Utilizing Computer Vision for the Analysis of Manufacturing Processes

Abstract

hroughout the manufacturing industry, video recordings are used to help standardize work and develop training material for companies. Solme AB develops a software suite named AviX which aids in the analysis of these recordings. This report aims to evaluate how computer vision technology could be utilized to increase the functionality of the AviX suite. Furthermore, the report will evaluate how the technology could be used to automate analysis currently performed manually in the program. The evaluated features are face blur, tool highlighting, ergonomic risk detection, and footstep counting. A software platform is developed in Java, primarily with the use of OpenCV, to serve as a proof-of-concept for Solme. To support the flexibility of changing the set of enabled features, the application was constructed modularly and the features were implemented independently. The thesis concludes that there is potential to extend the functionality of the AviX suite by utilizing computer vision. Automated face blurring has been achieved with a considerable success rate, increasing the privacy of people appearing in the video recordings. Moreover, the automation of ergonomic risk detection showed promising results which indicate that manually performed analysis can indeed be automated.