Weed Recognition of the Machine Vision Based Weeding Robot

Abstract

A new segmentation algorithm of image was proposed to discriminate weeds from background by using color information in RGB space.Firstly,the plant-soil-residue image was analyzed and calculated pixel by pixel,pixel was considered as weed when GR and GB and its value was set to 1;otherwise,the pixel was considered as background and its value was set to 0.Afterwards,8-neighborhood method was applied to eliminate the isolated points and the locations of weed regions were obtained.The software of weed recognition for weeding robot was developed by means of Visual C++ and a model of robot structure was designed.Experiments show that this algorithm is robust to the variation of the natural light and takes less time in image processing.The mechanical arm of weeding robot could be precisely positioned and carry out weeding activities.