Traffic Speed Estimation Device Based on Dual-Axis Magneto-Impedance Sensors for Complex Traffic Conditions

The development of autonomous driving has created new demands for real-time road surveillance. Portable, high-precision, low-power road infrastructure is needed to enable real-time communication of collected vehicle data with passing vehicles. Vehicle recognition and speed estimation are key challenges in this context. We propose a traffic detection device using dual-axis magneto-impedance (MI) sensors installed roadside to detect vehicles in adjacent lanes. Experiments were conducted by driving various vehicle types under a variety of complex road and traffic conditions, analyzing how different measurement conditions and driving patterns affect magnetic signatures. We identified the magnetic signatures most strongly correlated with vehicle recognition and speed estimation. We developed corresponding solutions to address issues as repeated misrecognition and the influence of non-adjacent lanes, which are frequently highlighted in related works. Ultimately, we achieved an overall recognition accuracy of 96.66% in random measurements on complex traffic conditions, reducing recognition errors related to these issues to 3.22%. Additionally, we proposed two speed estimation methods: the time-domain magnetic signature differential method and the dominant frequency method, utilizing multiple sensor nodes and a single node, respectively. Both speed estimation methods were analyzed and compared through controlled experiments, and their performance was evaluated in random measurement experiments. Both methods showed similar estimation accuracy on straight roads, with 100% and 98.8% of results falling within a ±5 km/h error range, respectively. In low-speed curved sections where multiple sensor nodes could not be deployed, we achieved speed estimation using a single sensor by analyzing frequency-domain signals. Finally, under complex traffic conditions, the device demonstrated a minimum Mean Absolute Error (MAE) of only 2.09 km/h for speed estimation.