Research On The Recognition Technology Of QR Code In Complex Background Motio

Two-dimensional codes can be used in product traceability,production management and product anti-counterfeiting in industrial production.However,most industrial production environments have features of uneven illumination,cluttered background,and fast product movement.It is easy to affect the rapid and accurate identification of the Two-dimensional code,resulting in misreading and missed identification,which affects the use effect and production efficiency.How to identify two-dimensional codes stably and efficiently in an industrial environment is important.Taking QR code as an example,this paper conducts an in-depth study on the recognition technology of QR code according to complex background and motion.This paper analyzes the components of complex background from illumination,color,background texture,and combines the features of motion blur and deformation to establish a QR code dataset based on complex background.Analyze and compare three deep learning target detection algorithms and focus on improving Yolo V4 from convolution kernel replacement and network structure adjustment,and propose a QR code recognition algorithm based on Yolo V4,which is verified by the server,the experiment shows that the improvement effect is good.In image restoration,combined with gradient grayscale prior and regularization method,the motion blur restoration of QR code is carried out.L₀regularization constraint term with gradient as the main and gray as the auxiliary is constructed,and the blur kernel and the restored image are solved alternately.Then through morphological processing,Hough line detection and corner positioning,perspective transformation,the linear deformation QR code is deformed and restored.Experiments show that the restorative QR code recognition rate is greatly improved.Finally,a test device composed of camera and embedded GPU platform is constructed to study the transplantation and optimization of QR code recognition algorithm,and test the effectiveness of the algorithm.The results show that the comprehensive performance of the QR code recognition algorithm based on the Tensor RT framework is the best;by locating and recognizing multiple sets of complex background motion QR codes,and analyzing the detection results and experimental data,the average recognition accuracy can reach 90%and above,and the average recognition accuracy can reach 90%or more.FPS is around 24.The algorithm studied in this paper has achieved good results in the recognition of complex background motion QR codes,and the accuracy and real-time performance can meet the requirements.