Automated brand inspection of plastic IC packages through computer vision using neural network

• Main Author: Arciaga, Generoso A.
• Format: text
• Language: English
• Published: Animo Repository 1995
• Subjects: Neural network; Computer vision; Brand name products; Plastics in packaging; Image processing; Quality control; Network analysis (Planning); Defect correction methods (Numerical analysis); xx6 Quality assurance; Engineering
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/1669
• Institution: De La Salle University

The study is made to investigate the industrial applicability and performance of neural network-based computer vision system designed for inspection of actual brand images marked on top of IC packages. There were three devices considered in this thesis based on the limited and controlled number of these units provided by the company. Using the available limited samples of devises, brand names were captured using a camera and processed in the neural network using various representations to investigate thoroughly the effect of neural network parameters to network performance. As a result, using the neural network technique trained by backpropagation in brand defect recognition proved to be very effective within the scope and limitations of the concluded study.