Model And Application Of Optimized Convolutional Neural Network Based On Granular Computing

The convolutional neural network model has the advantages of simple structure,high classification accuracy,and strong adaptability.It is widely used in image processing.Because,the input data of the model is often an image where the target pixel and the background pixel are fused,training to extract the image data features will take a long time.The granular computing model's ability to deal with fuzzy and inaccurate problems can help segment uncertain pixels and make the feature extraction speed up.Therefore,the granular computing model was chosen to optimize the convolutional neural network model for reducing the training time and the cost of image data mining.The main work of the model and application of optimized convolutional neural network based on granular computing was as follows:Firstly,the granular layer was constructed by the granular computing model.In order to avoid the problem of incomplete consideration caused by similar models,three methods of principal component analysis,fuzzy C-means clustering and cloud model were selected to constructed the granulation layer.Three models were used to segment the image,then the segmentation result was evaluated by the evaluation index.Secondly,according to the structure and principle of the convolutional neural network,the granulation layer was placed between the input layer and the convolutional layer.The pre-processed image entered the granulation layer for granulation and segmentation,then entered the convolution layer for convolution operation.This was a key step of the granular computing optimization model.The determination of the granular layer position played a vital role in improving the accuracy of the model classification.Finally,in order to avoid a single data set from affecting the experimental results,some images from Oxford-IIIT Pet Dataset and Dogs vs.Cats to form new data were selected for experiment randomly.Two types of satisfaction standards were set according to requirements,then the time to reach these two types of satisfaction in the last iteration of the model was calculated respectively.A suitable model of optimized convolutional neural networkl was selected by combining the segmentation effect for comparative analysis.Because the segmentation effect of the cloud model was the best,and the time to reach the two types of satisfaction was the shortest,the cloud model was determined to be the construction model of the granulation layer.The model and application of optimized convolutional neural network based on granular computing mainly used granular segmentation to simplify image features when classifying images for improving the training speed of the model.The model shortened the training time greatly,and helped to promote the further development of knowledge discovery in image data field,and it was of great significance for large sample data mining.Figure 40;Table 3;Reference 64...