Aluminum profile surface defect detection system integrating deep learning and industrial internet of things

Under the background of the continuous development of Industry 4.0 and intelligent manufacturing, the deep integration of industrial Internet of Things (IIoT) technology and artificial intelligence is promoting the evolution of traditional quality inspection links towards intelligence and automation. As a key structural material, the surface defects of aluminum profiles directly affect the mechanical properties and service life of the product. This paper thoroughly explores the significance of aluminum profile surface defect recognition and the limitations of traditional methods, focusing on the application of various deep learning models for this task. Utilizing the dataset provided by the 2018 Guangdong Industrial Intelligence Big Data Intelligent Algorithm Competition on the Tianchi Feiyue Cloud Platform, we experimentally compare and analyze the performance of ten deep learning models: DarkNet-19, DarkNet-53, EfficientNet-b0, GoogLeNet, MobileNetv2, NasNet-Mobile, Places365-GoogLe, ResNet-50, ShuffleNet, and SqueezeNet. The aim is to provide efficient solutions for aluminum profile surface defect recognition, thereby enhancing the quality of aluminum profiles in industrial production. The research results show that a variety of lightweight models have good reasoning efficiency while maintaining a high recognition accuracy, which is suitable for real-time detection systems integrated in industrial Internet of Things environments. This study establishes the first benchmark of 10 classical deep learning models for aluminum profile defect recognition and proposes a ‘accuracy-speed-scenario’ selection framework, bridging the gap between academic models and industrial deployment.