Applicative Framework for End-to-End AOI Implementation
Figure 2-3 : Software topology blueprint of the designed solution .
A considerable number of images are needed to develop the AI models . Frames are retrieved directly from the physical process through the installed camera and first saved in a dedicated pc local folder , then , once properly renamed , duplicated into a shared folder .
At constant intervals of time , the shared folder is analyzed , and new contents are transferred into Azure Data Lake , a cloud data warehouse service suitable for unstructured data , where they are stored in different locations based on the metadata embedded in their names ( id , timestamp , information from PLC ).
The pipeline used for moving data across different data sources is based on Azure Data Factory cloud service . This workflow allows for keeping track of images from different applications and makes them easily available for processing , AI model parametrization and training .
The new trained model is then imported in the main application which runs as a service in a dedicated container managed with Docker platform and located in the dedicated pc : in this way it performs continuously the analysis on the real process .
AOI FOR MATERIAL MIXING PREVENTION Add Main Chapter A Section
To maximize competitiveness on the market , often manufacturing companies opt for the adoption of highly flexible and automated multi-product lines , and standardized products designed to generate components to be as similar as possible . At the same time maintaining the
IIC Journal of Innovation 9