Research on Image Data Augmentation and Accurate Classification of Waste Electronic Components Utilizing Deep Learning Techniques

The escalating accumulation of waste printed circuit boards (WPCBs) underscores the urgent need for efficient recovery of valuable resources. Notably, WPCBs harbor a considerable number of intact electronic components that remain functional or could be repurposed. Nevertheless, the automated recognition and sorting of these components remain highly challenging, owing to their miniature dimensions, diverse model types, and the absence of publicly available, high-quality datasets. To address these challenges, this paper introduces a novel image dataset of discarded electronic components and proposes a deep learning-based data augmentation model that combines classical augmentation methods with DCGAN and SRGAN to achieve dataset size augmentation. This paper further conducts Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) evaluation on the generated images to ensure their suitability for downstream classification tasks. Experimental results demonstrate significant improvements in classification accuracy, with AlexNet, VGG19, ResNet18, ResNet101, and ResNet152 achieving increases of 6.6%, 9.7%, 4%, 5.4%, and 6.2%, respectively, compared to classical augmentation. This method enables precise identification to facilitate the downstream recovery of intact electronic components, thereby contributing to the conservation of natural resources and the effective mitigation of environmental pollution.