QR Code Recognition In Complex Acquisition Environment

Today,smart mobile devices are widely used,and the 2D barcode has entered the public's view more and more frequently.So that people's lives and work are inseparable from the support of 2D barcodes.QR code is currently the most popular and widely used type of 2D barcodes.Thanks to the convenience of QR codes and the improvement of QR codes identification technology,the QR code developed quickly from industrial production demand to every corner of the public life.From the transmission of logistics information to mobile payment,and from the verification of bill information to the dissemination of multimedia information,QR codes have brought a tremendous transformation to the way people live and work.While the QR code is widely used,it also brings with a identification problem.A variety of application environments and image acquisition devices increase the complexity of the QR code recognition process.And the conventional QR code identification methods gradually fail to meet people's needs.In order to improve the robustness and adaptability of the QR code identification method in various complex acquisition environments,and stimulate the development potential of QR codes,we make an in-depth study of QR code positioning and image restoration technology.The main contents of this thesis are as follows:(1)By studying the QR code application environment and acquisition conditions,the QR code's complex collection environment was analyzed qualitatively and quantitatively.We created an image data set of QR codes with the most robust and complex collection environment,which was used for the training and testing of the QR code positioning method.(2)We proposed a QR code positioning method based on SSD model.The complete QR code image area is used as a training sample.The K-means algorithm is used to perform dimensional clustering on the data set.And we used the clustering results to optimize the SSD target detection model to achieve the positioning of the QR code in a complex environment.The experimental results show that our QR code positioning method has achieved good positioning performance.(3)We deblurred the motion blurred QR code image that cannot be correctly decoded.Bilateral filtering algorithm is used to denoise the blurred QR code image.Then we create the intensity and gradient prior model of the blurred QR code image,and use a deblurring algorithm based on L0 regularization prior to realize the recovery process of the motion blurred QR code image.Through the statistics of the correct decoding rate of the QR code before and after the image deblurring,the effectiveness of our QR code image deblurring algorithm is verified.(4)We proposed a QR code distortion correction algorithm based on morphological operations and Hough lines detection.By analyzing the encoding process and texture features of the QR code,we connected the QR code areas by morphological operations.Then we obtain the straight lines and corners of the QR code area by Hough lines detection.Thereby the automatic restoration of the distorted QR code image is realized.Through the statistics of the correct QR decoding rate before and after the distortion restoration,the effectiveness of our QR code distortion recovery algorithm is verified.