The Design and Implementation of an Intelligent Paid Waste Recycling Bin based on Computer Vision

Abstract

In the context of rapid urbanisation leading to a significant increase in household waste and the low efficiency of traditional recycling methods, this study has designed an intelligent paid rubbish recycling box integrated with computer vision technology. The system uses OpenMV for image acquisition, employs convolutional neural networks (CNN) to achieve automatic rubbish classification, and completes payment through 3D face recognition. The experiments show that CNN achieves a 98.5% recognition accuracy on the MNIST dataset, and data augmentation techniques further enhance the model's generalisation ability. The 3D face recognition algorithm has an accuracy rate of 98.5% which is superior to DeepFace's 97.3%, and the combination of rigid region matching and exclusion algorithm can balance the recognition speed and accuracy, reducing 65.1% of the computational load. The point reward mechanism of the recycling box effectively promotes user participation. This study not only improves the efficiency and classification accuracy of rubbish recycling but also lays a foundation for intelligent rubbish management and the development of the circular economy.