Vehicle Axle Detection from Under-Sampled Signal through Compressed-Sensing-Based Signal Recovery

Abstract

In traffic data collection, sampling design should satisfy the requirements of identifying prominent pulses corresponding to vehicle axle passage. Insufficient measurement leads to signal distortion and attenuation, reducing the quality of signal pulses. This study exploits the value of under-sampled data by applying compressed sensing (CS) methods to recover signal components that are critical for vehicle axle detection. Two CS methods are investigated in this study to recover the strain signal pulses from inside-pavement instrumented sensors at high-speed traversals. The CS methods successfully recovered the signal pulses from all axles of the truck used for testing. A comparison of the measured axle distances with the reference measurements validated the effectiveness of signal recovery methods. Therefore, the CS methods have the potential of reducing the cost, energy consumption, and data storage space, and improving the data transmission efficiency in practical implementations by enabling sampling devices designed for static measurements to achieve dynamic measurements.