Research on Garbage Classification Based on Improved MNFN Algorithm

The lack of public participation, high technological costs, and inadequate policies and regulations seriously hinder the efficient promotion of garbage classification and processing as an indispensable part of modern urban civilization management. The paper proposes a garbage classification algorithm MNFN (Mobile Neuro Fuzzy Network) based on the fusion of MobileNet and Cascade Neuro-Fuzzy Network. The collected garbage images are processed by filtering to remove noise and improve image quality, and then segmented using Squeeze U-SegNet to enhance model convergence speed. To accurately represent the differences between different types of garbage, this paper extracts ten types of feature information from garbage images as inputs to the classifier. The SMOTE (Synthetic Minority Over-sampling Technique) algorithm is utilized to address the issue of imbalanced samples in garbage image data, followed by garbage classification using the MNFN algorithm. Finally, an accuracy evaluation model is established to compare the performance of common classification algorithms and validate the superiority of the algorithm proposed in this paper.