Page 94 - Undergraduate Research Opportunities Program Proceedings 2017-18
P. 94
School of Engineering
Department of Industrial Engineering and Decision Analytics
Department of Industrial Engineering and Decision Analytics
Quality Control via Vision System
Supervisor: TSUNG Fugee / IEDA
Student: LIN Chuan-en / COMP Course: UROP1100, Spring
This research project presents a crack identification methodology. With deployment of high definition cameras on
hangers and drones, a building’s exterior surface may be captured and stored as a series of image frames. The frames
are then converted to grayscale by utilizing morphological image processing techniques and thresholding operations.
Using edge detection and object classification algorithms, we overlay the frames with guiding points of possible crack
locations and allow expert users to select the valid cracks manually for use as the dataset of correct images to build our
learning model. By applying transfer learning to Google’s Inception CNN, we can build a customized classifier for
identifying cracks.
84
