Application of spectral reflectance for increasing plant discrimination speed in precision agriculture

Increasing the speed of weed/crop discrimination sensor engines is an increasingly challenging research area in precision agriculture (PA). Data collection, modelling, and real-time operation are currently the major challenges for accurate plant classification and effective weed control. In the current study, a new low-resolution spectral reflectance sensing is proposed for data collection and applied in conjunction with state-of-art convolutional neural network (CNN) algorithm for real-time weed detection. Experimental results demonstrate that the speed of the algorithm is ten times faster than typical spatial imaging based counterparts, while its discrimination accuracy is almost the same.