EVALUATION OF DEEP NEURAL NETWORKS FOR CAMERA GEOMETRIC CHARACTERIZATION IN IMAGE-BASED NAVIGATION

The geometric characterization of cameras requires laboratory methods during construction and field methods during operation, where measuring multiple points in images is essential to ensure redundancy and reliable results. However, conventional characterization is labor-intensive and demands specific data collection. Recently, efforts have shifted towards deep learning-based solutions to replace repetitive manual tasks. Various learning strategies, neural networks, geometric priors, and datasets have been investigated. This paper focuses on analyzing the performance of some of these AI-based camera characterization techniques, highlighting their advantages, limitations, and potential applications to enhance image-based navigation results.