Break | AIChE


Image analysis is a fundamental aspect of modern research, as many laboratory test methods consist of visual output that requires manual inspection or rating to assess product performance. The breadth of image analysis applications is expansive across a number of industries, including, for example, quantitative assessments (such as object or pattern recognition, color evolution, defect analysis, size determination), time-lapsed visual comparisons, and qualitative assessments (such as aesthetic judgements or ratings comparing relative performance or past experience). The automation of image processing provides efficiency gains and standardization for these methods, ultimately improving the consistency and quality of data in a research setting. Experience-based visual methods in particular suffer from low efficiency, lack of consistency, and poor reproducibility, which can be substantially improved with automated image analysis. Tests and results can be stored digitally for better knowledge retention and access across an organization. Furthermore, the available information in the database can be built-up over time to accommodate big data analytics and in-depth statistical analyses in the future. As an industry leader, Dow has developed in-house expertise and methodologies that leverage digital image processing across many businesses.

In this talk, we will demonstrate the scope of image analysis problems and discuss one specific tool development in detail: microscope image stitching of film samples in the plastics industry. We will highlight the context of the test method, technical challenges, the reason to choose in-house development, and the strategies and challenges of method development and validation.