Shadow identification and height estimation of defects by direct processing of grayscale images

• Published in: International journal of advanced manufacturing technology 2019-03, Vol.101 (1-4), p.317-328
• Main Authors: Galan, Ulises, Ahuett-Garza, Horacio, Orta, Pedro, Kurfess, Thomas
• Format: Article
• Language: English
• Subjects: Aluminum; CAE) and Design; Casting defects; Castings; Computer-Aided Engineering (CAD; Datasets; Engineering; Gray scale; Industrial and Production Engineering; Inspection; Mechanical Engineering; Media Management; Original Article; Shadows
• ISSN: 0268-3768; 1433-3015
• DOI: 10.1007/s00170-018-2933-6

This paper presents a visual inspection system designed to identify and indirectly measure defect heights on aluminum castings. In the proposed approach, a metallic surface is illuminated at three different positions and grayscale images are captured. Unlike other methods that employ a threshold value from one image, the method shown in this article processes information from the three grayscale images to identify shadows. The data from these images is stored in matrices, and four data sets are obtained from element by element operations among the sets. The data sets exhibit a specific pattern at the position of a shadow, and it is possible to identify the pixels where it starts and ends. The method allows discriminating between actual shadows and dark spots. Height estimation was made based on the shadow’s size and they were compared to the actual height with an average error of 5.9%. The system uses a camera connected to a Jetson TX1 board, and the data is processed in parallel on a GPU to obtain results in real time.