Matched filter for microwave-based detection of dielectric objects in powders
Paper in proceedings, 2017
We present a microwave measurement system for detection of dielectric objects in inhomogeneous powder flows in metal pipes. The system includes a non-intrusive microwave sensor that uses multiple cavity modes to measure the permittivity inside the pipe, and a fast-sampling microwave transmitter and receiver instrument. In this paper, we study a matched filter detection algorithm that takes temporal and inter-channel correlation of the noise into account. The target waveforms used in the filter are constructed from measurements on single object passages. Based on repeated gravity-fall measurements with polyethylene powder at flow rates up to 109 kg/h, we find that wood and plastic objects of 5mm size can be detected with high accuracy, and that the matched filter detector yields minor improvements as compared to a power detector that ignores the temporal correlation.
Maximum likelihood estimation