Evaluation of low-power vision platform for robotic industrial application

• Main Author: Aung, Ye Lin
• Other Authors: Nachiket Kapre
• Format: Final Year Project
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/66767
• Institution: Nanyang Technological University

Cracking in plane turbine blades is the significant defect for the airplanes. Cracking usually occurs because of high pressure and temperature during manufacturing processes. In this project, automatic crack inspection system will be developed and implemented on real-time system. Inspection system is implemented on Jetson Tk1 embedded hardware and Robot Operating System (ROS). Many researched methods will be compared and inspection algorithm is developed based on the comparison results. Inspection algorithm includes a sequence of image processing methods and machine learning classifier to correctly output the defect location. For the final step, defects location coordinates will then return to ABB Industrial Robot for further executions and corrections. Accuracy rate of 89% was achieved at the final classification stage of the system with the average processing time less than 1 second.