A Wafer Segmentation Method Using the Closest Affine Iterative Point

A novel wafer segmentation method based on the closest affine iterative point (ICP) is proposed to address the problems inherent in wafer segmentation relying on the hardware instruments, such as high production cost, complicated process and unstable segmentation effects. The method is independent on hardware facilities, and establishes the feature point sets of the wafer template image and the target image by utilizing the Canny operator to extract edges of the images and obtains the rough wafer rectangle information through Hough's straight line detection. The affine ICP algorithm is then used to perform precise registration with the initial value that is the coordinates of the upper left corner of the rectangle. The model image is matched with the reference image which results in a rapid and accurate wafer segmentation. Theoretical analysis and experimental results show that the computational complexity of the method is low and a single sample segmentation takes no more than 0.9 seconds, while the sample segmentation accuracy is significantly higher than those of other algorithms. Therefore, the proposed method satisfies the real-time detection requirements of automatic production lines.