Location method of airborne plant disease source based on a non-local-interpolation algorithm

Jing Zhang¹Linglan Zhu¹Yifang Wang²Si Chen³Yafei Wang⁴Shifa Li¹Lu Xiao¹Ning Yang^1*

• ¹School of Electrical and Information Engineering, Jiangsu University, Zhenjiang, China
• ²Faculty of Innovation and Design, City University of Macau, Taipa, Macao, Macao, SAR China
• ³Research Center of Fluid Machinery Engineering and Technology, Jiangsu University, Zhenjiang, Jiangsu Province, China
• ⁴School of Agricultural Equipment Engineering, Jiangsu University, Zhenjiang, Jiangsu Province, China

The early stage pathogens of plant diseases have the characteristic of low concentration and difficult detection, which exacerbates the difficulty of tracing the disease, leading to rapid spread and difficulty in effective control. Currently, common plant disease detection techniques such as imaging and spectroscopy can only be applied after the occurrence and manifestation of diseases, and it is difficult to accurately locate the source of disease outbreaks during spore germination or propagation stages. Therefore, this paper proposes a method for locating the source of airborne plant diseases based on the non-local-interpolation algorithm. Firstly, a highly sensitive concentration sensor was designed based on Mie scattering theory to accurately count spores in plant diseases, and a multi-sensor collaborative computing network model was constructed. Secondly, by collecting spore quantity data at different locations, a particle diffusion model is established to summarize the propagation patterns of particles under specific regional conditions. Finally, a non-local-interpolation algorithm coupled with improved power-law equations was designed for precise localization of airborne plant disease sources under different wind and direction conditions. The experimental results in the greenhouse show that the maximum error of light scattering counting does not exceed 10%; Under windless and windy conditions, our method achieved localization accuracies of 94.7% and 92.9%, respectively, with approximately three nodes per square meter. This provides new ideas and insights for early diagnosis and precise localization of plant diseases.