Portable Image Analyzer for Indicators with OpenCV

Some diagnostic tools rely on human eyesight to determine their results such as urine dipsticks or pH strips. Ideally, in these cases, human eyesight needs to be verified by other people, or technology, as misinterpretations may lead to dangerous outcomes. Hence, to alleviate this issue, in this proof-of-concept we developed a portable and compact computer vision system using Raspberry Pi 3B+ and a digital camera as a powerful tool to objectively analyze physical indicators such as pH test strips. The algorithm is based on a series of thresholding, localization, color extraction and calculations that averages the multiple BGR parameters into a single value for comparison with a custom pH library. Thus, it compares BGR values to determine the most appropriate pH level based on the smallest difference in BGR between a test strip and a library. Case examples are described where inaccuracies of typical pH strip readings are compensated and corrected by the system.