Real-time Detection of Between-row Weeds Using Machine Vision

Abstract

A system used machine vision to detect between-row weeds was developed and tested in laboratory with outdoor lighting conditions. A software system named �Between-row Weeds Detection System� was developed to process the images. The proposed algorithms used color information to discriminate between plants and background, whilst novel analysis techniques were applied to distinguish between crop and between-row weeds by use of the information of plants’ location within the field. Firstly, the excessive green algorithm was adopted to gray the source images. Secondly, the gray-level image was transformed to binary image by use of the algorithm of the maximum variance optimal threshold selection. Finally, crops and weed were segmented by use of the seed-fill algorithm. It was indicated that the DWB system had the superiority in real-time.