A Robotic System Employing Deep Learning for Visual Recognition and Detection of Weeds in Grasslands

Abstract

In this paper, we describe the vision system of a robot prototype that operates in dairy farm grasslands and detects the presence of the harmful Broad-leaved dock (Rumex obtusifolius L.). Image data were collected using the prototype robot from 3 fields in 2 different countries under real conditions. The proposed recognition and detection system is using solely 2D visual input and is based on state-of-the-art Convolutional Neural Networks (CNNs), which were used for high level feature extraction in combination with various examined classifiers. Gathered data were used to experimentally show that the proposed system yields state-of-the-art detection and recognition performance, while being able to keep low false-positive rates under challenging operation conditions.