Proceedings Article | 7 June 2024
KEYWORDS: Cameras, Long wavelength infrared, Near infrared, Short wave infrared radiation, Monte Carlo methods, Panoramic photography, Global Positioning System, Mid-IR, Imaging systems, Mathematical optimization
The ability to accurately ascertain an observer’s position directly from imaged scenery is an important technological capacity, especially given the susceptibility of global positioning system (GPS) signals to interference. Horizon matching has been demonstrated as a method of geolocation in regions with sufficiently rich topography, such as in mountainous terrain. In this method, digital elevation data is processed to yield simulated imagery of horizon features, and this is then matched to real scenery images to back-calculate the imager position. The effectiveness of this method depends on several attributes of the imaging system, including its resolution, range performance, contrast-to-noise ratio, and others. As such, cameras operating in different spectral bands – visible, near-infrared, mid-wave infrared, and longwave infrared – offer distinct advantages and drawbacks for this task.
Here, we demonstrate geopositioning via horizon matching with a vehicle-integrated, gimbal mounted multi-band imaging platform that facilitates this functionality with visible, NIR, SWIR, MWIR and LWIR imagers. Using FOV-matched cameras, performance of the method was evaluated for single images and for panoramic imagery. We present a statistical method of evaluating this and similar approaches by executing Monte Carlo simulations. We then exploit this approach to find multiple solutions and down select these based on an error/fit metric. By exploiting the fit metric, we demonstrate accuracies for each band of 100m or better, which compares well with comparable geolocation approaches. We compare performance of the different bands and generally assert that the accuracy of each band depends on how well the fit metric and position error correlate for that band.
