False Detection Filtering Method for Magnetic Sensor-Based Vehicle Detection Systems

Abstract

Magnetometer-based vehicle detection systems offer many advantages compared to other technologies. These systems can effectively detect vehicle presence, but vehicles with high metallic content passing in the neighboring lane can cause false detections. In this work, a new method is proposed for the filtering of these false detections. Filtering is based on different rules constructed using various data types extracted from the signals. All data types were tested separately to find the parameters with most influence, and later these parameters were combined in more complex rules to achieve more accurate results. The optimization of the parameter values was realized using genetic algorithms. The obtained results show that using properly constructed rules, 97% of false detections can be filtered with losing only nearly 0.3 % of good detections, and even a single parameter can be sufficient to reliably filter the false detections.