A real-time plant discrimination system utilising discrete reflectance spectroscopy

AbstractAn advanced, proof-of-concept real-time plant discrimination system is presented that employs two visible (red) laser diodes (635nm, 685nm) and one near-infrared (NIR) laser diode (785nm). The lasers sequentially illuminate the target ground area and a linear sensor array measures the intensities of the reflected laser beams. The spectral reflectance measurements are then processed by an embedded microcontroller running a discrimination algorithm based on dual Normalised Difference Vegetation Indices (NDVI). Pre-determined plant spectral signatures are used to define unique regions-of-classification for use by the discrimination algorithm. Measured aggregated NDVI values that fall within a region-of-classification (RoC) representing an unwanted plant generate a spray control signal that activates an external spray module, thus allowing for a targeted spraying operation. Dynamic outdoor evaluation of the advanced, proof-of-concept real-time plant discrimination system, using three different plant species and control data determined under static laboratory conditions, shows that the system can perform green-from-green plant detection and accomplish practical discrimination for a vehicle speed of 3km/h