Description
This paper has presented an image-based approach for detecting ICs on fully populated PCBs. An image-processing architecture was proposed that leveraged a MAP detector and morphological filtering techniques to segment IC regions, a polygon detection heuristic to reconstruct the vertices of ICs, and a SIFT descriptor matching technique to identify IC part numbers.
Design trade offs and implementation details at each stage of the processing pipeline were examined on the basis of application performance. Finally, the paper presented the measured results of the proposed image-based IC detection approach. Future identification rate performance can be achieved by training the algorithm on both higher resolution images and a greater diversity of PCBs (across lighting and color variations).
Reviews
There are no reviews yet.