Abstract

Automated robotic weeding of grassland will improve the productivity of dairy and sheep farms
7 while helping to conserve their environments. Previous studies have reported results of machine
8 vision methods to separate grass from grassland weeds but each use their own datasets and
9 report only performance of their own algorithm, making it impossible to compare them. A
10 definitive, large-scale independent study is presented of all major known grassland weed detec-
11 tion methods evaluated on a new standardised data set under a wider range of environment
12 conditions. This allows for a fair, unbiased, independent and statistically significant comparison
13 of these and future methods for the first time. We test features including linear binary pat-
14 terns, BRISK, Fourier and Watershed; and classifiers including support vector machines, linear
15 discriminants, nearest neighbour, and meta-classifier combinations. The most accurate method
16 is found to use linear binary patterns together with a support vector machine