Real-Time Weed Detection using Machine Learning and Stereo-Vision

In modern days, weed identification / plant detection in plants is more difficult. There has been little work so far to identify weeds while planting vegetables. Traditional approaches for the identification of agricultural weed were primarily directed at directly identifying weed; nevertheless, the variations in weed species are significant.

This study presents a novel technique, which merges deep learning with imaging technology, as opposed to this method. First, the YOLO v2 model was trained to identify and draw boundary boxes for plants surrounding it. Then, the remaining green items fell from the border boxes like weeds. In this approach, just the crops are identified and other weed species are thereby prevented from being handled.

These experiment results demonstrate the feasibility of using the proposed method for the ground-based plant identification in vegetable plantation.

Keywords: Plant identification, deep learning, image processing, deep learning, YOLO v2.