Research On Recognition Algorithm In The Automatic Lost And Found System Of Smart Cards

Smart cards is making for an important role in people's lives, hence,the loss and retrieval of smart cards are drawing more attention from society nowadays. The current retrieval of smart cards relies heavily on cumbersome and time-consuming manual operations and lacks efficient management. This paper designs and implements an automated system for the retrieval of smart cards based on imaging analysis and machine learning.The retrieval system consists of the following parts: the collection box,the server-side, and the client-side(mobile phone APP). The collection box is designed to collect and transfer image data of smart cards; the identification program on server-side identifies the information of the smart card and records it into the database; people could query card information and provide real-time feedback via the mobile phone APP to achieve an automatic retrieval.This paper set up the hardware platform of the smart card retrieval system; implemented the identification algorithm on server-side; and built the overall web framework, the database, and android API using Apache,Python, and Django.This paper mainly studies the identification algorithm on the server:The images are preprocessed using imaging smoothing and angle correction. For different types of smart cards, appropriate methods are used to achieve classification. Next, we designed algorithm for background removal, digit locating, and digit segmentation. We also established the digital data set for digit recognition. Experiment results showed a 99.7%accuracy can be reached using identification algorithm for digit recognition.This paper implements an automated system for the retrieval of smart cards based on imaging analysis and machine learning, which combines card information collection, ID recognition, and data communication, and greatly improves retrieval efficiency.The innovations of the paper are as follows:(1) Came up with a detection method based on Fast algorithm and image convolution. Greatly improved the accuracy of target-region locating.(2) Designed a background removal algorithm based on edge probability density. Combined local binary classification and morphological processing for detecting digit location in a complex background.(3) Established the digit data set containing digits of different fonts under different lumen, position, or noise conditions, which outperforms the others as better representing real-life card digit features.