The Application Of Dimensionality Folding Method In Garbage Classification

The principle of domestic waste management in China is ”reduce,resource and harmless”,i.e.,to avoid generating waste as much as possible,and to take care of the waste already generated.Waste should be recycled as much as possible,followed by incinera-tion and finally landfill disposal.However,at present,the coverage of the recycling facilities is not extensive and the waste is not sorted carefully,and the residents are mainly unable to sort their waste correctly.Sorting garbage,lack of awareness and knowledge about garbage sorting,and other problems.To this end,this paper proposes an intelligent recyclable garbage sorting algorithm that can automatically identify the category to which the garbage belongs based only on the garbage image information obtained from a photo-graph.Thus,smart waste sorting algorithms can correctly sort recyclable waste on the one hand and improve the resource efficiency of recyclable waste on the other.The recyclables that commonly exist in China are mainly waste paper,scrap metal,waste glass,and waste plastic.Therefore,this article first collects image data of four types of recyclable garbage: plastic,metal,cardboard,and glass.Using Lanczos algorithm to uniformly process the data into 192 × 256 × 3 size images,and then expand the sample size through data augmentation methods such as rotation,light,contrast,and blur.Next,the 192 × 256 × 3 size image is converted into a 192 × 256 size two-dimensional array,that is,the explanatory variables are data in the form of a matrix,and the data set is divided into a training set and a validation set.On the training set,the traditional PCA,DFPCA,and DFPFC three dimensionality reduction methods are used to reduce the dimensionality of the high-dimensional image data,and the validation set data is reduced by the parameters trained on the training set based on the three dimensionality reduction methods.The traditional PCA dimensionality reduction method first vectorizes the matrix form data and then performs dimensionality reduction without considering the inherent structure of the matrix.The two-dimensional dimensionality reduction method of DFPCA and DFPFC is to directly reduce the dimensionality of the explanatory variables of the matrix values,and the explanatory variables after the dimension are still in the form of a matrix,so that the inherent correlation structure of the matrix can be well preserved.Compared with DF-PCA,DFPFC uses the information of the explanatory variables and belongs to supervised dimensionality reduction.The results of the three dimension reduction methods on recy-clable garbage data show that DFPFC has the best dimension reduction effect,followed by DFPCA,and finally the traditional PCA.Finally,vectorize the data after three dimension reduction methods as explanatory vari-ables of recyclable garbage,and use the category plastic,metal,cardboard,and glass to which the recyclable garbage belongs as explanatory variables.The Light GBM is used for model fitting,and PCA-Light GBM,DFPCA-Light GBM,and DFPFC-Light GBM are used to evaluate the performance of the model on the validation set.The results show that the DFPFC-Light GBM has the best classification effect,and its average-macro index is more accurate.The macro-average precision is 0.93,the macro-average recall is 0.80,and the macro-average F1-score is 0.85.Therefore,for a new recyclable garbage image,the DFPFC-Light GBM can be used to determine the probability of the category which garbage belongs to.This also shows that for image data of the recyclable garbage,the effect of dimensionality reduction and reclassification using matrix values is better than the effect of vectorized dimensionality reduction and reclassification,thereby further improving the accuracy of recyclable garbage identification and resource utilization.