| dc.description.abstract | In this paper, we use feature analysis to identify the scratch errors of the wafer map, and use the Hough transform to find the defects of straight lines and arcs on the wafer map. The actual wafer we used is the WM-811K wafer database provided by TSMC, and the error patterns can be divided into the following nine types: Center, Donut, Scratch, Edge-Ring, Edge-Loc, Loc, Near-Full, Random, None. And this paper is the part to distinguish Scratch.
First, we perform a noise reduction operation on the wafer map, remove discrete defects on the wafer map, and leave the more obvious cluster points on the wafer map. After the noise is reduced, the wafer map input map is converted into a gray-scale pattern, and the wafer map is adjusted and scaled to a uniform format array, so that it is convenient for parameter setting during Hough transformation.
Then, when looking for the shape of the straight line on the wafer map, we use the equation of the polar coordinate system instead of the linear equation to avoid that when the straight line is perpendicular to the X axis of the plane coordinate system, the slope is infinite and cannot be calculated. When looking for the shape of the arc on the wafer map, the ellipse equation is used to calculate. By applying these two formulas to the Hough transform algorithm to find the scratch pattern on the wafer map, it can be achieved based on feature analysis. Ways to identify the scratches on the wafer map.
Finally, the scratch patterns are identified by these two Hough transform methods, and part of the wafer map is taken from the WM-811K wafer map database for testing, and the numerical performance of Accuracy, Precision, and Recall is used to show the model of this paper. Judgment accuracy. | en_US |