Optical Sensing of Weed Infestations at Harvest.

Judit Barroso,Dan Long,John Mccallum

doi:10.3390/s17102381

Judit Barroso, Dan Long + Show 1 more

Open Access

https://doi.org/10.3390/s17102381

Copy DOI

Abstract

Kochia (Kochia scoparia L.), Russian thistle (Salsola tragus L.), and prickly lettuce (Lactuca serriola L.) are economically important weeds infesting dryland wheat (Triticum aestivum L.) production systems in the western United States. Those weeds produce most of their seeds post-harvest. The objectives of this study were to determine the ability of an optical sensor, installed for on-the-go measurement of grain protein concentration, to detect the presence of green plant matter in flowing grain and assess the potential usefulness of this information for mapping weeds at harvest. Spectra of the grain stream were recorded continuously at a rate of 0.33 Hz during harvest of two spring wheat fields of 1.9 and 5.4 ha. All readings were georeferenced using a Global Positioning System (GPS) receiver with 1 m positional accuracy. Chlorophyll of green plant matter was detectable in the red (638–710 nm) waveband. Maps of the chlorophyll signal from both fields showed an overall agreement of 78.1% with reference maps, one constructed prior to harvest and the other at harvest time, both based on visual evaluations of the three green weed species conducted by experts. Information on weed distributions at harvest may be useful for controlling post-harvest using variable rate technology for herbicide applications.

Highlights

Weeds are often distributed in farm fields as many irregular patches [1,2] that are relatively stable from one season to the [3,4,5,6]
Before growers can benefit from site-specific weed management (SSWM), they must know where weeds are located within their fields
Modern spray delivery systems can be controlled either (i) indirectly from a weed map derived from remotely sensed information that had been acquired at an earlier time [9], or (ii) directly in response to the outputs from a sensing system mounted on the treatment vehicle [10,11]. These technologies rely upon machine vision systems capable of detecting green plants and discriminating weeds from the crop

Summary

Introduction

Weeds are often distributed in farm fields as many irregular patches [1,2] that are relatively stable from one season to the [3,4,5,6]. Modern spray delivery systems can be controlled either (i) indirectly from a weed map derived from remotely sensed information that had been acquired at an earlier time [9], or (ii) directly in response to the outputs from a sensing system mounted on the treatment vehicle [10,11]. These technologies rely upon machine vision systems capable of detecting green plants and discriminating weeds from the crop

Objectives

Methods

Results

Conclusion