Abstract: Abstract: Panicle blast is one of the most serious diseases in the rice production process. Because of its rapid transmission, difficult prevention and control, and strong destruction, it has the greatest impact on yield. The Unmanned Aerial Vehicle (UAV) hyperspectral remote sensing can not only realize the accurate monitoring of diseases and insect pests in a larger range and with higher spatial resolution but also promoted the application of the hyperspectral theory of rice blast. In this study, a field experiment on rice blast was conducted in Yongji, Jilin from April to September 2019. Jiyujing (code name: ji90-g4) was selected as the experimental variety. In order to maximally stimulate the natural onset of rice blast, Mongolian rice inoculated with Pyricularia oryzae was used as the inducing plant to infect healthy rice. The UAV hyperspectral remote sensing platform (UAV: DJI M600 Pro; Imaging spectrometer: Cubert S185) was used to collect hyperspectral image data of the entire experimental area. At the same time, plant protection experts were invited on the ground to classify the 30 sampling points according to the health, mild, moderate, and severe simultaneous disease severity. ENVI 5.3 was used for the geometric correction of the image. According to the GPS positioning points determined by ground sampling, each corresponding sampling area was extracted into a Region Of Interest (ROI) according to the 30 × 30(pixels) rectangular area, and the corresponding ground spatial resolution was 0.9 m × 0.9 m. The spectral data of all pixels in each ROI were averaged, and different spectral preprocessing and mathematical transformations were carried out as the input of the model. The samples were randomly divided into the modeling set and verification set according to the ratio of 2:1, and then Random Forest (RF) model was used for modeling. RF model avoided the overfitting problem when there were few sampling points. The overall reflectance of rice spectral curve with different panicle blast grades showed a downward trend, and change at 670 nm was the strongest correlation with the change of rice blast grade; Continuum Removal (CR) treatment further improved the spectral difference of target objects, and there were three obvious inflection points of reflection and absorption between 466 and 750 nm, with 498, 534, and 666 nm as the center points. Based on a variety of the Combination of Vegetation Indices (CVIs) which reflected the changes in rice physiological parameters, the best results were obtained in RF modeling. The highest accuracy of modeling was 90% and the Kappa coefficient was 0.86. At the same time, it explained the changes in plant physiological parameters such as chlorophyll, carotenoid, nitrogen content, cell structure, red edge, and so on. The relationship between the spectrum of panicle blast and the variation of plant parameters were established. The Principal Component Analysis (PCA) method for data processing and modeling, which was often used in previous studies on rice blast spectrum, did not achieve ideal results in this study, which might be due to the difference between field and laboratory environmental conditions. How to reduce environmental noise and extract more effective disease information would be the key to the next step of research. Compared with the indoor spectral theory research of rice blast, the UAV hyperspectral remote sensing monitoring experiment is the theoretical research, and it is the key link to the field rice blast quantitative remote sensing monitoring and early warning grading, filling the gap between the theory and practice of rice blast monitoring.