Reagent Label Text Detection Using the Stroke Width Transfom

We implemented an algorithm to extract text from reagent labels to facilitate the retrieval of safety information in a laboratory using an Android mobile device. Our algorithm combined the stroke width transform and a variety of connected component filters to detect text candidates. These were then processed using Tesseract, an open-source optical character recognition engine. We concluded that Tesseract benefits from our implementation when reagent labels contain many non-text objects.