Physical form handling in challenging imaging environment

Even in a highly digitized world most of us live in, paper or physical forms (survey, multiple choices exam, etc.) are still playing an important roles. They are not only cheap and convenient, but also the only legal form of documents according to laws in many countries. Currently, these physical forms are mostly handled by hand. Otherwise, we need special forms with special filling instructions. To ease this process, this research proposed a non-supervised learning technique for handing physical forms in challenging imaging environment. This technique works on an assumption that a typical filled form has more blank entries than filled entries and image in each group (blank and filled) are closer to each order in image space. This technique considers special designed forms which can be printed in any paper and filled without any special instruction. With a standard webcam, the proposed method is demonstrated its capability of physical forms handling in challenging imaging conditions.