Research On Image Classification Method Based On Convolutional Neural Network

In recent years,with the progress of science and technology and the continuous development of Internet technology,image classification technology has attracted extensive attention of many scholars,and has become the research focus of computer vision,which is widely used in aviation,transportation,security inspection and other fields.Although a series of achievements have been made in the field of image classification accuracy and efficiency,the classification accuracy and efficiency of small sample image,fine-grained image and multi factor interference image still need to be further improved.Therefore,this paper puts forward the corresponding improvement methods for the above problems(1)Aiming at the problem of insufficient feature information extracted due to the small amount of sample image data.This paper proposes an image classification method based on an improved convolutional neural network.Firstly,a multi-scale filter bank of 11?,33?,55? is used to obtain the image’s texture,shape and other characteristic information from multiple aspects.Then,by adding a residual learning module to the convolutional layer,the number of information transfer layers is reduced,and the image features are directly transferred to the deeper layers of the network.Experimental results on multiple data sets show that this method can increase the accuracy of image classification and reduce the time complexity of image feature information transmission(2)Aiming at the problem of low classification accuracy caused by the small difference between fine-grained image classes,this paper proposes a fine-grained image classification method based on multi-layer weights and annotation sets.First,set the corresponding weights in the multiple convolutional layers of the VGG-16 model according to the different feature information obtained,so as to obtain different feature information of the image from multiple angles;then create a label set with multi-category labels,so that The classifier can be used for multiple classification training in one propagation.Experimental results show that this method can increase the difference between fine-grained image classes and improve the accuracy of classification.(3)Since the multi-factor interference image will be affected by factors such as illumination,expression,posture,etc.during feature extraction,this paper proposes an image classification method based on the combination of Gabor wavelet transform and Alex Net convolutional neural network.First,LBP algorithm is used to preprocess the image information to remove interference factors,and then the image information is divided into sub-modules,and the Gabor wavelet transform is used for local feature extraction,so as to obtain the local feature information of the entire image.Secondly,the global features of the image are obtained through the convolution operation of the Alex Net convolutional neural network model,and the global features and local features are feature-fused,and the Softmax classifier is used to obtain the optimal number of kernels for classification learning.Experiments have verified that this method is effective for improving the classification accuracy of multi-factor interference images.Figure[45] Table[15] Reference[70]...