根据SURF特征，在标准印铁图像中选择了若干特定的子区域，进而在检测印件图中选择与其大致对应的子区域，并提取了其中的SURF特征。将所提取的特征和标准图中相应区域中的SURF特征进行匹配，根据匹配结果，利用最小二乘法获得了2幅图的变换参数。随机选择一定数量的匹配像素点，计算了各自8邻域RGB平均值，并转换到了L*a*b*颜色空间，根据L*a*b*色差可判断印铁图像是否合格。实验结果表明了该方法的有效性，同时判别速度满足实际需要。

Several special sub-areas were selected according to the SURF feature of the standard image printed on tin slice, and the roughly similar ones in the image to be checked were selected. The SURF features in image to be checked were extracted and matched with those in the sub-areas in standard image. Then according to matched features the transformation parameters between these two images were acquired with least squares method. A certain number of matched pixel pairs were chosen randomly and the RGB averages of their 8 neighborhoods were computed and transferred into L*a*b* model. Then the L*a*b* chromatic aberrations were applied to judge the qualification of tin printing. Experimental results showed that the method is valid and the judgment speed can meet practical requirement.