Object Distance Determination by Means of Technical Vision

The purpose of this paper is to evaluate the effectiveness and accuracy of an integrated approach for object distance estimation based on a camera model and machine learning libraries in the context of advanced driver assistance systems (ADAS) and autonomous driving. This paper examines how camera model parameters (sensor resolution, focal length, field of view, and aperture characteristics) influence the quality of input data for algorithms implemented using machine learning libraries and, consequently, the overall distance estimation accuracy. Particular attention is paid to the practical application of modern machine learning algorithms, including deep neural networks, to solve the problem of monocular depth estimation from a single image. This study aims to identify the potential dependence of distance estimation accuracy on object distance. In particular, it will analyze how distance increasing can lead to a disproportionate increase in estimation error, which will help identify potential limitations of current implementations of machine learning algorithms when working with distant objects. Factors, contributing measurement accuracy reduction at long distances will be considered, such as increased distortion during image capture and processing, as well as issues of adapting algorithms to the characteristics of data from remote objects.