A monitoring method for pine wilt disease infected discolored and deceased pine trees removal information based on DDPTnet network and Bi-temporal UAV imagery

Abstract

Pine Wilt Disease (PWD) is spreading globally, and the failure to timely detect and clear Discolored and Deceased Pine Trees (DDPT) carrying the pine wilt will further exacerbate the spread of PWD in affected regions. To address the necessity of recognizing the DDPT removal information, this research constructs a highly robust Discolored and Deceased Pine Trees network (DDPTnet) for DDPT detection. Additionally, by integrating detection results from DDPTnet with dual-period UAV Digital Orthophoto Models (DOM), Digital Surface Models (DSM), and a scene classification algorithm (DDPTnet-cls) based on DDPT, the current gap in DDPT removal information monitoring has been addressed. In the three monitoring areas, DDPTnet achieved an average F1 score (F1) of 87.31 % for DDPT detection. By integrating the detection results of DDPTnet with dual-period DSM, the average F1 score (F1-la) for logged area extraction was 94.59 %, and the average F1 score (F1-l) for identifying logged DDPT was 82.08 %. The DDPTnet_cls classification method, after fine-tuning, achieved an average classification accuracy (Acc) of 77.91 %. Finally, based on the above results, The “DDPT Removal Information Thematic Map” were produced. These outcomes can provide objective and effective decision support for the prevention and control of PWD outbreaks.