Image acquisition for weed detection and identification by digital image analysis

Publication Type

Contribution to conference

Authors

M. Sökefeld and R. Gerhards and H. Oebel and R.-D. Therburg

Year of publication

2007

Published in

Precision agriculture '07

Editor

J.V. Stafford

Pubisher

Wageningen Academic Publishers , Netherlands

Band/Volume

6/

ISBN / ISSN / eISSN

978-90-8686-024-1

Page (from - to)

523-529

Conference name

6th European Conference on Precision Agriculture (ECPA)

Site-specific weed control requires information on weed infestation and distribution within agricultural fields. For the application of selective herbicides, detailed information on the distribution of all weed species present is needed. Digital image analysis is described as a tool for weed identification. For these purposes, images of the field surface were taken. A high resolution bi- spectral camera system was developed which simultaneously takes two pixel congruent images in the visible wave band at 610 - 690 nm and in the near infrared wave band > 700 nm. This camera system removed the effect of elements that disturb the images such as organic materials like straw and root residues, stones and shade. Only green plants were represented in the images. This reduced the time for pre-processing of the images and eliminated misclassification due to disturbance. An automatic and dynamic camera control for changing light conditions by atmospheric conditions was developed. Therefore, image acquisition was independent of artificial light and shading. For tests under field conditions, the bi-spectral camera system was mounted on a vehicle equipped with GPS and an on-board computer.