An intelligent optical scanner based on visual AI algorithms was experimentally designed to realize automatic correction when scanning skewed placed documents and automatic correction of paper bending due to spine binding when scanning books. The use of the algorithm instead of the bracket and laser assist module in the traditional scanner further promotes the process of a paperless office and realizes the protection of renewable resources such as trees and water, while reducing the difficulty of operation for the user, decreasing the cost, and increasing the types of documents supported by the scanner. Therefore, compared with the traditional scanner, this work has the advantages of energy conservation, high portability, and ease of promote. Skew correction part of the first separately for each region to select a threshold for binarization, select the largest three text contour traversal, get the coordinates of the four corner points after the calculation of the transformed location of the corner points with the transformation matrix, and further use the perspective transformation to achieve the skew correction. The surface modification part finds 6 feature points creates a TPS deformer and inputs the mapping points corresponding to the feature points into the lists. The two lists are reshaped to the shape required by OpenCV and then the transformation matrix M is further estimated and the surface correction is done using perspective transformations.