Quality assessment of Lithium-ion batteries using light optical microscopy is of high relevance, as this type of batteries are widely used in range of applications like consumer electronics, electric vehicles, and energy storage systems. Ensuring quality is critical to prevent performance problems, safety hazards and premature aging. Light optical microscopy provides a detailed examination of the microstructure of the cells, allowing for the identification and characterization of defects and anomalies. Incorporating this into an at-line quality assurance process allows manufacturers to enhance battery performance, extend lifespan, and minimize safety risks. This approach is also vital for OEMs, who need to select suppliers based on stringent quality standards. Moreover, microscopy enables researchers to uncover the root causes of defects, refine manufacturing processes, and contribute to the development of innovative materials and designs for future-generation batteries.
While supervised machine learning approaches like object detection or instance segmentation yield good results for defect detection in Li-ion batteries, the time and effort that has to be put into labelling huge amounts of data is a drawback. Unsupervised methods can significantly reduce workload. In this work, we give an overview of different unsupervised approaches for defect and anomaly detection applied to light optical microscopy images of different Li-ion batteries.