Research On QR Code Automatic Recognition Algorithm In Complex Imaging Environment

In the era of informationization and automation,QR codes have been widely used in production and life by virtue of their advantages of large storage space,good fault tolerance and strong security,and have brought many conveniences to people’s lives in modern society.The increasingly diversified and complicated application environment has also caused various difficulties for the automatic identification of QR codes.Aiming at the recognition problem of QR code in the case of motion blur and plane distortion,the performance of the current algorithm is not satisfactory,and deep learning provides another feasible solution for the above problems.Therefore,this article focuses on the recognition of motion blur plane distortion QR code,and based on deep learning and traditional digital image processing technology,the development of related automatic recognition algorithms is carried out.The main research work is as follows:Aiming at the problem of deblurring the QR code with motion blur plane distortion,a related image deblurring model is built based on the generation of confrontation network,and related improvements are made to the Deblur GAN model.The main improvement measures are: adding a residual path to the residual block of the generator and adding a loss to control the low-frequency information of the image closer to the content loss function of the generator.The simulation results show that the related improvements to the Deblur GAN model in this article are effective,which can further improve the deblurring effect of the model while ensuring the efficiency of the model.Aiming at the problem of detecting the distorted QR code in the image,this paper builds a related target detection model,focusing on related improvements to the YOLOv3 model.The main improvement measures are: lightweight model based on depth separable convolution,a priori box size optimization based on K-means clustering algorithm,and cross-scale upsampling feature fusion optimization for multi-scale prediction networks.The simulation resul ts show that the related improvements in this article for the YOLOv3 model are effective,which can further improve the detection efficiency of the model while ensuring the positioning accuracy of the model.Aiming at the problem of location and correction of QR code with plane distortion,this paper designs a set of algorithm flow.The algorithm first performs semantic segmentation on the distorted QR code image based on the Seg Net network,then uses the Canny algorithm to perform edge detection on the sem antic segmentation map,and then performs line detection on the edge detection map based on the Hough transform and uses the computational geometry method to obtain the distorted QR code quadrilateral The coordinates of the four corners of the region are t hen corrected based on the anti-perspective transformation and bilinear interpolation to complete the correction of the distortion QR code and the pixel value filling,and finally the corrected QR code is binarized by the OTSU algorithm.The simulation res ults show that the planar distortion QR code positioning and correction algorithm designed in this paper can well complete the positioning and correction of the distortion QR code in a variety of application scenarios,while ensuring that the corrected QR code can be based on a variety of current commercial software Such as: We Chat,Weibo and Alipay can quickly scan and read.