A Monocular Vision Ranging Method Related to Neural Networks

This paper proposes a neural network-based monocular vision ranging method for the situation of large camera calibration and distance variation in monocular vision ranging. The imaging size of the corresponding target under different distances of the same camera is recorded, and the distance variation is recorded according to the change of the imaging size, and a dataset is made accordingly. The ranging network model is established by referring to the neural network and trained on the dataset. The yolov7 target detection network is combined with the ranging network, and real-time ranging is performed according to the real-time target frame output by the target detection network. The monocular vision ranging method in this paper avoids the complex calibration of the camera’s internal parameters, has a simple structure, fast operation speed, low cost and easy implementation. The training results of this paper’s ranging method show that the average distance error is 0.1m within 20m range, which meets the accuracy requirements and verifies the feasibility and effectiveness of this method by real-time ranging experiment.