基于CNN的手写中文数字识别研究

Handwritten Chinese Digits Recognition Based CNN

为提高手写中文数字识别准确率,提出了一种基于TensorFlow平台的卷积神经网络模型.该模型使用反阈值二值化算法对图像进行预处理,以提高运算速度;通过调整卷积神经网络中的超参,以优化该模型的识别性能;使用多分类交叉熵损失函数对模型的损失函数进行度量,以验证模型的有效性.仿真分析结果显示,该模型识别手写中文数字的整体准确率达到99.58%、召回率达到99.53%,因此该模型在手写中文数字自动识别场景中具有良好的应用价值.

To improve the accuracy of handwritten Chinese digits recognition,a convolutional neural network model based on TensorFlow platform is proposed.In the proposed model,the inverse threshold binarization algorithm is applied to preprocess the image to improve the computing speed,and the hyperparameters in convolutional neural network are adjusted to optimize the performance of the model.The multi-classification cross entropy loss function is used to measure the loss function of the model to verify the validity of the model.The simulation results show that the overall accuracy rate and recall rate of handwritten Chinese digits recognition by this model reach 99.58%and 99.53%respectively,so the model has a good application value in the handwritten Chinese digits automatic recognition scenes.