Intelligent Vehicle Speed Estimation System Integrating with Deep Learning from the Perspective of Unmanned Aerial Vehicles (UAVs)

Abstract

This study designs an intelligent vehicle speed estimation algorithm relying on the perspective of unmanned aerial vehicles (UAVs)
and deep learning technology to address the limitations encountered by conventional Ground-Based Speed Measurement Methods in complex traffic environment. This system acquires traffic surveillance video footage from aerial birds-eye view data captured by UAVs and then
employs the deep learning framework to realize the entire automated processing flow of vehicle detection, tracking, and speed estimation.
Experiments have shown that in urban roads and highways, this newly proposed method exhibits significantly higher speed measurement accuracy compared to conventional approaches, with a mean absolute error of less than 3km/h, and has better environmental adaptability and
anti-interference ability, providing reliable empirical evidence for the fields of intelligent traffic management and accident analysis.