Identification of Weed Species, Kikuyu Grass (Pennisetums) in Crops Using Machine Vision

Abstract: This article shows the results of applying image processing to estimate the presence of weed species Kikuyu grass in crops. A study was conducted in field (in site), within more than 30 locations in the area Cundinamarca - Boyaca in Colombia; an analysis of classical methods of image classification is performed to determine the best way to identify arvense species. It is concluded that at present, the benefits presented by analysis with neural networks, allow ensure a correct classification despite the difficulties that arise with environmental variables. This document is to disseminate the results obtained from the analysis of images taken in the field in order to estimate automatically the presence of weed species, Kikuyu grass, by using computer vision and traditional approaches of classification.