Detection of pests and diseases in vegetable crops using hyperspectral sensing: a comparison of reflectance data for different sets of symptoms

Apan, Armando and Datt, B. and Kelly, Rob (2005) Detection of pests and diseases in vegetable crops using hyperspectral sensing: a comparison of reflectance data for different sets of symptoms. In: 2005 Spatial Sciences Institute Biennial Conference: Spatial Intelligence, Innovation and Praxis (SSC2005), 12-16 Sep 2005, Melbourne, Australia.

[img]
Preview
PDF (Published Version)
Apan_etal_SSIMelb2005_2.pdf

Download (549Kb)

Abstract

The aim of the study was to examine the potential of hyperspectral sensing to detect the incidence of pests and diseases in vegetable crops. The specific objectives were: a) to test if symptoms of pests and diseases of vegetable crops can be detected by hyperspectral sensing, b) to determine the best spectral bands relevant to pest and disease detection, and c)to compare the spectral responses obtained from the symptoms of two different pest and disease. Using a handheld spectrometer, sample measurements of diseased/infested and healthy leaves were collected separately from tomato and eggplant crops. The tomato crops were affected by a fungal 'early blight' disease (Alternaria solani), with symptoms characterised by a yellowing or chlorosis of leaves. Conversely, the eggplants exhibited skeletal interveinal damage ('holes') on leaves, caused by the 28-spotted ladybird (Epilachna vigintioctopunctata). To overcome the problems of using traditional regression techniques, the Partial Least Squares (PLS) regression technique was used. The cross-validated results showed that the incidence of disease on tomato and pest on eggplant could be predicted with 82% accuracy. Different sets of symptoms (i.e. chlorosis vs. loss of leaf area) provided different sets of optimum number of principal components and significant predictor variables. The most significant spectral bands for the tomato disease prediction corresponded to the reflectance red-edge (690nm-720nm), as well as the visible region (400nm-700nm), and part of near-infrared wavelengths (735nm-1142nm). For the eggplant, the near-infrared region (particularly the bands 732nm-829nm) was identified by the regression model to be as equally significant as the red-edge (694nm-716nm) in the prediction. However, the inclusion of the shortwave infrared bands (1590nm-1766nm) as significant variables in the eggplant regression model has indicated the contributing role of other factors. Results such as these confirmed the utility of hyperspectral data to diagnose pests and diseases of vegetables that can improve detection speed and provide opportunity for non-destructive sampling.


Statistics for USQ ePrint 553
Statistics for this ePrint Item
Item Type: Conference or Workshop Item (Commonwealth Reporting Category E) (Paper)
Refereed: Yes
Item Status: Live Archive
Depositing User: epEditor USQ
Faculty / Department / School: Historic - Faculty of Engineering and Surveying - Department of Surveying and Land Information
Date Deposited: 11 Oct 2007 00:24
Last Modified: 02 Jul 2013 22:32
Uncontrolled Keywords: hyperspectral sensing, pests and diseases, vegetables, PLS regression
Fields of Research (FOR2008): 09 Engineering > 0909 Geomatic Engineering > 090905 Photogrammetry and Remote Sensing
07 Agricultural and Veterinary Sciences > 0706 Horticultural Production > 070603 Horticultural Crop Protection (Pests, Diseases and Weeds)
URI: http://eprints.usq.edu.au/id/eprint/553

Actions (login required)

View Item Archive Repository Staff Only