Application of 3d models in intelligent visual inspection of product anomalies

Due to its advantages of non-contact, high efficiency, and digitization, machine vision has been extensively utilized in product appearance anomaly inspection. The advancement of deep learning has significantly propelled the digitization and intelligence process of visual inspection. Nonetheless, the current learning paradigm still heavily relies on a large quantity of high-quality training samples, and exhibits limited capability in detecting anomaly categories beyond the training set (also known as open-set detection). Aiming at these challenges and inspired by the model-driven digital inspection idea, this article advocates for the thorough exploration of the auxiliary potential of 3D models in visual inspection. Diverging from the traditional paradigm, which mainly leverages digital models as data management centers, this article explicitly highlights two additional critical functions of 3D models: knowledge carriers and detection references. The efficacy of the proposed approach is demonstrated through case studies.