Handwritten Digits Recognition And The Design Of Test Paper Management System Based On CNN

Handwritten digits recognition is a hot topic in the field of character recognition currently,and the digital processing technology has become an important research content in the field of information technology.In the modern teaching process,the automatic identification and storage of student test paper scores are important parts of the standardized construction of school teaching management.Automatic recognition of test paper scores can effectively reduce the workload of teachers and improve work efficiency.This thesis mainly focuses on three aspects of automatic recognition for test paper scores including test paper image processing,handwritten digit recognition and the design of test paper automatic recognition system.1.Test paper image processing is the basis of automatic recognition of test paper scores.This thesis implements test paper segmentation,image enhancement,tilt correction,handwritten digit(score)segmentation,and final normalization.The experimental results show that the preprocessing and segmentation of the test paper images can improve the accuracy of handwritten digit recognition;2.A CNN-based handwritten digit classification method is achieved.At the same time,for the recognition of error-prone or altered digits on the test paper images,or the unclear digits,the post-processing method using expertise on previous output is proposed to improve the recognition accuracy;3.With the application of Flask and Caffe frameworks based on Python and the use of Qt Creator development tools,this thesis designs and implements an automatic recognition system for test paper scores based on convolutional neural network,which can directly display score statistics.