The invention relates to a method for determining a remote sensing optimal diagnosis time period of the harm of hyphantria cunea. The method comprises the following steps of (1) determining a diagnosis threshold of the harm of the hyphantria cunea based on measured data and a mutation theory; (2) obtaining a multi-time-point remote sensing image based on the available multi-time-point remote sensing image; and obtaining the remote sensing response characteristics including the leaf area index (LAI), the characteristic spectrum index (CSI), the normalized difference mountain vegetation index (NDMVI), the global vegetation humidity index (GVMI) and the like of bamboo poison moth hazard detection respectively, establishing the time sequence data of all the indexes according to the characteristics, and extracting the bamboo poison moth hazard level; and (3) counting insect pest risk evaluation index values, drawing a curve graph, and respectively determining the optimal monitoring and early warning time period of the insect pests. The method can provide an important basis for remote sensing detection and accurate prevention and control of the hyphantria cunea insect pests.