Segmentation of vehicle signatures from inductive loop detector (ILD) data for real-time traffic monitoring

Inductive loop detectors (ILD) are one of the most popular traffic detectors in use. It works based on the principle of mutual inductance and detects vehicles by measuring the change in inductance due to its presence on top of the sensor. The change in voltage measured is usually called as vehicle signature and is the raw output from the detector system. Proper processing of output data will lead to accurate information about the type and nature of the vehicles movement. This processing needs careful attention, and this is particularly true when it is used under the heterogeneous and lane-less traffic conditions. Overall objective of this work is to identify and segment the signatures of different vehicles from the noisy data, which is the first step for classified counting of vehicles. This work proposes a simple and effective threshold-based approach following a two step procedure for ILD data segmentation. In the first step threshold value for segmentation is determined through a statistical characterization of the historical data corresponding to no-vehicle region. Consequently in the second step, standard deviation is estimated for complete raw data using the mean absolute deviation measure using moving window of data. The developed algorithm was tested, and results showed high accuracy in vehicle count. A guideline for selecting the optimal value of the threshold is also presented.