Research And Implementation Of The Technology Of Image Restoration And Identification Of QR Code

Two-dimensional bar code recognition is a new automatic identification technology generated with the development of information technology. As a type of two-dimensional bar code, QR code has the following features: high density, small footprint, rapid recognition and highly efficient coding of Chinese characters. It has been widely used in various fields. How to quickly and efficiently identify QR Code has become one of the research hot spots. The focus of this thesis is how to recover QR code image tilt, distortion, etc. caused by a variety of reasons, and to ensure the accuracy of QR code image decoding.Through the combination of theoretical analysis and simulation experiment, this thesis is mainly a study of the technology of QR code image restoration and recognition. The main work and points of innovation are as follows:(1) In the image preprocessing stage, for there are noises in the QR code images acquired through hardware devices, noise reduction has been done based on the principle of median-filter. According to the binaryzation feature of QR code images, and through a comparative analysis of surface fitting algorithm, Ostu algorithm and adaptive brightness algorithm, adaptive brightness algorithm has been selected to solve the complicated problems such as uneven illumination in QR code images, which helps the preliminary recognition of QR code image to achieve the desired effect.(2) By finding the position detection graphs of QR codes and with the method of Hough transform, this thesis has solved the problem of code image location and rotation. Besides, a modified method of Hough transform based on edge detection has been put forward to effectively reduce the computation and improve the efficiency of QR code image rotation correction.(3) As devices collect QR codes images from different perspectives, which results in the problem of image trapezoidal distortion, the thesis has put forward related solutions. First, the boundary point is scanned to set the control point, and then with the help of projection transformation of the geometric plane and gray-scale interpolation algorithm, which is done through bilinear interpolation, the trapezoidal distortion correction is achieved.(4) As there are also surface distortion problems caused by the curved surface of QR code images, the thesis has advanced a method to reconstruct QR code images, which is based on the variation of image gray scale. First, the 3D images of the collected QR code images is converted into the corresponding 2D images; then deformation caused by perspective projection is resumed along the X axis and that caused by orthogonal projection is resumed along the Y axis, and rotation orientation is done as the last step. With the above-mentioned method, we can correct the curved surface distortion of QR code images.By integrating the above-mentioned algorithms, the thesis has designed a prototype system of QR code identification, and has achieved the whole process of QR code image identification.