A Tomato Leaf Disease Recognition Method Based on Hybrid Attention Mechanism
The disease of tomato leaf becomes a serious threat to the yield and quality of tomatoes.In order to achieve early de-tection and prevention of tomato diseases,it is necessary to have an effective method for identifying tomato diseases.Shannon entro-py can be used as a measure ment of the amount of information in an image,but a huge calculation of channel Shannon entropy in deep convolutional neural networks can increase the training time of network.Since there is a positive correlation between image Shannon entropy and image variance,image variance can be calculated to measure the amount of information in an image.This pa-per proposes a deep convolutional neural network method based on the Squeeze-and-Excitation(SE)mechanism and channel vari-ance.Firstly,the variance data of the channels are calculated to measure the channel information,which a big variance of channel has more information,and vice versa.According to the variance values of the channels,weighted processing of the channels is car-ried out,followed by SE operations on the channels.The experiments show that the method proposed in this paper can effectively classify and identify types of tomato diseases,with an accuracy improvement of 4.65％compared to traditional methods.
万方数据
维普
