Home /Research /Recognition of Dandelion Weed via Computer Vision for a Weed Removal Robot
PERCEPTION

Recognition of Dandelion Weed via Computer Vision for a Weed Removal Robot

Ibrahim Babiker, Wenfang Xie, Guangyi Chen

Year
2019
Citations
8

Abstract

This paper introduces novel methods for distinguishing dandelion weed leaves from grass blades and identifying the centroids of dandelion weeds by means of a webcam. Firstly, a method of isolating the colour green using a Hue-Saturation-Value (HSV) scheme is proposed. Next, a technique for filtering the high frequency grass blades is developed to isolate individual low frequency of weed leaves. Subsequently, two novel methods: the lengthwise intersection method and mean centroid method are proposed to detect the location of the weed center based on the individual weed leaves. The detected location of the dandelion weed centroid has been used as the desired destination of autonomous driving system of a weed removal robot for successful dandelion removal.

Keywords

DandelionWeedHueCentroidRobotComputer scienceArtificial intelligenceComputer visionDroneAgronomy

Related papers

Browse all PERCEPTION papers