Clustering analysis on disease severity of wheat stripe rust based on SOM neural network

摘要

A SOM (Self-organizing Feature Maps) model was introduced to cluster and analysis on the disease severity of wheat stripe rust based on PHI (Pushbroom hyperspectral imager) data. By means of acquiring the spectral index data (SID) and spectral angle data (SAD) of the samples, combining with the samples' spectral average reflectance data (ARD), three two-dimensional data matrixes were obtained as the inputs of SOM model. After iterative learning and self-organized clustering, the models' outputs farthest approached to the reality in 3-dimensional severity space of wheat stripe rust. Then, with the net-trained, all data of the trial plot were simulated. The simulating results demonstrate that the division of wheat stripe rust severity is obviously. The whole trial spot was derived into four grades and the results are satisfactory.