Research And Implementation Of QR Code Image Restoration And Fast Recognition Method

The emergence of two-dimensional bar code has promoted the development of the Internet of Things and communication technology greatly, and it has become an important information carrier of the machine reading. QR Code, which is a widely-used two-dimensional bar code, not only possesses the common characteristics of 2D barcodes of high information density, large capacity, strong error correction capability, being able to be read from 360 degrees, but also have its own advantages, such as ultra-high-speed reading, being able to deal with Chinese and Japanese effectively. However, the current QR code recognition technology has some shortcomings:(1) previous QR code binary segmentation work mainly relates to improving the recognition rate of QR code in uneven illumination condition. There is few work considering fast binarization method to meet the need of fast recognition in industrial production and embedded applications; (2) in the process of inclination correction, the computational complexity of Hough Transform algorithm is relatively high; (3) the geometric distorted QR code images have lower recognition rate.In order to solve the above problems, we study the binary segmentation algorithms of QR code and adopt improved FCM algorithm (FFCM) to binarize QR code image, and use multi-constraint Hough transform algorithm to reduce the computational complexity of calculating the inclination angle and utilizing QR code characteristics, use mathematical morphology algorithm based on self-adapting structural element to obtain the geometric distorted vertexes accurately and revise the distortion. The specific design method includes following three points:1. Fuzzy C-means(FCM) algorithm is a classic data segmentation algorithm in unsupervised learning field which can be used to aggregate similar characteristics of the sample. Utilizing the advantage of FCM algorithm, in this paper, we adopt improved FCM algorithm for binary segmentation of QR code image. Through data reduction, FCM would have better performance in QR code binarization which can not only assure the recognition rate, but also shorten the recognition time effectively.2. The computational complexity of Hough Transform in the calculation of inclination angle of QR code is high, but inclination angle constraint and edge-point constraint would reduce the computational complexity and improve the performance of the algorithm greatly. In this paper, utilizing the characteristics of QR code, we adopt multi-constraint approach. Firstly we constrain the transform angle in HT domain based on Freeman guidelines; secondly we analyze three structural features of straight line edge in the digital image and use primitive element to replace the target edge points to constrain the contour points. Our multi-constraint method can improve the speed of the calculation of QR code inclination angle effectively.3. The mathematical morphology algorithm based on adaptive structural element can obtain the distorted vertexes of QR code more precisely. Firstly run-length histogram is used to calculate the module size of QR code. Then the module size is used as the parameter of structural element in the mathematical morphology process to locate the distorted vertexes and finally revise the distortion. The algorithm can locate the distorted vertexes precisely and improve the accuracy of QR code image restoration and recognition rate.Based on the research results above, we build a QR code recognition system which has achieved the whole process of QR code image recognition. Having been tested, the results show that our system can recognize QR code image in practical application rapidly and accurately. It means our system can provide valuable application for embedded system and mobile terminal device.