Real-Time Vehicle Tracking using Magnetic Field Fingerprinting

This paper focuses on developing a vehicle tracking algorithm for vehicle inspection systems. Tracking small vehicle movements in small enclosed systems can be challenging, but offers tremendous benefits to optimize the camera capture moment. Therefore, we focus on locating the vehicle position to ensure that pictures can be taken with equal spacing, limiting the likelihood of missed damages and making it easier to compare different inspections. We propose the usage of sensor fusion between Magnetic Field sensors and Time-of-Flight Lidar sensors and develop a fingerprinting method that accurately tracks vehicle movement through the system. Unlike previous research, we developed a unique approach that does not require offline data collection, prior knowledge of the vehicle, or any sensors to be attached to the vehicles. Using sensor fusion, we establish the fingerprint in real-time to ensure that the system does not require storing and collecting fingerprints in a database. We benchmark the proposed solution against varying vehicle brands, models, and driving behavior. The findings highlight the algorithm’s exceptional performance, which tracks vehicles with less than 0.32 sensor distances in accuracy, or almost 3 centimeters when 64 sensors are used. We show that the algorithm can scale in the number of sensors to improve tracking granularity, with almost no compromise on the algorithm’s accuracy. By evaluating different driving scenarios, we show that the proposed algorithm is robust against different driving speeds, speed changes, and complete stops of the vehicle during the drive-through process. Even when vehicles change driving direction, the proposed system demonstrates the capability to track them effectively.