Abstract
LED chips positioning is a key process in chip quality classification. With the use of pattern-matching mechanism, we proposed a new method for high accuracy positioning system which can be used in detection, sorting and crystal binding machines. Firstly, we used the Otsu and morphological filtering algorithms to gather all kinds of LED chips image. The boundary characteristic points of image were then obtained to bring the feature vector to generate a pattern template. Similar characteristic points were then searched in the sample images to achieve accurate chip positioning. Finally, the Euclidean distance error between the template and target feature vector was compared. As the pattern template is only dependent on the boundaries and independent of the grayscale, the positioning robustness and accuracy can be drastically improved, achieving a positioning error of less than 1 μm and a positioning speed of over 574 grains/second.
Sign-in/Register to access full text options 
Free) 
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
