Agro-Weed Identification via Deep and Shallow Learning with Texture Features

Weeds are one of the most significant elements affecting agricultural production. The waste and pollution in farm caused by spraying chemical herbicides are becoming increasingly evident. If weeds are identified from crops, there will be a tremendous improvement in the agricultural production level, thereby achieving precise spraying only for weeds. This paper gives its contribution to weed identification in two publicly available datasets. The weeds are segmented and their texture features are extracted. The obtained texture features are classified using deep learning and shallow learning classifiers. The experiments substantially proved that the proposed method with deep learning classifier achieves approximately 4% higher than shallow learning classifiers. It also achieves 0.02% higher accuracy than recent methods on the compared dataset.