Development Of A Micro-featured Password Recognition System Based On Machine Vision

Nowadays,with the increasing competition of commodities,fake and shoddy commodities emerge in an endless stream.Fake and shoddy commodities not only affect the sales volume of the company's commodities,but also affect the health of consumption.Therefore,how to identify the fake and shoddy commodities and how to reduce the ready-made fake and shoddy commodities are hot topics.This project relies on the horizontal project of the enterprise.According to the way of hiding the password at the bottom of the bottle,a micro feature password recognition system based on machine vision is designed.By means of specially designed hardware device and software algorithm,the anti-counterfeiting code hidden in the micro feature graph(point size between 50 and 200 microns)of laser marking is recognized.First of all,this paper introduces the principle of hiding micro anti-counterfeiting password in production date,and introduces the theoretical basis of image recognition,which provides a reference for the next research of micro password recognition algorithm.Then,according to the sample detection method given by the enterprise,the hardware of the password recognition system is built,including camera,light source,fine-tuning platform,bracket and acquisition box shell.Then,according to the way of hiding password,image recognition is carried out.The algorithm steps include the location of production date,digital cutting and recognition,the application of watershed algorithm to segment the image,sorting the segmentation points,and finally the anti-counterfeiting secret code is obtained by sorting algorithm.At the same time,considering the problem of image scratch and the increase of data in the later stage,the neural network password recognition is preliminarily explored and compared with the traditional algorithm results of this system.Secondly,considering the requirements of software design and the independence and confidentiality of software application,the function of software module is designed,the software is packaged,and the software installation and use process are described based on MATLAB.Finally,based on the measured data of the enterprise,it is shown that the effective recognition accuracy of the system is as high as 99.5% after removing the invalid marking samples,and the software operation is simple and stable,which meets the requirements of the enterprise production encryption.