Classification And Detection Of Household Garbage Based On Deep Learning

Household garbage is a misplaced resource.Effective garbage classification can not only protect the environment,but also realize the reuse of resources.China has one of the largest population in the world,so people produce a large amount of garbage every year,so our country has been facing the problem of garbage disposal.In our country since the founding of garbage disposal problem has been given high attention and publish relevant policies in China in recent years,with the 2020 national recycling rate of 35% as the goal,relevant departments of the selection of nearly 50 cities across the country ahead of living garbage classification of compulsory pilot work,across the country to carry out the garbage classification for the future work to make sufficient preparations.Since compulsory garbage classification has never been implemented in China,the classification of domestic garbage is faced with many difficulties in both the source and recycling.For example,the meaning of garbage classification of residents is weak and the garbage classification error rate is high.However,there are many problems such as heavy workload and low sorting efficiency in centralized garbage classification.With the rapid development of economy and the emergence of artificial intelligence,garbage classification should be combined with advanced science and technology to better realize garbage recycling and utilization.For the current situation of garbage classification in China,the intelligent and efficient garbage classification has become an urgent problem to be solved.The content of this paper is to apply the optimized deep learning object detection model to garbage classification,so as to realize accurate and efficient automatic garbage classification.This paper mainly applies the deep learning target detection model to the household garbage classification,and optimizes the target detection network model in three aspects to improve the accuracy of household garbage classification.Finally,the optimized model is compared with various deeplearning target detection models.The main research results of this paper are as follows:(1)research shows that the attention mechanism in convolutional neural network can generally improve the accuracy of the model by 1 to 2percentage points.In this paper,the attention mechanism is added to the target detection network to improve the accuracy of the model.(2)the clustering grouping normalization is proposed.The output of each layer in the neural network may lead to the deviation of data distribution,thus affecting the training speed of the network model.The normalization operation can reduce the internal covariance deviation,smooth the loss function and thus accelerate the training.This paper proposes a clustering grouping normalization based on the similarity of the features extracted by convolution kernel.(3)the optimized model is used to finally achieve the positioning and classification of domestic garbage.