Research On Intelligent Garbage Classification System Based On Deep Learning

Faced with the increasing volume of domestic waste removal,the wide variety of domestic waste,the difficulty of correct classification of different types of waste,the knowledge reserves related to the overall garbage classification of citizens and the overall quality of the people need to be improved,and many other problems,indirectly lead to the deterioration of environmental resources and affect Harmony between man and nature.At the same time,Accompanied by computer skill’s exaltation,it has expedite of deep learning.The mainstream convolution Neural networks perform well in large-scale image classification tasks.However,for some garbage that is similar but does not belong to the same category,the basic convolutional neural network has a certain degree of difficulty in fine feature extraction.There is an urgent need for the network The model is improved accordingly to meet the research requirements.In order to study the application of deep learning in the field of environmental protection,especially for the specific problem of garbage classification,this article will be based on the Res Net-50 convolutional neural network,which has a better training effect in the field of image classification,and focus on how to improve the refinement of the network.The extraction of features will further improve the accuracy of network classification and carry out research work in this direction.The following are the contents of the research and the results obtained:First,comparing the accuracy of AlexNet,GoogleNet,VGG-16 and ResNet-50 on self-built data sets of several types of convolutional neural networks,and finally choosing Res Net-50 with the highest model accuracy as the basic algorithm model of this article.And added 1x3 convolution and 3x1 convolution to its model to improve the model’s ability to extract detailed features of the target to be detected.At the same time,the original pooling layer of Res Net-50 is replaced with multiple pooling layers of different scales to improve the network model’s ability to obtain contextual information for association.The improved Res Net-50 network model is trained and tested on the data set,and an accuracy rate of 93.87% is obtained.Second,so as to heighten the feature extraction of the target to be detected and further heighten the discern accuracy,this paper introduces a cyclic recursive model to superimpose the features extracted for the first time in the network with the original image,and use cyclic recursion as the new network input.Finally,the mean value output by different sub-networks is used to judge the similarity of each result.Increased success rate of the detection results,the attention mechanism is introduced after the last convolutional layer of the classification network in each scale branch network,which is used to extract the IOU region information from the feature information extracted by the convolution operation,and then The region is cropped and enlarged,so as to be used as the input of the new classification network,and the convolution operation is performed again in a cyclic manner.Finally,using the improved cyclic convolutional neural network model proposed in this paper and verifying it on the data set,an accuracy rate of 95.26% is obtained.Third,the classification results identified by the improved cyclic convolutional neural network model are used as the different inputs of the controller.Through the formulated control strategy,the motor is driven to control the operation of garbage on the conveyor belt,thereby realizing the automatic classification process of garbage.Therefore,this paper proposes a system that can realize the automatic garbage classification function,and through the visual simulation experiment,the method proposed in this paper can well solve the problem of automatic classification of domestic garbage in daily life,and further alleviate the increasing number of garbage types.The problems of multiple and difficult classification have promoted the beautiful situation of environmental protection and harmonious coexistence between man and nature.