It is difficult to meet the engineering requirements of curved plates through a single curved surface processing due to the strong nonlinearity of the forming of the ship curved plate. Therefore, it is necessary to evaluate the intermediate shape of the processed curved plate, in order to obtain the forming deviation of the curved plate and thus improve the automatic forming of the curved plate. A multi-directional scanning point cloud merging method is proposed based on the measured point cloud data. The point clouds of the curved plates scanned at different positions can be accurately registered for best fit between different point clouds. The processed scanning surface point cloud is positioned with the target surface. The vertical, longitudinal and transverse shape deviations of the processed curved plate are obtained through the proposed surface forming deviation calculation method. The simulation experiment is then carried out to find that the calculation accuracy of the shape deviation of the curved plate meets the engineering requirements of the ship curved plates. An experiment is designed to measure a processed sail plate to verify the practicability of the method. The research results provide a processing method for the uneven distribution of the measured point cloud data and a calculation method for the forming deviation of the processed curved plate in different directions.