实现图像的预处理，包括灰度值转换，二值化，梯度锐化，去离散噪声，整体倾斜调整，字符分割，尺寸归一，紧缩重排等功能，经过神经网络训练后可以字符识别。-image preprocessing, including gray-scale conversion value, the value of the gradient sharpening to discrete noise, overall tilt adjustment, character segmentation, to a size and tightening rearrangements, and other functions, the neural network can be trained to identify characters.