Detection of Helminthosporium Leaf Blotch disease based on UAV RGB imagery

Authors: HUASHENG, HUANG - South China Agricultural University; LAN, YUBIN - South China Agricultural University; DENG, JIZHONG - South China Agricultural University; YANG, AQING - South China Agricultural University; Zhang, Huihui; ZHANG, LEI - South China Agricultural University

Submitted to: Applied Sciences
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/4/2019
Publication Date: 2/8/2019
Citation: Huasheng, H., Lan, Y., Deng, J., Yang, A., Zhang, H., Zhang, L. 2019. Detection of Helminthosporium Leaf Blotch disease based on UAV RGB imagery. Applied Sciences. 9(3):558. https://doi.org/10.3390/app9030558.
DOI: https://doi.org/10.3390/app9030558

Interpretive Summary: In order to evaluate a serious disease of wheat cultivation, Helminthosporium Leaf Blotch, RGB imagery taken by an unmanned aerial vehicle was used to determine severity degree of the disease in field. Different image classification methods were tested and verified by ground-truth measurements. The results show that neural networks method performs the best among other methods on discriminating infested area from non-infested area and precisely distinguishing RGB images from different disease categories.

Technical Abstract: Helminthosporium Leaf Blotch (HLB) is a serious disease of wheat cultivation causing yield reduction in the world. HLB disease is subject to site-specific disease control including infected vegetation removal and pesticide spraying. In this case, automatic disease detection over the whole field can provide important information for growers. The objective of this paper was to evaluate the potential of UAV remote sensing technology for HLB detection. UAV RGB imagery acquisition and ground investigation were conducted in Central China in April 2017. Four disease categories (normal, light, medium, and heavy) were established based on different disease severity degrees. A Convolutional Neural Network (CNN) was proposed for HLB disease classification. The network architecture, data preprocess and hyper-parameter optimization was introduced. The classification accuracy of our CNN architecture was up to 91.8%, which outperformed other methods. Experimental results showed that the HLB infected and healthy areas can be precisely discriminated based on UAV remote sensing data, indicating that UAV remote sensing can be proposed as an efficient tool for HLB disease detection.