Abstract
This paper introduces a methodology for online inspection of surface mount components using vision and infrared sensors. The complementary nature of vision and infrared sensors allows for separation of solder joint defects into surface level defects and solder mass related defects for defect detection. The vision sensor can provide reliable information for surface level defects while infrared sensors are capable of providing information about solder mass related defects. An experimental facility is described and software routines developed for defect defection. Using both oblique and flat illumination techniques, the 2-dimensional gray scale images of the printed circuit boards are captured and preprocessed for suitable feature extraction. A sampling scheme has been developed for infrared inspection to reduce the inspection time. For defect detection a neural network construct is used to compare incoming data with stored templates. In addition, a classifier based on fuzzy relations and linguistic labels has been used to perform classification of linear misalignment defect accommodating the uncertainty associated with it. The strategy is demonstrated with inspection of a number of solder joint defects.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call