Techniques are presented for
processing LIDAR data from a sensor mounted for example on a gantry over a roadway. The
processing method disclosed classifies blank data points into "active blank" (which is supposed to be caused by a nonreflective object) or "inactive blank" (which is supposed to be a caused by a faulty sensor). By classifying blank data points in this way,
processing is better able to distinguish, for example, two vehicles following close behind from a
single vehicle having a poorly-reflective roof. Data points are classified at the same time as downsampling with little computational expense by taking
advantage of the higher resolution data available before downsampling. The result of the combined downsampling and classification is a
time series for each "pixel" where there will be far fewer pixels than there are raw point measurements. A
LIDAR sensor (10) is typically mounted on a gantry (12) over a roadway (14). Typical arrangements are on a
toll road having multiple
LIDAR sensors across the gantry to detect vehicles at any point across the
full width of the road. Scan lines (16a, 16b, 16c) typically extend substantially across the width of a lane, at three positions along the length of the lane. Moving averages associated with
time series of sensor data are maintained wherein a first and a second
moving average are measures of how often a
data point in the
time series has been classified as blank and non-blank, respectively.