Effective Adversarial Examples Detection Improves Screw Detection in Dust Pollution Within Industry System

Screws play a critical role as essential components across various industries. To meet market demands and standards, screw manufacturers are embracing digital transformation and leveraging artificial intelligence (AI) techniques. AI models have also been proposed for identifying defective products. However, most studies are limited to controlled environments, and the presence of dust particles during screw manufacturing poses challenges for visual-based AI applications. To address this, we propose a solution involving the use of adversarial examples (AE). These examples are employed to simulate dust particles on camera lenses. We introduce a robust AE detection method designed to enhance the accuracy of screw recognition. This approach aims to improve the overall efficiency of the screw manufacturing industry. Experimental results show that our proposed mechanism can effectively improve screw identification accuracy in dust-polluted environments.