Μεταπτυχιακές Εργασίες

Μόνιμο URI για αυτήν τη συλλογήhttps://pyxida.aueb.gr/handle/123456789/7

Περιήγηση

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω

Τώρα δείχνει 1 - 1 από 1

Computer vision on piping and instrumentation diagrams: towards the identification of their components
(08-12-2021) Androvitsanea, Anna; Ανδροβιτσανέα, Άννα; Athens University of Economics and Business, Department of Informatics; Vassalos, Vasilios; Koutsopoulos, Iordanis
Piping and Instrumentation Diagrams (P&IDs) are schematic representations of equipment, pipelines, instrumentation, and control systems. They appear in process environments, such as Oil Refineries, Chemical Plants, Paper Mills, and Cement Plants, etc. The identification of each element constituting a P&ID, along with the way they are interconnected, is an important task that has not been automated yet. In this work we study a methodology and develop the respective algorithm towards the identification of these components. This identification aims to the classification of the elements based on their representation as images as well as to the identification and translation of the codes included in the diagrams.In order to achieve this goal a combination of methods are employed. Using the OpenCV library the outlines of the P&ID are calculated. An algorithm is developed, which based on the coordinates of the outlines, delivers snapshots of the elements constituting the P&ID. In the sequel, these elements are classified by a suitably designed classifier, to one out of 53 classes. The classifier is a convolutional neural network (CNN), implemented using the TensorFlow and Keras libraries, which was trained on a data set of 2970 images that belong to one out of 53 classes.Textual information contained in the P&ID are identified, using the pytesseract library and stored into an array. Then, they are passed to an algorithm that implements the ANSI/ISA-5.1.-1984 (R1992) standards and deciphers the textual tags, by providing as output the name, function, modifier etc. of each element. The model is able to successfully identify an image and attribute it to the right class, which is a great step towards solving the challenging problem of the identification of the elements constituting a P&ID.cPiping and Instrumentation Diagrams (P&IDs) are schematic representations of equipment, pipelines, instrumentation, and control systems. They appear in process environments, such as Oil Refineries, Chemical Plants, Paper Mills, and Cement Plants, etc. The identification of each element constituting a P&ID, along with the way they are interconnected, is an important task that has not been automated yet. In this work we study a methodology and develop the respective algorithm towards the identification of these components. This identification aims to the classification of the elements based on their representation as images as well as to the identification and translation of the codes included in the diagrams.In order to achieve this goal a combination of methods are employed. Using the OpenCV library the outlines of the P&ID are calculated. An algorithm is developed, which based on the coordinates of the outlines, delivers snapshots of the elements constituting the P&ID. In the sequel, these elements are classified by a suitably designed classifier, to one out of 53 classes. The classifier is a convolutional neural network (CNN), implemented using the TensorFlow and Keras libraries, which was trained on a data set of 2970 images that belong to one out of 53 classes.Textual information contained in the P&ID are identified, using the pytesseract library and stored into an array. Then, they are passed to an algorithm that implements the ANSI/ISA-5.1.-1984 (R1992) standards and deciphers the textual tags, by providing as output the name, function, modifier etc. of each element. The model is able to successfully identify an image and attribute it to the right class, which is a great step towards solving the challenging problem of the identification of the elements constituting a P&ID.

Περιήγηση

Πλοήγηση Μεταπτυχιακές Εργασίες ανά Συγγραφέα "Androvitsanea, Anna"