Design And Implementation Of Digit Recognition Software Based On Android Platform

In recent years,with the expansion of Android market,the demand of its application market also diversifies increasingly;the emergence of new hotspots,such as mobile payment and Vehicle License Plate Recognition,make the requirements of digital identification technology getting higher.The problems faced by digital identification applications include the simultaneous to improve the recognition speed and recognition rate is difficult,the recognition rate in extreme environments and the identification of damaged numbers is low and the majority of applications do not support the recognition of damaged digital numbers.In order to solve the problems above,this thesis designs and implements a digital recognition software based on Android platform.The algorithm is the core part of the digital recognition software.Firstly,the existing digital recognition algorithm are studied in the theisi and their advantages and disadvantages are anlaysed.According to the design requirements,the Convolution neural network and Hopfield neural network are selected by the thesis to deal with the conventional digital identification and damaged numbers identification separately.And the compiled the model in to Dynamic-link library for Android application.Secondly,the development of software is completed from three parts:image preprocessing,image recognition and data storage.In the image preprocessing module,the system overhead is reduced by bit operation.In the image recognition module,the identify process is completed by the neural network model;in the data store module,the results is stored by SQLite.The results of inspection and test show that the digital recognition software realized in this paper has reached the intended design goal,and the functions can be operated correctly and stably.For the ideal situation,the numerical recognition rate is higher than the design target,reaching 95.83%,and the recognition speed is controlled within 1 second;for the non-ideal state digital recognition,it also reached 75.29%recognition rate;and for the residual digital recognition,the 58.69%.