Abstract: In industrial applications, it is necessary to accurately detect scratches on the
surface of a quad flat no-lead package (QFN) chip in real time. A rapid chip
surface scratch detection and location method was proposed. According to image
segmentation algorithm, a defect image can be acquired firstly. Then by combining
with the contour extraction algorithm, the chip surface scratch location can be
achieved. At the same time, in order to ensure real-time detection of scratches
on the chip surface, image segmentation is further completed by using the
Otsu multi-threshold algorithm based on particle swarm optimization (PSO)
algorithm, which not only makes the defect area in the image more obvious,
but also shortens the scratch detection time on the chip surface. Compared
with the direct use of the Otsu algorithm, the scratch detection time on the
chip surface is reduced from seconds to milliseconds, and the chip quality
detection efficiency has been improved greatly. It is shown that the method
has important reference value for the development and application of
software systems for chip detection equipment.

Key words: quad flat no-lead package chip, scratch detection, multi-threshold segmentation, particle swarm optimization algorithm