Using UAV Multispectral Remote Sensing with Appropriate Spatial Resolution and Machine Learning to Monitor Wheat Scab

This study took the wheat grown in experimental area of Jiangsu Academy Agricultural Sciences as research object and used unmanned aerial vehicle (UAV) to carry Rededge-MX multispectral camera obtain scab image with different spatial resolutions (1.44 cm, 2.11 3.47 4.96 6.34 7.67 cm). The vegetation indexes (VIs) texture features (TFs) extracted from UAV were screened for high correlation disease index (DI) investigate impact resolution on accuracy monitoring. Finally, best monitoring was determined be then, based cm image, VIs TFs input variables, three algorithms partial least squares regression (PLSR), support vector machine (SVR), back propagation neural network (BPNN) establish scab, models. findings demonstrated that fusion model more appropriate scabs by remote sensing had better fitting than single data source during filling period. SVR algorithm has effect multi-source (VIs TFs). training set identified 0.81, 4.27, 1.88 coefficient determination (R2), root mean square error (RMSE), relative percent deviation (RPD). verification 0.83, 3.35, 2.72 R2, RMSE, RPD. In conclusion, results this provide a scheme field crop diseases area, especially classification variable application near-earth