Research On Hyperspectral Image Recognition Method Of Wheat Scab Based On Ensemble Learning

In recent years,large-scale outbreaks of wheat scab in many areas have affected wheat yields.Studies have shown that hyperspectral detection technology can detect wheat scab and other diseases with the good results.However,hyperspectral image has the characteristics of large amount of data and many dimensions.So this paper chooses deep learning method to recognize wheat scab by constructing deep neural network model.In order to improve the accuracy of deep neural network recognition of wheat scab,this paper analyzes the structure of deep model network model and the many methods of ensemble model.Then construct a variety of ensemble models which evaluates and analyzes the accuracy of recognition.In addition to this,the generalization ability of the model itself in order to find an optimal ensemble model for recognizing wheat scab.The main research contents determined according to the research objectives are as follows:1.A deep neural network model for recognizing wheat scab is studied.Four deep neural network models with different structures were constructed and structural analysis was performed on these four models.Based on the hyperspectral image data of wheat scab,the four models were trained and tested,and the overall performance of the model was analyzed and evaluated through training results and test results.The four deep neural models are convolutional neural network models like VGG₁ and VGG₂ and recurrent neural network models like LSTM and GRU.2.The ensemble model based ensemble learning method and the optimal model combination method are studied.Four deep neural networks are arranged and combined by Stacking algorithm.Then eleven different integration models are constructed.The training results and testing results of the ensemble model are analyzed and evaluated.To improve the performance of ensemble models,they are optimized and optimized by implementing ensemble algorithms such as random forest,Adaboost,XGboost and gradient boosting decison tree.The ensemble learning results of multiple model combinations are analyzed and evaluated.3.The optimal depth neural network model and the ensemble model are compared.The testing results of deep neural network model and ensemble model structure are analyzed which to establish an ensemble model with high recognition accuracy and generalization ability for wheat scab in the field.At the same time,the research target which improve the recognition accuracy of wheat scab is realized by the method of ensemble learning.