Weed Identification Using Deep Learning and Image Processing in Vegetable Plantation

Weed identification in vegetable plantation is more challenging than crop weed due to their random plant spacing. So far, little work has been found on identifying weeds plantation. Traditional methods of used be mainly focused directly; however, there a large variation species. This paper proposes new method contrary way, which combines deep learning and image processing technology. Firstly, trained CenterNet model was detect vegetables draw bounding boxes around them. Afterwards, the remaining green objects falling out were considered as weeds. In this focuses only thus avoid handling various Furthermore, strategy can largely reduce size training dataset well complexity detection, thereby enhancing performance accuracy. To extract from background, color index-based segmentation performed utilizing processing. The employed index determined evaluated through Genetic Algorithms (GAs) according Bayesian classification error. During field test, achieved precision 95.6%, recall 95.0%, $F_{1}$ score 0.953, respectively. proposed -19R + 24G -2B ≥ 862 yields high quality with much lower computational cost compared wildly ExG index. These experiment results demonstrate feasibility using for ground-based