Poster Presentation Science Protecting Plant Health 2017

Phenotyping crown rot in wheat using imaging technologies (#228)

Cassandra Percy 1 , Tai Nguyen-Ky 2
  1. University of Southern Queensland, Centre for Crop Health, Toowoomba, Qld, Australia
  2. University of Southern Queensland, Computational Engineering and Science Research Centre, Toowoomba, Qld, Australia

Crown rot caused predominantly by Fusarium pseudograminearum significantly limits yield in Australian winter cereals. Breeding and pre-breeding for crown rot resistance and tolerance in wheat has been a national focus. While strong levels of partial resistance and tolerance exist, the complex nature of inheritance and the difficulties associated with phenotyping this disease, has resulted in slow progress of the availability of commercial cultivars with high levels of resistance and tolerance to crown rot. When screening for crown rot in field trials currently multiple tillers are rated on approximately 20 plants randomly sampled from within each experimental plot at maturity. This process is laborious and often difficult to achieve consistent results between seasons. Alternative methods of measuring the host response to Fusarium pseudograminearum infection are being tested in field trials in the northern region. Inoculated and non-inoculated plots of 12 wheat genotypes ranging in susceptibility to crown rot were assessed in the 2016 season.

A new method for data collection was designed using multispectral, thermal and visible cameras. Images were analysed by the wavelet and filter band methods to extract the differences between plus and minus inoculated plots under field conditions. A new index (PWDI: Phenotyping Wheat Disease Index) is proposed to detect the differences between the plus and minus inoculated plots; and rank 12 wheat genotypes into three classes such as most resistant, middle, and susceptible classes, with the accuracy from 80% to 90% when comparing with the expected crown rot ranking. Artificial intelligence and machine learning models will be developed for integration into the dataset to validate results against traditional disease assessments performed at maturity. Both researchers and breeding companies would greatly benefit from such technologies in Australia and internationally. Significant benefits would be further provided to growers to allow appropriate management strategies to be implemented on farm.