Research on Monocular Vision Target Detection and Localization Under Non-Orthographic Conditions

Abstract

Aiming at the target localization problem in monocular vision, this paper proposes a nonlinear target localization method under non-orthoptic conditions. First, an image is captured using a monocular camera and a network framework based on YOLOv7 is constructed to detect the target in the image. Then, based on the principle of aperture imaging, the imaging models of orthophoto and non-orthophoto are established, and the nonlinear relationship between pixel coordinates and world coordinates in the image is deduced, so as to calculate the relative position coordinates of the target. In order to verify the validity of this nonlinear imaging model, we choose different shapes and numbers of ship targets for verification. The experimental results show that the target localization accuracy of monocular vision can reach more than 90% under non-orthoptic conditions.