Estimating Vehicle Length Using FMCW Radar: A Micro-Doppler Signature Analysis Approach

Effective traffic management and safety in Intelligent Transportation Systems (ITS) significantly benefit from various data points, with vehicle length being one important aspect. Current radar-based methods often fall short of demonstrating the capability to assess vehicle lengths in diverse traffic conditions. To bridge this gap, we introduce a novel algorithm employing time-frequency analysis of radar signals at a fixed range bin, enabling the estimation of the time vehicles of varying lengths and speeds take to cross this range bin. Assuming known vehicle velocities, this method allows the estimation of vehicle length. We develop a signal model of the vehicle's bulk motion extracted from its micro-Doppler signature. We then introduce three duration estimation methods: Maximum Likelihood Estimation (MLE) assuming Gaussian noise and two heuristic approaches utilising threshold calculations. Through testing on synthetic data and further validation using both simulated and real radar data, we evaluate the accuracy and computational efficiency of these methods. The results underscore the potential of our time-frequency analysis approach for vehicle length estimation, highlighting the prospect of non-invasive radar-based data collection systems in ITS.