Machine vision App for automated cotton insect counting: initial development and first results

Long, Derek and Grundy, Paul and McCarthy, Alison (2019) Machine vision App for automated cotton insect counting: initial development and first results. In: Australian Association of Cotton Scientists 2019 Australian Cotton Research Conference (AACS 2019), 28-30 Oct, 2019, Armidale, Australia.

[img]
Preview
Text (Presentation slides)
AACS Presentation Derek Long - 20191024.pdf

Download (1MB) | Preview
[img]
Preview
Text (Accepted Version)
2019 AACS abstract draft_20190913_AM.pdf

Download (74kB) | Preview

Abstract

Silverleaf whitefly, cotton aphids and spider mites cause cotton yield loss through plant feeding and lint contamination from waste secretions. Agronomists determine if control action is required from weekly monitoring of changes in pest counts. This manual sampling is labour-intensive as hundreds of leaves are sampled at 20-30 leaves per 25 hectares of cotton and examined by eye for the presence and density of each pest. Machine vision has potential to automate the pest counting on each leaf using infield cameras and image analysis software. There is potential to transfer the machine vision algorithms to a mobile device App for agronomist to enable real-time photo capture and analysis for pest counting. This App would standardise pest counting between different observers, improve chemical control decisions, provide a convenient method for logging and viewing data for each field, and inform Area Wide Management from silverleaf whitefly nymph counts.

Data collection and software development have been conducted to develop the image analysis algorithms for detecting silverleaf whitefly nymphs. A dataset of training images was captured from glasshouses cultures of whitefly and commercial cotton farms in southern Queensland with three smartphone models. Image analysis algorithms were developed to extract numbers of silverleaf whitefly nymphs (3rd/4th instar) on each leaf. Two image analysis methods were implemented: a segmentation-based approach, and a machine learning approach. The segmentation-based approach and machine learning approach detected silverleaf whitely nymphs with up to 67% and 79% accuracy, respectively. The image analysis algorithms will be refined through parameter optimisation and incorporated into an App that will be evaluated by agronomists in the 2019/20 season. The image analysis algorithms will be extended to cotton aphids and mites as all three insects can occur simultaneously.


Statistics for USQ ePrint 37860
Statistics for this ePrint Item
Item Type: Conference or Workshop Item (Commonwealth Reporting Category E) (Speech)
Refereed: No
Item Status: Live Archive
Faculty/School / Institute/Centre: Current - Institute for Advanced Engineering and Space Sciences - Centre for Agricultural Engineering (1 Aug 2018 -)
Faculty/School / Institute/Centre: Current - Institute for Advanced Engineering and Space Sciences - Centre for Agricultural Engineering (1 Aug 2018 -)
Date Deposited: 14 Aug 2020 05:30
Last Modified: 08 Oct 2020 06:18
Uncontrolled Keywords: machine vision; pest counts; cotton insects; silverleaf whitefly; cotton aphids; spider mites; cotton yield loss
Fields of Research (2008): 09 Engineering > 0906 Electrical and Electronic Engineering > 090602 Control Systems, Robotics and Automation
08 Information and Computing Sciences > 0801 Artificial Intelligence and Image Processing > 080104 Computer Vision
07 Agricultural and Veterinary Sciences > 0703 Crop and Pasture Production > 070308 Crop and Pasture Protection (Pests, Diseases and Weeds)
07 Agricultural and Veterinary Sciences > 0703 Crop and Pasture Production > 070302 Agronomy
Socio-Economic Objectives (2008): D Environment > 96 Environment > 9609 Land and Water Management > 960905 Farmland, Arable Cropland and Permanent Cropland Water Management
URI: http://eprints.usq.edu.au/id/eprint/37860

Actions (login required)

View Item Archive Repository Staff Only