Child perception-based plant species identification

Automation of mechanically pulling the weeds out of the crop row not only copes well with the new European high environmental standards but also removes the high cost of mere conventional alternative as high hand-weeding. The objective of this research work is to propose a similar distinguishing methodology of a weeding labour when discriminating the weeds species to choose and remove the undesired ones. The method is governed by a systematic analysis carried out on recognition method of an immaturely trained human brain. In other words, a number of children, who have never seen a maize agricultural field, were examined while recognizing a maize pattern using at most five sample images. The proposed method works mainly based on morphological operators for extraction of fundamental plant features in the image. The advantage of the proposed method is producing similar results to human labour which is approximate identification. This final decision was made by a fuzzy classifier which generate the degree of membership to either of weed or crop plant groups. Unlike the very popular research trend for object classification in the literature, our proposed methodology neither requires huge sample of images nor a high capacity processor. The accuracy of maize plant discrimination from typical Mediterranean weeds under extreme July sun illumination was observed as 95% for a scene of a single plant and 85% for a scene containing multiple objects.