The invention provides a
wafer detection method and
system based on multichannel C scanning, and aims to solve the problems of multichannel sensitivity difference, Z-axis height deviation, splicing artifacts and the like in a traditional
system. The maximum defect point and the defect-free point of the standard sheet are selected before detection, the
gain of each channel is adjusted to a unified amplitude, and the echo amplitude is recorded, so that sensitivity calibration is realized; in the detection process, a
laser displacement sensor is used for collecting
wafer surface height data in real time, modeling is conducted in combination with an ARIMA + Kalman filtering or
Gaussian process regression
algorithm, the Z-axis focusing distance is dynamically adjusted through PID feedback, and
signal attenuation caused by surface warping is compensated; during
image splicing, channel response is optimized by adopting three gray mapping
modes of linear self-adaption, Gamma self-adaption (automatic alignment channel mean value after reference gamma setting) and manual setting, sub-pixel level offset is calculated based on a two-dimensional cross-
correlation function to realize coordinate alignment, and a 5% overlapping area is set at a channel junction. Gray transition is completed through
linear gradient weight weighted averaging.