Solder paste inspection using region-based defect detection

This paper introduces a new method to detect for solder paste defects on printed circuit board. This method not only recognizes volumetric defect but also shape deformity. It involves the use of the region-based classification technique. The 3D profile of solder paste block is first reconstructed using phase shift profilometry. The reconstructed 3D profile is then divided into several regions by geometric segmentation. The resulted regions include edges and top surface. Several features which represent their regions most are extracted and analyzed from the solder paste blocks. Regional classifiers are applied to process these feature spaces. The classifier outputs are then combined to form final detection result for each solder paste block. The performance of the region-based defect detection method is evaluated using 150 and 200 mu m solder paste blocks.