Semi-Automated Vision-Based Construction of Safety Models from Engineering Drawings

摘要

The work describes a concept application that aids a safety engineer to create a layup of equipment models by using an image scan of a schematic and a library of predefined standard component and their symbols. The approach uses image recognition techniques to identify the symbols within the scanned image of the schematic from a given library of symbols. Two recognition approaches are studied, one uses General Hough Transform; the other is based on pixel-level feature computation combining both structure and statistical features. The application allows the user to accept or edit the results of the recognition step and allows the user to define new components during the layup step. The tool then generates an output file that is compatible with a formal safety modeling tool. The identified symbols are associated to behavioral nodes from a model based safety tool. The model based safety modeling tool, together with the associated library is used during the last step for the user to link the nodes together and to ensure that the behavior of the equipment is correct. The tool improves the time to create a layup of equipment, and helps make the model of the equipment look like the original schematic to help engineers remain familiar with the model when compared with the original schematic. This helps a safety engineer to create a formal safety model for an existing aircraft based on an existing failure library of components.