Having the ability to compare inspections of the same infrastructure over many years enables engineers to recognize the evolution of flaws that might become dangerous. This can readily be done with modern digital data, but is a slow, error-prone process to do with earlier data that is on handwritten inspection forms. YoJonesy was engaged to find a way to automatically identify and interpret these forms from these old reports.
Having the ability to compare inspections of the same infrastructure over many years enables engineers to recognize the evolution of flaws that might become dangerous. This can readily be done with modern digital data, but is a slow, error-prone process to do with earlier data that is on handwritten inspection forms. YoJonesy was engaged to find a way to automatically identify and interpret these forms from these old reports.
By breaking the problem down into component parts, YoJonesy was able to apply multiple fields of machine learning, including localized computer vision and cloud-based AI recognizers, to develop an integrated custom solution. We also built a UI that allows users to visually define areas of interest and thresholds for interpretation confidence.
The Virtual NDE (non-destructive evaluation) software application simulates the conditions and functions of manual ultrasonic scans and allows users to gain experience performing inspections using techniques similar to those used with the instruments on actual specimens.
The resulting application can process multiple documents looking for multiple form types, each with user-defined fields of interest. It interprets, catalogs and captures the data in a database where it can be searched and analyzed together with more recent data. It also flags any interpreted fields with confidence levels below the user-defined threshold for manual confirmation.
