基于CNN网络的手写体数字识别系统的实现

Implementation of handwritten digit recognition system based on CNN network

手写体数字识别现在仍是图像识别分类的一个热点,而基于卷积神经网络的深度学习算法具有局部区域连接、权值共享、降采样的结构特点,使得卷积神经网络在图像处理领域有出色表现。以实现手写体数字高精度识别为目标,设计并实现一个基于卷积神经网络的高精度手写体数字识别系统。首先,通过Pyqt5平台设计一个人机交互的GUI界面,其次进行手写体数字图像的采集与预处理,变换成规范的三维向量输入到CNN网络卷积层中,接着进行各个网络层的运算处理,最后通过Softmax输出分类结果。仿真实验结果下MNIST数据集识别模式下的识别率为99.9%,手写输入识别模式下的识别率为98%。结果表明:基于CNN的神经网络识别准确率高,实现技术简单,实用性高。

Handwritten digit recognition is still a hot spot in image recognition and classification,and the deep learning algorithm based on convolutional neural network has the structural characteristics of local area connection,weight sharing and downsampling,which makes convolutional neural network perform well in the field of image processing.Aiming at realizing high-precision handwritten digit recognition,a high-precision handwritten digit recognition system based on convolutional neural network is designed and implemented.Firstly,use the Pyqt5 platform to build a GUI interface for human-computer interaction;secondly,carry out the collection and preprocessing of handwritten digital images,which are converted into standardized three-dimensional vectors and input into the convolution layer of the CNN network,and then start the operation processing of each network layer;finally,the classification results are output through Softmax.The simulation results show that the recognition rate in the MNIST dataset recognition mode is 99.9%,and the recognition rate in the handwriting input recognition mode is 98%.The results show that the neural network based on CNN has high recognition accuracy,simple implementation technology and high practicability.