm evaluates candidate positions within a search grid by projecting known terrain features from a co-registered database into the 2D image plane. This reference database is derived from a combination of Digital Surface Models (DSM) and Digital Elevation/Terrain Models (DEM/DTM), allowing the system to isolate the relative height of vertical structures (e.g., buildings, tree lines) from the underlying terrain profile. For a given candidate position, expected landmark edges are projected into the image to form dynamic 2D search windows. Within these windows, the algorithm samples and aggregates localized intensity gradients (e.g., via Sobel filters) and texture gradients (e.g., via Structured Edge Detection, SED) independently to compute a robust alignment score without sign-cancellation artifacts. By conducting parallel East/West and North/South sweeps over the candidate position space, the system identifies the absolute coordinates that maximize the feature contrast score, enabling drift-free navigation using low-cost hardware in GPS-denied environments.