Magnetic sensing has been widely investigated for detecting ferromagnetic objects in outdoor environments and is commonly used in traffic monitoring and intelligent transportation applications. Low-cost microelectromechanical systems (MEMS) magnetometers provide an attractive sensing solution owing to their compact size, low power consumption, and ease of integration into embedded platforms. However, measurements obtained from such sensors are often affected by noise, baseline drift, and slowly varying geomagnetic disturbances, which may degrade vehicle detection reliability. In this paper, an experimental evaluation of magnetic vehicle detection using a low-cost LIS3MDL triaxial MEMS magnetometer is presented. A lightweight signal processing method based on moving-average baseline compensation is applied to suppress slowly varying geomagnetic variations while preserving transient disturbances generated by passing vehicles. Roadside experiments were conducted at a sampling frequency of 200 Hz. The results show that the proposed processing method improves measurement stability and increases the signal-to-noise ratio by more than 11 dB. Repeated experiments demonstrate reliable vehicle detection with a rate exceeding 96%.